Predictive Markers of Psychosis

Abstract

The invention relates to a method of determining the likelihood of an individual transitioning to a first episode of psychosis (FEP), the method comprising determining the level of selected markers in a bodily fluid sample from the individual, wherein the increase or decrease in the markers is predictive of the individual transitioning to a first episode of psychosis (FEP). The invention also relates to a method of predicting the functional outcome for an individual following a first episode of psychosis (FEP), the method comprising determining the level of selected markers in a bodily fluid sample from the individual, wherein the increase or decrease in the markers is predictive of an increased risk of functional disability outcome for the individual.

Claims

1. A method of determining the likelihood of an individual transitioning to a first episode of psychosis (FEP), the method comprising: determining the level of markers in a bodily fluid sample from the individual, wherein the markers are selected from one or more proteins of Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, Phospholipid transfer protein, C4b-binding protein alpha chain, Complement component 8 alpha chain, Vitamin K-dependent protein S, Ficolin-3, Transthyretin, Complement component 6, Retinol-binding protein 4, Beta-crystallin B2, Vitamin D binding protein, Inter-alpha-trypsin inhibitor heavy chain H1, Plasma protease C1 inhibitor, Alpha-2-antiplasmin, Fibulin-1, Clusterin, L-lactate dehydrogenase B chain, Extracellular matrix protein 1, disintegrin and metalloproteinase with thrombospondin motifs 13, Complement C1q subcomponent subunit C, and Alpha-crystallin A chain, coagulation factor XII, Carboxypeptidase N subunit 2, Complement C1s subcomponent, Alpha 1 anti-chymotrypsin, Plasminogen, Monocyte differentiation antigen CD14, Zinc alpha-2-glycoprotein, Attractin, Complement Factor I, Immunoglobulin lambda constant 3, Ceruloplasmin Antithrombin III, and N-acetylmuramoyl-L-alanine amidase, wherein an increase in the level of one or more markers selected from Complement component 8 alpha chain, Complement component 6, Retinol-binding protein 4, Beta-crystallin B2, Vitamin D binding protein, Inter-alpha-trypsin inhibitor heavy chain H1, Fibulin-1, Clusterin, L-lactate dehydrogenase B chain, Complement C1q subcomponent subunit C, and Alpha-crystallin A chain, coagulation factor XII, Carboxypeptidase N subunit 2, Alpha 1 anti-chymotrypsin, Plasminogen, Monocyte differentiation antigen CD14, Attractin, Zinc alpha-2-glycoprotein, Extracellular matrix protein 1, Complement C1s subcomponent, Ceruloplasmin, Antithrombin III and Complement Factor I; and/or a decrease in the level of one or more markers selected from Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, Phospholipid transfer protein, C4b-binding protein alpha chain, Vitamin K-dependent protein S, Ficolin-3, Transthyretin, Plasma protease C1 inhibitor, Alpha-2-antiplasmin, A disintegrin and metalloproteinase with thrombospondin motifs 13, Immunoglobulin lambda constant 3, and N-acetylmuramoyl-L-alanine amidase; is predictive of the individual transitioning to a first episode of psychosis (FEP).

2. The method according to claim 1, wherein the individual is an ultra-high risk (UHR) individual for psychosis.

3. The method according to claim 1 or claim 2, further comprising the assessment of clinical features.

4. The method according to any preceding claim, further comprising selecting the individual for therapeutic intervention and/or a follow-up check, if the individual is predicted to transition to a first episode of psychosis (FEP) and/or develop a functional disability.

5. The method according to any preceding claim, further comprising administering a therapeutic or preventative medication to the individual, if the individual is predicted to transition to a first episode of psychosis (FEP).

6. A method of predicting the functional outcome for an individual following a first episode of psychosis (FEP), the method comprising: determining the level of markers in a bodily fluid sample from the individual, wherein the markers are selected from one or more proteins of Alpha-2-macroglobulin, Phospholipid transfer protein, Immunoglobulin heavy constant mu, Fetuin-B, CD5 antigen-like, Pyruvate kinase, Inter-alpha-trypsin inhibitor heavy chain H1, Clusterin, Complement factor H, Pigment epithelium-derived factor, Insulin-like growth factor-binding protein 3, Histidine-rich glycoprotein, Galectin-3-binding protein, and Mannose-binding protein C, wherein an increase in the level of one or more markers selected from Fetuin-B, Inter-alpha-trypsin inhibitor heavy chain H1, Clusterin, Complement factor H, Pigment epithelium-derived factor, Insulin-like growth factor-binding protein 3, Histidine-rich glycoprotein, Galectin-3-binding protein, and Mannose-binding protein C; and/or a decrease in the level of one or more markers selected from Alpha-2-macroglobulin, Phospholipid transfer protein, Immunoglobulin heavy constant mu, CD5 antigen-like, and Pyruvate kinase, is predictive of an increased risk of functional disability outcome for the individual.

7. The method according to claim 6, further comprising the assessment of clinical features.

8. The method according to claim 7, wherein the method comprises the further assessment of one or more, or all, of the clinical features selected from BPRS: suspiciousness, SANS: impersistence at work or school, SANS: increased latency of response, SANS: blocking, SANS: grooming and hygiene, BPRS: excitement, SANS: sexual activity, and MADRS: suicidal thoughts.

9. The method according to any one of claims 6-8, further comprising selecting the individual for therapeutic intervention and/or a follow-up check, if the individual is predicted to develop a functional disability.

10. The method according to any one of claims 6-9, further comprising administering a therapeutic or preventative medication to the individual, if the individual is predicted to develop a functional disability.

11. The method according to any preceding claim, wherein determining the level of a marker comprises conducting an enzyme-linked immunosorbent assay (ELISA) to determine the level of one or more markers in the sample or by a Proximity Extension Assay (PEA).

12. The method according to any preceding claim, wherein the markers are detected by probes that are immobilised on a substrate.

13. A composition comprising a plurality (e.g. two or more) of probes capable of binding to protein markers in a bodily fluid sample, wherein the protein markers comprise two or more proteins selected from the group comprising Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, Phospholipid transfer protein, C4b-binding protein alpha chain, Complement component 8 alpha chain, Vitamin K-dependent protein S, Ficolin-3, Transthyretin, Complement component 6, Retinol-binding protein 4, Beta-crystallin B2, Vitamin D binding protein, Inter-alpha-trypsin inhibitor heavy chain H1, Plasma protease C1 inhibitor, Alpha-2-antiplasmin, Fibulin-1, Clusterin, L-lactate dehydrogenase B chain, Extracellular matrix protein 1, disintegrin and metalloproteinase with thrombospondin motifs 13, Complement C1q subcomponent subunit C, and Alpha-crystallin A chain, coagulation factor XII, Carboxypeptidase N subunit 2, Complement C1s subcomponent, Alpha 1 anti-chymotrypsin, Plasminogen, Monocyte differentiation antigen CD14, Zinc alpha-2-glycoprotein, Attractin, Complement Factor I, Immunoglobulin lambda constant 3, Ceruloplasmin, Antithrombin III and N-acetylmuramoyl-L-alanine amidase, Immunoglobulin heavy constant mu, Fetuin-B, CD5 antigen-like, Pyruvate kinase, Complement factor H, Pigment epithelium-derived factor, Insulin-like growth factor-binding protein 3, Histidine-rich glycoprotein, Galectin-3-binding protein, and Mannose-binding protein C.

14. The composition according to claim 13, wherein the composition comprises a plurality of probes capable of binding to protein markers in a bodily fluid sample, wherein the protein markers comprise two or more proteins selected from the group comprising Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, Phospholipid transfer protein, C4b-binding protein alpha chain, Complement component 8 alpha chain, Vitamin K-dependent protein S, Ficolin-3, Transthyretin, Complement component 6, Retinol-binding protein 4, Beta-crystallin B2, Vitamin D binding protein, Inter-alpha-trypsin inhibitor heavy chain H1, Plasma protease C1 inhibitor, Alpha-2-antiplasmin, Fibulin-1, Clusterin, L-lactate dehydrogenase B chain, Extracellular matrix protein 1, disintegrin and metalloproteinase with thrombospondin motifs 13, Complement C1q subcomponent subunit C, and Alpha-crystallin A chain, coagulation factor XII, Carboxypeptidase N subunit 2, Complement C1s subcomponent, Alpha 1 anti-chymotrypsin, Plasminogen, Monocyte differentiation antigen CD14, Zinc alpha-2-glycoprotein, Attractin, Complement Factor I, Immunoglobulin lambda constant 3, Ceruloplasmin, Antithrombin III and N-acetylmuramoyl-L-alanine amidase.

15. The composition according to claim 13 or 14, wherein a plurality of probes are provided for binding to one or more, or all, of the protein markers selected from Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, C4b-binding protein alpha chain, Phospholipid transfer protein, Transthyretin, Vitamin D binding protein, Beta-crystallin B2, Vitamin K-dependent protein S, Coagulation factor XII and clusterin; or wherein a plurality of probes are provided for binding to one or more, or all, of the protein markers selected from alpha-2-macroglobulin (A2M), immunoglobulin heavy constant mu (IGHM), C4b-binding protein alpha chain (C4BPA), vitamin K-dependent protein S, fibulin-1, transthyretin, N-acetylmuramoyl-L-alanine amidase, vitamin D-binding protein, clusterin and complement component 6 (C6); or wherein a plurality of probes are provided for binding to one or more, or all, of the protein markers selected from alpha-2-macroglobulin, Immunoglobulin heavy constant mu, C4b-binding protein alpha chain, complement component 8 alpha chain, Phospholipid transfer protein, ficolin-3, vitamin D binding protein, vitamin K-dependent protein S, beta-crystallin B2, and transthyretin.

16. The composition according to claim 13, wherein a plurality of probes are provided for binding to protein markers in a bodily fluid sample, wherein the protein markers comprise two or more proteins selected from the group comprising Alpha-2-macroglobulin, Phospholipid transfer protein, Immunoglobulin heavy constant mu, Fetuin-B, CD5 antigen-like, Pyruvate kinase, Inter-alpha-trypsin inhibitor heavy chain H1, Clusterin, Complement factor H, Pigment epithelium-derived factor, Insulin-like growth factor-binding protein 3, Histidine-rich glycoprotein, Galectin-3-binding protein, and Mannose-binding protein C.

17. The composition according to claim 16, wherein a plurality of probes are provided for binding to protein markers in a bodily fluid sample, wherein the protein markers comprise two or more proteins selected from the group comprising Alpha-2-macroglobulin, Phospholipid transfer protein, Immunoglobulin heavy constant mu, Fetuin-B, CD5 antigen-like, Pyruvate kinase, Inter-alpha-trypsin inhibitor heavy chain H1, Clusterin, Complement factor H, and Pigment epithelium-derived factor.

18. The composition according to any of claims 13-17, wherein the probes are provided as a panel of probes anchored to a surface.

19. A method of detecting the level of two or more proteins selected from the group comprising Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, Phospholipid transfer protein, C4b-binding protein alpha chain, Complement component 8 alpha chain, Vitamin K-dependent protein S, Ficolin-3, Transthyretin, Complement component 6, Retinol-binding protein 4, Beta-crystallin B2, Vitamin D binding protein, Inter-alpha-trypsin inhibitor heavy chain H1, Plasma protease C1 inhibitor, Alpha-2-antiplasmin, Fibulin-1, Clusterin, L-lactate dehydrogenase B chain, Extracellular matrix protein 1, disintegrin and metalloproteinase with thrombospondin motifs 13, Complement C1q subcomponent subunit C, and Alpha-crystallin A chain, coagulation factor XII, Carboxypeptidase N subunit 2, Complement C1s subcomponent, Alpha 1 anti-chymotrypsin, Plasminogen, Monocyte differentiation antigen CD14, Zinc alpha-2-glycoprotein, Attractin, Complement Factor I, Immunoglobulin lambda constant 3, Ceruloplasmin, Antithrombin III and N-acetylmuramoyl-L-alanine amidase, in a bodily fluid sample; and optionally the bodily fluid sample may be of an ultra-high risk (UHR) individual for psychosis.

20. A method of detecting the level of two or more proteins selected from the group comprising Alpha-2-macroglobulin, Phospholipid transfer protein, Immunoglobulin heavy constant mu, Fetuin-B, CD5 antigen-like, Pyruvate kinase, Inter-alpha-trypsin inhibitor heavy chain H1, Clusterin, Complement factor H, Pigment epithelium-derived factor, Insulin-like growth factor-binding protein 3, Histidine-rich glycoprotein, Galectin-3-binding protein, and Mannose-binding protein C, in a bodily fluid sample from an individual; and optionally the bodily fluid sample may be of a ultra-high risk (UHR) individual for psychosis.

21. A method for treating an individual to prevent a transition to a FEP, the method comprising the steps of: determining whether the individual is predicted to transition to a FEP by: obtaining or having obtained a sample from the individual; and performing or having performed the method according to any one of claims 1-12 to determine if the individual is predicted to transition to a FEP; and if the individual is predicted to transition to a FEP, then administering medication as described herein to the individual.

22. A method for treating an individual to prevent a transition to a FEP, the method comprising the steps of: receiving results of a test performed according to the method of any one of claims 1-12 to determine if the individual is predicted to transition to a FEP; and if the individual is predicted to transition to a FEP, then administering medication as described herein to the individual.

23. A method for treating an individual to prevent a functional disability following a FEP, the method comprising the steps of: determining whether the individual is predicted to develop a functional disability by: obtaining or having obtained a sample from the individual; and performing or having performed the method according to any one of claims 1-12 to determine if the individual is predicted to develop a functional disability; and if the individual is predicted to develop a functional disability, then administering medication as described herein to the individual.

24. A method for treating an individual to prevent the development of a functional disability following a FEP, the method comprising the steps of: receiving results of a test performed according to the method according to any one of claims 1-12 to determine if the individual is predicted to develop a functional disability; and if the individual is predicted to develop a functional disability, then administering medication as described herein to the individual.

25. Use of one or more proteins as a predictive biomarker for an individual to transition to a FEP, wherein the protein is selected from the group comprising Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, Phospholipid transfer protein, C4b-binding protein alpha chain, Complement component 8 alpha chain, Vitamin K-dependent protein S, Ficolin-3, Transthyretin, Complement component 6, Retinol-binding protein 4, Beta-crystallin B2, Vitamin D binding protein, Inter-alpha-trypsin inhibitor heavy chain H1, Plasma protease C1 inhibitor, Alpha-2-antiplasmin, Fibulin-1, Clusterin, L-lactate dehydrogenase B chain, Extracellular matrix protein 1, disintegrin and metalloproteinase with thrombospondin motifs 13, Complement C1q subcomponent subunit C, and Alpha-crystallin A chain, coagulation factor XII, Carboxypeptidase N subunit 2, Complement C1s subcomponent, Alpha 1 anti-chymotrypsin, Plasminogen, Monocyte differentiation antigen CD14, Zinc alpha-2-glycoprotein, Attractin, Complement Factor I, Immunoglobulin lambda constant 3, Ceruloplasmin, Antithrombin III and N-acetylmuramoyl-L-alanine amidase.

26. Use of one or more proteins as a predictive biomarker for predicting the likelihood of a functional disability outcome for an individual following a FEP, wherein the protein is selected from the group comprising: Alpha-2-macroglobulin, Phospholipid transfer protein, Immunoglobulin heavy constant mu, Fetuin-B, CD5 antigen-like, Pyruvate kinase, Inter-alpha-trypsin inhibitor heavy chain H1, Clusterin, Complement factor H, Pigment epithelium-derived factor, Insulin-like growth factor-binding protein 3, Histidine-rich glycoprotein, Galectin-3-binding protein, and Mannose-binding protein C.

27. The use according to any of claim 25 or 26, wherein the use is determining the levels of the protein(s) in a bodily fluid sample from the individual, optionally wherein the individual is an UHR individual for psychosis.

Description

[0194] Embodiments of the invention will now be described in more detail, by way of example only, with reference to the accompanying drawings.

[0195] FIG. 1: Class prediction for Model 1a stratified by EU-GEI site.

[0196] FIG. 2: Receiver-operating characteristic curve for Model 1a.

[0197] FIG. 3: STRING network analysis of protein-protein interactions for differentially expressed proteins (following FDR adjustment) between CHR-T and CHR-NT in EU-GEI.

[0198] The network nodes are proteins (proteins implicated in the complement and coagulation cascades are highlighted in red (marked with *)). The edges represent functional associations between proteins, and the colour of each edge represents the source of evidence for that association, including fusion evidence; co-occurrence evidence; experimental evidence; and text mining evidence.

[0199] FIG. 4: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 1b. T: ultra high-risk participants who transitioned to first episode psychosis; NT: ultra high-risk participants who did not transition.

[0200] FIG. 5: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 1c.: ultra high-risk participants who transitioned to first episode psychosis; NT: ultra high-risk participants who did not transition.

[0201] FIG. 6: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 1d in training dataset (all sites except London): ultra high-risk participants who transitioned to first episode psychosis; NT: ultra high-risk participants who did not transition.

[0202] FIG. 7: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 2. T: ultra high-risk participants who transitioned to first episode psychosis; NT: ultra high-risk participants who did not transition.

[0203] FIG. 8: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 3. Poor: General Assessment of Functioning disability score ≤60 at 2 years; Good: General Assessment of Functioning disability score >60 at 2 years.

[0204] FIG. 9: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 4. T: ultra high-risk participants who transitioned to first episode psychosis; NT: ultra high-risk participants who did not transition.

[0205] FIG. 10: Mean algorithm scores (A) and receiver-operating characteristic curve (B) for Model 5. PE: Definite psychotic experiences at 18; No PE: No psychotic experiences at 18.

[0206] FIG. 11a: mean algorithm scores for Model 11a. FIG. 11b: Receiver-operating characteristic curve for Model 1a.

[0207] FIG. 12: Mean algorithm scores and class predictions (A) and receiver-operating characteristic curve (B) for Model 1b: clinical data. T: clinical high-risk participants who transitioned to first episode psychosis; NT: clinical high-risk participants who did not transition.

[0208] FIG. 13: Mean algorithm scores and class predictions (A) and receiver-operating characteristic curve (B) for Model 1c: proteomic data. T: clinical high-risk participants who transitioned to first episode psychosis; NT: clinical high-risk participants who did not transition.

[0209] FIG. 14: Mean algorithm scores and class predictions (A) and receiver-operating characteristic curve (B) for Model 2a: proteomic (non-London). T: clinical high-risk participants who transitioned to first episode psychosis; NT: clinical high-risk participants who did not transition.

[0210] FIG. 15: Mean algorithm scores and class predictions (A) and receiver-operating characteristic curve (B) for Model 2b: parsimonious (10-predictor) proteomic model, training data (non-London).

[0211] FIG. 16: Mean algorithm scores and class predictions (A) and receiver-operating characteristic curve (B) for Model 3: replication. T: clinical high-risk participants who transitioned to first episode psychosis; NT: clinical high-risk participants who did not transition.

EXAMPLE 1—DEVELOPMENT OF PROTEOMIC PREDICTION MODELS FOR OUTCOMES IN THE CLINICAL HIGH RISK STATE AND PSYCHOTIC EXPERIENCES IN ADOLESCENCE: MACHINE LEARNING ANALYSES IN TWO NESTED CASE-CONTROL STUDIES

Summary

[0212] Background: Biomarkers for prediction of outcomes in people at risk of psychosis would inform the clinical management of this group.
Methods: We conducted two nested case-control studies. The first study was nested within the European Network of National Schizophrenia Networks Studying Gene-Environment Interactions (EU-GEI) and comprised 133 clinical high-risk (CHR) participants, of whom 49 transitioned to psychosis. The second study was nested within the Avon Longitudinal Study of Parents and Children (ALSPAC) and comprised 121 participants who did not report psychotic experiences (PEs) at age 12, of whom 55 later reported PEs at age 18. Baseline plasma samples in EU-GEI and age 12 plasma samples in ALSPAC were analysed using mass spectrometry. Support vector machine algorithms were used to develop models for prediction of transition and functional outcome in EU-GEI, and PEs at age 18 in ALSPAC.

[0213] Outcomes: In the CHR sample, using 65 clinical and 166 proteomic features a model demonstrated excellent performance for prediction of transition status (area under the receiver-operating curve [AUC] 0.96, positive predictive value [PPV] 81.8%, negative predictive value [NPV] 94.9%). A model based on the ten most predictive proteins accurately predicted transition status in training (AUC 0.97, PPV 84.8%, NPV 95.7%) and withheld data (AUC 0.93, PPV 80.0%, NPV 90.9%). A model using the same 65 clinical and 166 proteomic features predicted functional outcome with AUC 0.72 (PPV 67.6%, NPV 47.6%). In the general population sample, a model using 265 proteomic features predicted psychotic experiences at age 18 with AUC 0.76 (PPV 69.1%, NPV 74.2%).

[0214] Interpretation: Proteomic markers may contribute to prediction of outcomes in individuals at risk of psychosis.

[0215] Funding: Health Research Board Clinician Scientist Award to DRC; European Community's Seventh Framework Programme (EU-GEI); UK Medical Research Council, Wellcome Trust (ALSPAC).

Introduction

[0216] Early detection of people with psychotic disorders may improve their clinical outcomes..sup.1 There has been a focus on the clinical high-risk (CHR) state.sup.2 with the aim of identifying vulnerable individuals and offering clinical interventions..sup.3, 4 16-35% of CHR individuals develop first-episode psychosis (FEP) within 3 years.sup.5, 6 and the CHR state is often associated with co-morbid depressive and anxiety disorders.sup.7, 8 as well as functional impairment..sup.9, 10 Studies have also characterised an ‘extended psychosis phenotype’ which includes individuals with psychotic experiences (PEs),.sup.11 psychotic symptoms that may occur in the general population or at less severe points on the psychosis continuum, with or without help-seeking. PEs are associated with increased risk of psychotic and non-psychotic disorders,.sup.12 suicidal behaviour.sup.13 and reduced functioning..sup.14, 15

[0217] Biomarkers may aid prediction of outcomes for at-risk individuals..sup.16 Blood-based studies associate inflammatory and immune-related processes with development of psychosis..sup.17-19 This is supported by proteomic studies in schizophrenia implicating the acute phase response, glucocorticoid receptor signalling, coagulation and lipid metabolism..sup.20, 21 Proteomic studies in those who develop PEs provide evidence for early dysregulation of the complement system,.sup.22, 23 which has been implicated in schizophrenia.sup.24, 25 and other mental disorders..sup.26

[0218] We aimed to apply proteomic methods to assess differential protein expression in CHR individuals who do and do not develop psychosis, and to develop predictive models for transition and functional outcome. We employed similar methods for the broader phenotype of PEs in a general population sample. For predictive modelling, we used machine learning techniques such as have been used for prediction of functional outcome in CHR.sup.27 and increasingly in psychiatry..sup.28,29

Methods

Study 1: CHR Sample

Participants and Study Design

[0219] The European network of national schizophrenia networks studying Gene-Environment Interactions (EU-GEI) is a collaborative project studying gene-environment interactions in schizophrenia. Work Package 5 comprises a prospective study of CHR individuals followed for up to 6 years, across 11 sites in Europe, Australia and Brazil..sup.30,31, 32 Participants with CHR symptoms were referred by their local mental health service and were eligible to participate if they met CHR criteria as defined by the Comprehensive Assessment of At-Risk Mental States.sup.33 (CAARMS). Exclusion criteria were: current or past psychotic disorder; symptoms explained by a medical disorder or drug or alcohol use; IQ<60. Plasma samples were obtained at baseline and participants followed clinically for up to six years, with clinical assessments performed at baseline, 12 and 24 months. Accrual began in September 2010 and the last baseline assessment was performed in July 2015. The present study was a nested case-control study of participants who provided plasma samples at baseline, comparing samples from CHR participants who transitioned to psychosis on follow-up (CHR-T, n=49) with a control group of randomly-selected participants who did not (CHR-NT, n=84).

Outcomes

[0220] Transition status: Transition was defined as the onset of non-organic psychotic disorder as determined by the CAARMS. Assessors were not systematically blinded to transition status since, in some cases, clinical services had contacted the research team to advise that transition had occurred. For participants who developed psychosis after 24 months, transition status was determined by contacting the clinical team or from clinical records. Functional outcome: We used the General Assessment of Functioning (GAF).sup.34, 35 disability subscale, recorded at follow-up assessment closest to two years from baseline. For use as a classification target variable (and in line with previous approaches.sup.36) the score was dichotomised into ‘poor functioning’ (≤60 points; i.e. moderate or severe impairment) or ‘good functioning’ (>60 points; i.e. mild or no impairment).

Clinical Measures

[0221] Baseline clinical data included sociodemographic data, the GAF subscales for symptoms and disability,.sup.34, 35 the Scale for Assessment of Negative Symptoms (SANS),.sup.37 the Brief Psychiatric Rating Scale (BPRS).sup.38 and the Montgomery-Asberg Depression Rating Scale (MADRS)..sup.39

Sample Preparation

[0222] Protein digestion and peptide purification was performed as previously described.sup.40. Laboratory staff were blind to case/control status.

Proteomic Analysis

[0223] We used discovery-based proteomic methods, namely data-dependent acquisition (DDA), as described previously.sup.22. Briefly, 5 μl from each sample was injected on a Thermo Scientific Q-Exactive mass spectrometer, connected to a Dionex Ultimate 3000 (RSLCnano) chromatography system, and operated in DDA mode for label-free liquid chromatography mass spectrometry..sup.22, 23, 40-42

Enzyme-Linked Immunosorbent Assay (ELISA) Validation

[0224] We assessed nine proteins in plasma samples from the same CHR-T and CHR-NT participants using ELISA.

Replication

[0225] In a partial replication of the mass spectrometry experiment, we analysed samples from mostly the same group of CHR-T participants (2 of the 49 cases were different, otherwise the CHR-T participants were the same) and an entirely different group of CHR-NT participants (n=86).

Study 2: General Population Sample

Participants and Study Design

[0226] The Avon Longitudinal Study of Parents and Children (ALSPAC) is a prospective birth cohort study..sup.43, 44 Pregnant women resident in Avon, UK with expected dates of delivery between 1 Apr. 1991 to 31 Dec. 1992 were invited to participate with a total sample of over 15,000 pregnancies. The study website contains details of available data (http://www.bristol.ac.uk/alspac/researchers/our-data/). We previously analysed age 12 plasma samples from young people who did and did not report PEs at age 18,.sup.22, 23 finding several differentially expressed proteins in a data-independent acquisition (DIA) analysis focused on proteins of the complement pathway. In the present study we performed DDA analyses rather than DIA to achieve broader proteome coverage.

Outcome

[0227] Psychotic experiences: PEs were assessed at 12 and 18 years using the semi-structured Psychosis-Like Symptom Interview.sup.11 and rated as not present, suspected or definitely psychotic. Cases were participants who did not report PEs at age 12 but reported at least one definite PE at age 18. Controls were randomly selected age-matched participants who did not report PEs at age 12 nor 18.

Sample Preparation

[0228] Age 12 plasma samples were prepared for mass spectrometry as previously described..sup.22

Bioinformatics and Statistical Analysis

Demographic and Clinical Data

[0229] Baseline demographic and clinical data were tested for differences using the 2-sided t-test for continuous and χ.sup.2 test for categorical variables in SPSS v.25 (Armonk, N.Y., USA) with α=0.05. In Study 1, baseline clinical data comprised 65 variables including the GAF subscales for symptoms and disability, and total and individual item scores for the SANS, BPRS and MADRS. In Study 2, data were obtained for sex, ethnicity, maternal social class and body mass index (BMI) at age 12. Missing clinical data were replaced using the mean (for continuous) or mode (for categorical variables).

Proteomic Data

[0230] Label free quantification (LFQ) was performed in Max Quant (v.1.5.2.8).sup.45, 46 as described..sup.40 Proteins that were identified with at least two peptides (one uniquely assignable to the protein) and quantified in >80% of samples were taken forward for quantification. LFQ values were log 2-transformed and missing values imputed using the imputeLCMD package v.2.0.sup.47 in RStudio (Boston, Mass., USA; http://www.rstudio.com/). Values were converted to z-scores and winsorised within ±3z.

Differential Expression

[0231] To determine differential expression, analysis of covariance (ANCOVA) was performed in Stata 15 (College Station, Tex., US) comparing mean LFQ value in cases and controls for each identified protein. In Study 1, covariates were age, sex, BMI and years in education. In Study 2, covariates were sex, maternal social class and age 12 BMI. P-values were corrected for multiple comparisons using the Benjamini-Hochberg procedure.sup.48 with false discovery rate (FDR) of 5%.

Support Vector Machine (SVM) Models

[0232] We used the open-source machine learning software Neurominer v.1.0 (https://www.pronia.eu/neurominer/) for MatLab 2018a (MathWorks Inc, USA) to generate predictive models, with area under the receiver-operating curve (AUC) as the performance criterion for evaluation and optimisation. Continuous variables were converted to z-scores for feature scaling and winsorised within ±3z. Random-label permutation analysis.sup.49-51 with 250 permutations was used to verify predictive models against null models and derive p-values for model significance.

SVM Model 1

[0233] We used an L2-regularised SVM algorithm to develop a classification model predicting transition outcome. We incorporated geographical generalisability using repeated nested cross-validation with a leave-site-out approach in the outer loop as previously described..sup.50 We used the LIBLINEAR program with L2 regularisation to attenuate risk of over-fitting.sup.52 whereby weightings of non-predictive features are minimised, but not reduced to zero (thus more closely modelling the biological effects of functionally inter-related proteins). Given the unbalanced group sizes, the hyperplane was weighted (increasing the misclassification penalty in the minority class) which reduces the risk of bias.′ A priori covariates were age, sex, BMI and years in education.

SVM Model 2

[0234] Concentrations derived from ELISA were used as features in an L2-regularised SVM algorithm with cross-validation and covariates as for Model 1.

SVM Model 3

[0235] We used an L2-regularised SVM algorithm to derive a classification model predicting functional outcome at 2 years: poor (GAF≤60) vs. good (GAF>60) functioning. Features and covariates were as for Model 1. Compared to transition status, fewer participants had data available for functional outcome (n=79). Therefore, this model used five-fold repeated nested cross-validation with five inner and five outer folds, irrespective of study site.

SVM Model 4

[0236] We developed an L2-regularised SVM model for prediction of transition status using the clinical and proteomic features in the replication dataset. This model used leave-site-out repeated nested cross-validation as for Model 1. In addition to age, sex, BMI and years in education, we also adjusted for ethnicity and tobacco use due to evidence of baseline differences for these characteristics.

SVM Model 5

[0237] We developed an L2-regularised SVM model predicting PEs at age 18 in ALSPAC using DDA proteomic data at age 12. Repeated nested cross-validation with five inner and five outer folds was used to derive the model. Sex, maternal social class and age 12 BMI were covariates.

Results

Study 1: EU-GEI (CHR Sample)

Sample Characteristics

[0238] The EU-GEI cohort included 344 CHR participants, of whom 65 (18.9%) developed psychosis on follow-up. 57 transitioned within two years as defined by the CAARMS. In the eight who transitioned after two years, transition status was determined by contact with the clinical team or from clinical records.

[0239] Our subsample comprised 49 CHR-T and 84 CHR-NT participants. Characteristics of included and non-included participants are compared in Table 5. At baseline, included participants had higher mean total SANS composite and global scores and total BPRS score compared to non-included participants, but were otherwise comparable.

[0240] Among included participants, there was evidence that the mean baseline total BPRS score was higher in included CHR-T compared to CHR-NT. There was no evidence of differences between the groups on other symptom measures, socio-demographic features, cannabis use or medication use (Table 1). The median duration from baseline to transition was 219 days (interquartile range 424 days).

Differential Expression

[0241] Of 345 proteins identified, 166 were quantified in >80% of samples. There was nominally significant (p<0.05) differential expression for 56 proteins in CHR-T vs. CHR-NT, of which 35 remained significant after 5% FDR adjustment (Table 6. Proteins with FDR-adjusted p<0.001 (and associated mean fold change in CHR-T vs. CHR-NT) included: alpha-2-macroglobulin (Alpha-2-macroglobulin, 0.33), immunoglobulin heavy constant mu (Immunoglobulin heavy constant mu, 0.41), complement C8 alpha chain (Complement component 8 alpha chain, 1.48), vitamin D binding protein (VTDB, 1.43), complement C1q subcomponent subunit C (Complement C1q subcomponent subunit C, 1.53), plasminogen (1.29), clusterin (1.29), fibulin-1 (1.52), phospholipid transfer protein (Phospholipid transfer protein, 0.67), complement C1r subcomponent (1.27). FIG. 3 shows results of a STRING database.sup.54 protein interaction network analysis for differentially expressed proteins. The topmost implicated pathway was the complement and coagulation cascade (Table 7).

SVM Model 1

[0242] An SVM classification model predicted transition status based on 65 clinical and 166 proteomic features (Model 1a) with AUC 0.96 (p<0.004). Further performance metrics are presented in Table 2. FIG. 1 shows class prediction stratified by site and FIG. 2 the receiver-operating characteristic curve. Table 3 lists the 10% highest-weighted features based on mean feature weighting across all models selected in the inner loop.

[0243] We examined the predictive value of the clinical and the proteomic data separately by generating models based on each dataset individually. The clinical model (Model 1b) poorly predicted transition outcome (AUC 0.47, p=0.7; Table 2 and FIG. 4). The proteomic model (Model 1c) demonstrated excellent predictive performance (AUC 0.97, p<0.004; Table 2 and FIG. 5).

[0244] We next sought to develop a more parsimonious model based on a subset of 10 predictive features, and to test such a model in unseen data. As the largest site, we chose London to be the test site. To derive the ten highest-weighted proteins, an L2-regularised SVM model was trained using the proteomic data from all sites except London (CHR-T n=30, CHR-NT n=50), with leave-site-out cross-validation and adjustment for the same covariates as for Model 1a. The resulting AUC was 0.94, p<0.004 (sensitivity 86.7%, specificity 94.0%, balanced accuracy 90.3%, PPV 89.7%, NPV 92.2%, positive likelihood ratio 14.4, negative likelihood ratio 0.1). The ten highest-weighted features were: Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, C4b-binding protein alpha chain, Phospholipid transfer protein, transthyretin, VTDB, vitamin K dependent protein S (PROS), Beta-crystallin B2, coagulation factor XII and clusterin. A reduced model using these ten features (Model 1d) was trained on data from all sites except London with AUC 0.97, p<0.004 (Table 2 and FIG. 6). This model predicted transition status in the withheld London sample (CHR-T n=19, CHR-NT n=34) with AUC 0.93 (Table 2).

ELISA Validation

[0245] Three of the nine proteins assessed by ELISA showed statistically significant mean differences between CHR-T and CHR-NT (Alpha-2-macroglobulin p=0.0016, C1r p=0.0084, plasminogen p=0.0196; Table 8).

SVM Model 2

[0246] Complete ELISA data were available for 126 participants (CHR-T n=44, CHR-NT n=82). This model predicted transition status with AUC 0.76, p<0.004 (Table 2; FIG. 7).

SVM Model 3

[0247] In the 79 participants with outcome data available, this model predicted 2-year functional outcome with AUC 0.72, p=0.008 (Table 2; FIG. 8). The 10% highest-weighted features are listed in Table 3.

SVM Model 4

[0248] Sample characteristics of the replication dataset are described in Tables 9 and 10. Of 485 proteins identified, 119 were quantified in >80% of samples. In ANCOVA adjusted for age, sex, BMI, years in education, tobacco use and ethnicity, 82 proteins showed nominally significant differential expression (p<0.05) in CHR-T vs. CHR-NT of which 78 remained significant following 5% FDR adjustment (Table 11).

[0249] Due to differences in protein identifications, it was not possible to apply Models 1a-d in the replication dataset. We generated a new L2-regularised SVM model (adjusted for age, sex, BMI, years in education, tobacco use and ethnicity) using these 119 proteomic features and the same 65 clinical features as for Model 1a. This model demonstrated excellent performance for prediction of transition status (AUC 0.98, p<0.004; Table 2 and FIG. 9). The highest-weighted 10% of features are listed in Table 12. Proteins among the highest-weighted 10% of features in both Model 1a and Model 4, and weighted in similar directions, included Alpha-2-macroglobulin, Immunoglobulin heavy constant mu, clusterin, C4b-binding protein alpha chain and complement component 6 (Complement component 6).

Further Models

[0250] A model based only on the top 5 proteins when adjusting for age, sex, BMI and years in education (comprising A2M, IGHM, C4BPA, PLTP, transthyretin) was trained on the non-London samples (n=80) and predicted transition status with AUC 0.95 and balanced accuracy 94.3% (PPV 87.9%, NPV 97.9%). When tested on the withheld London samples (n=53) this model predicted transition status with AUC 0.92 and balanced accuracy 91.8% (PPV 89.5%, NPV 94.1%).

[0251] Another model based only on the top 5 proteins unadjusted for covariates (comprising A2M, IGHM, C4BPA, vitamin K dependent protein S, fibulin-1) was trained on the non-London samples (n=80) and predicted transition status with AUC 0.97 and balanced accuracy 86.3% (PPV 70.7%, NPV 97.4%). When tested on the withheld London samples (n=53) this model predicted transition status with balanced accuracy 90.1% (PPV 84.2%, NPV 93.9%).

[0252] A model based only on the top 2 proteins when adjusting for age, sex, BMI and years in education (comprising A2M and IGHM) was trained on the non-London samples (n=80) and predicted transition status with AUC 0.94 and balanced accuracy 86.0% (PPV 75.0%, NPV 93.2%). When tested on the withheld London samples (n=53) this model predicted transition status with AUC 0.91 and balanced accuracy 87.4% (PPV 77.3%, NPV 93.5%).

Study 2: ALSPAC (General Population Sample)

Sample Characteristics

[0253] The total sample comprised 65 cases and 67 controls. Eleven samples were excluded due to poor protein identification profiles, resulting in 55 cases and 66 controls. There was evidence that cases were more likely to be female, but no evidence of differences in ethnicity, maternal social class or age 12 BMI (Table 13).

Differential Expression

[0254] Of 506 proteins identified, 265 were quantified in >80% of samples. There was nominally significant (p<0.05) differential expression of 40 proteins at age 12 (Table 14) of which five remained significant after 5% FDR adjustment (mean fold change in cases vs. controls): C4b-binding protein alpha chain (0.77), serum paraoxonase/arylesterase 1 (0.80), Immunoglobulin heavy constant mu (0.78), inhibin beta chain (1.31) and clusterin (0.92).

SVM Model 5

[0255] An SVM model based on proteomic features from age 12 plasma samples predicted PE status at age 18 with AUC 0.76, p<0.004 (Table 2 and FIG. 10). The 10% highest-weighted features are listed in Table 3.

Discussion

[0256] We report evidence of differential expression of multiple proteins at baseline between CHR individuals who developed psychosis compared to those who did not, with particular implication of the complement and coagulation cascade. We used machine learning algorithms incorporating clinical and proteomic data to predict transition (AUC 0.96). Proteomic features were of greater predictive value than the included clinical features. A reduced model using data for 10 highly predictive proteins showed excellent predictive performance in training (AUC 0.97) and testing on withheld data (AUC 0.93). We also developed models for prediction of functional outcome in the same CHR population (AUC 0.72) and for prediction of PEs in a longitudinal birth cohort (AUC 0.76). Our results have clinical and aetiopathogenic implications.

[0257] Although a minority of CHR individuals transition to FEP,.sup.5, 6 the CHR state is a strong risk indicator for psychosis.′ Accurate identification of those at greatest risk of transition would facilitate targeting preventative interventions. Models based on clinical data have previously shown some value for prediction of transition and functional outcome in CHR..sup.56-60 A previous study combined CHR criteria with data on cognitive disturbances, achieving AUC of 0.81 for prediction of transition..sup.61 Accuracy has been further augmented using neuroimaging.sup.36, 62-64 and neurocognitive.sup.65 data. However, blood-based tests have the advantage of greater accessibility. A previous investigation found a panel of 15 proteins using immunoassays that distinguished between CHR individuals who did and did not transition with AUC 0.88..sup.66 A further study used blood-based biomarkers to predict onset of schizophrenia with AUC 0.82, which improved to 0.90 when the CAARMS positive symptoms subscale was included in the model..sup.19

[0258] We developed a parsimonious model using data from 10 highly predictive proteins which accurately predicted transition outcome. With further validation, these markers may contribute to individualised prognostic scores and stratification strategies to improve risk estimation.′ Notably, the models for transition that incorporated proteomic data had high sensitivity. In implementation, this would require balancing against the costs of unnecessary treatment of false positive individuals. However, interventions in CHR generally have a psychosocial focus.sup.68-71 and may have utility even in those who will not ultimately transition.

[0259] Regarding pathogenesis, our study provides the first mass spectrometry-derived evidence of differential protein expression associated with transition in CHR. We find particularly strong indication for the complement and coagulation cascades, which have previously been implicated in schizophrenia.sup.20, 24, 72-77 and preceding PEs..sup.22, 23 The primary causes of these changes remain unknown, but are consistent with evidence for raised inflammatory tone preceding psychosis and other mental disorders.sup.19, 66, 78-84 and the vulnerability associated with genetic variation of complement C4 in schizophrenia.′ In our study, several complement pathway proteins emerged as important predictors of transition including C4b-binding protein alpha chain, Complement C1q subcomponent subunit C, C1r of the antibody-antigen complex mediated pathway, key regulatory protease CFI, ficolin-3 and terminal pathway components Complement component 6 and Complement component 8 alpha chain. These proteins arise from common proteolytic pathways or interact with coagulation proteins plasminogen (positively associated with transition) and vitamin K-dependent protein S (negatively associated with transition), supporting hypotheses of activation of coagulation in psychosis..sup.75

[0260] The strongest predictor of transition was Alpha-2-macroglobulin (reduced in CHR-T), a protease inhibitor with diverse functions including inhibition of pro-inflammatory cytokines such as IL1β,.sup.85 which has been shown to be raised in FEP..sup.86 Alpha-2-macroglobulin is also a key coagulation inhibitor.sup.87, 88 and thus links functionally to our observations of elevated plasminogen in CHR-T. This is intriguing in light of evidence that blood-derived plasminogen drives brain inflammation.sup.89 and complement activation..sup.90 In models of multiple sclerosis, blood-brain barrier disruption facilitates transfer of fibrinogen into the brain where it is deposited as fibrin, causing local inflammation..sup.91 Our findings suggest a procoagulant phenotype in CHR-T and, given evidence for blood-brain barrier disruption in psychosis,.sup.92 the effects of fibrin provide possible aetiopathogenic mechanisms and novel therapeutic avenues,.sup.93 but this will require further confirmation.

[0261] We validated differential expression of Alpha-2-macroglobulin, C1r and plasminogen using ELISA. The ELISA-based SVM model demonstrated acceptable, though reduced, predictive accuracy. This may reflect the reduced sensitivity of ELISA and the inability to accurately quantify specific protein isoforms. It is intriguing that several proteins in the highest-weighted 10% of features in the original dataset were similarly highly weighted (and in similar directions) for prediction of PEs in the general population (C4b-binding protein alpha chain, vitamin K-dependent protein S, Alpha-2-macroglobulin and Immunoglobulin heavy constant mu). This could tentatively suggest a degree of similarity in certain proteomic changes between non-clinical youth who develop PEs and help-seeking CHR individuals who develop psychosis, but will require further investigation. More widely, our results are in keeping with studies in bipolar disorder and depression reporting reductions in Alpha-2-macroglobulin, IgM and C4b-binding protein alpha chain.′ Thus these changes may be in keeping with a general vulnerability to psychiatric disorder, outside of the psychosis spectrum.

[0262] Our study is not without its limitations. Firstly, the poor performance of the clinical model likely relates to the included features. It is probable that other clinical data (for example, individual CAARMS items) would lead to improved predictive ability. However, our primary aim was to investigate the role of proteomic predictors. Secondly, we could not access a similar sample in which we could test the external predictive ability of the models. Thirdly, our replication experiment was partial, since we were unable to access an entirely new set of CHR-T cases. Finally, it is possible that childhood adversity at least partially mediates the changes we observe,.sup.23, 95-97 but this will require further study.

[0263] In conclusion, we have developed models incorporating proteomic data to contribute to prediction of transition and functional outcome in CHR. In a longitudinal birth cohort, several of the same proteins also contributed to prediction of PEs. Further studies are required to validate these findings, evaluate their causes and elucidate amenable targets for prediction and prevention of psychosis.

Evidence in Context

[0264] Evidence Before this Study

[0265] Schmidt et al (2017) conducted a systematic review of studies (published until October 2015) that developed predictive models for onset of psychosis in people at clinical high risk (CHR) using clinical, biological, cognitive and environmental predictors or combinations thereof. To identify further studies published since this review, we performed a PubMed search using the following search terms: psycho* OR prodrom* OR “ultra high risk” OR “at risk mental state” OR “clinical high risk” AND prediction. The search was restricted to peer-reviewed studies published in English from October 2015 to November 2019. In their review, Schmidt et al identified 25 studies that generated models predicting onset of psychosis. Among biological models, the highest positive predictive value (83%) was achieved by a neuroimaging model (grey matter volume reduction on MRI). Schmidt et al also examined the role of sequential testing, finding that the highest positive predictive value was obtained using three models sequentially: a combined model (clinical plus electroencephalography), then structural MRI, followed by blood biomarkers. Since this review there have been several further studies describing predictive models for outcomes in CHR. These have incorporated data (or combinations of data) from several different modalities including socio-demographic, clinical, neuroimaging, linguistic and biological sources. With regard to prediction of development of psychosis, published models vary with regard to their predictive performance, with area under the receiver-operating curve typically in the range 0.60-0.90.

Added Value of this Study

[0266] In this study, we report the development of prediction models using support vector machine learning techniques based on proteomic data obtained from mass spectrometry of baseline plasma samples. We found that the clinical variables included in our study did not usefully predict development of psychosis. Proteomic data were highly predictive, and a model based on the 10 most predictive proteins performed well for prediction of transition outcome in training and test data. Proteomic data also contributed to prediction of functional outcome, though with less accuracy in comparison to transition outcome. Analysis of differentially expressed proteins provided particular evidence for implication of the complement and coagulation cascades. We also developed a prediction model based on proteomic data in a general population sample, with several of the same proteins weighted highly for prediction of outcomes in both samples.

Implications of all the Available Evidence

[0267] Proteomic features may helpfully contribute to outcome prediction in CHR individuals, and in particular for prediction of development of psychosis. However, the models we have developed require validation in external samples to assess their validity and applicability in the clinical setting.

TABLES

[0268]

TABLE-US-00001 TABLE 1 Descriptive statistics for EU-GEI CHR-T and CHR-NT groups Missing data, CHR-T CHR-NT n (%) N = 49 N = 84 t/χ.sup.2 p Baseline age 0 22.2 (5.0) 22.9 (4.2) −0.824 0.412 in years, mean (SD) Sex, n (%) 0 26 male (53.1%) 42 male (50.0%) 0.116 0.733 23 female (46.9%) 42 female (50.0%) Baseline BMI 20 (15.0%) 24.5 (4.5) 24.4 (6.1) 0.116 0.908 in kg/m.sup.2, mean (SD) Baseline years 14 (10.5%) 14.1 (3.4) 14.4 (3.0) −0.625 0.533 in education, mean (SD) Ethnicity, 0 33 white (67.3%) 58 white (69.0%) 2.370 0.306 n (%) 8 black (16.3%) 7 black (8.3%) 8 other (16.3%) 19 other (22.6%) Ever used 3 (2.3%) 36 yes (73.5%) 65 yes (77.4%) 0.051 0.821 cannabis, 11 no (22.4%) 18 no (21.4%) n (%) 2 not known (4.1%) 1 not known (1.2%) Baseline 29 (21.8%) 15 yes (30.6%) 26 yes (31.0%) 0.030 0.862 cannabis 22 no (44.9%) 41 no (48.8%) use, n (%) 12 not known (24.5%) 17 not known (20.2%) Baseline 14 (10.5%) 21 yes (42.9%) 43 yes (51.2%) 0.373 0.541 tobacco 21 no (42.9%) 34 no (40.5%) use, n (%) 7 not known (14.2%) 7 not known (8.3%) Baseline 3 (2.3%) 35 yes (71.4%) 58 yes (69.0%) 0.071 0.790 alcohol use, 13 no (26.5%) 24 no (28.6%) n (%) 1 not known (2.0%) 2 not known (2.4%) Baseline 31 (23.3%) 19 yes (38.8%) 32 yes (38.1%) 0.042 0.839 medication Antidepressant 12 Antidepressant 17 use, n (%) Antipsychotic 5 Antipsychotic 4 Hypnotic 1 Hypnotic 6 Other 1 Other 5 20 no (40.8%) 31 no (36.9%) 10 not known (20.4%) 21 not known (25.0%) Baseline GAF 12 (9.0%) 52.4 (10.3) 56.0 (10.0) −1.906 0.059 symptoms score, mean (SD) Baseline GAF 5 (3.8%) 52.3 (12.4) 54.8 (11.3) −1.148 0.253 disability score, mean (SD) Baseline SANS 19 (14.3%) 20.9 (14.0) 16.2 (11.6) 1.903 0.060 total composite score, mean (SD) Baseline SANS 11 (8.3%) 6.6 (4.1) 5.8 (3.7) 1.158 0.249 total global score, mean (SD) Baseline BPRS 10 (7.5%) 49.1 (11.5) 44.2 (10.2) 2.452 0.016 total score, mean (SD) Baseline 7 (5.3%) 20.3 (10.4) 19.2 (9.2) 0.657 0.512 MADRS total score, mean (SD) 2 year GAF 62 (46.7%) 42.3 (13.2) 62.2 (10.3) −7.125  <0.001 symptoms score, mean (SD) .sup.a 2 year GAF 54 (40.6%) 44.7 (9.1) 64.5 (12.8) −8.024  <0.001 disability score, mean (SD) .sup.b 2 year GAF 54 (40.6%) 29 poor (59.2%) 18 poor (21.4%) 27.734  <0.001 disability functioning functioning score, 1 good (2.0%) 31 good (36.9%) dichotomous functioning functioning outcome .sup.c 19 not known (38.8%) 35 not known (41.7%) .sup.a Data available for 71 of 133 participants (CHR-NT n = 44, CHR-T n = 27) .sup.b Data available for 79 of 133 participants (CHR-NT n = 49, CHR-T n = 29) .sup.c Poor functioning: GAF disability score ≤60; good functioning: GAF disability score >60 Tobacco use was defined as daily use for at least 1 month over the previous 12 months. Alcohol use was defined as at least 12 or more alcoholic beverages over the previous 12 months. Missing data excluded in hypothesis tests.

[0269] EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions; CHR-T: clinical high risk, transitioned to psychosis; CHR-NT: clinical high risk, did not transition to psychosis; BMI: body mass index; GAF: General Assessment of Functioning; SANS: Scale for the Assessment of Negative Symptoms; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery Asberg Depression Rating Scale

TABLE-US-00002 TABLE 2 Performance metrics for support vector machine models Bal- True False True False anced AUC Positive Negative Positive Negative posi- nega- nega- posi- Sensi- Speci- accu- (95% predic- predic- likeli- likeli- tives, tives, tives, tives, tivity, ficity, racy, confidence tive tive hood hood, Model description n (%) n (%) n (%) n (%) % % % interval) value, % value, % ratio ratio Model 1a: clinical 45  4 74 10 91.8 88.1 90.0 0.96 81.8 94.9 7.7 0.1 and proteomic (92%)  (8%) (88%) (12%) (0.92-1.00) Dataset: EU-GEI Features: 65 clinical and 166 proteomic Target: transition status N: 49 transition, 84 non-transition Model 1b: clinical 22 27 40 44 44.9 47.6 46.3 0.47 33.3 59.7 0.9 1.2 Dataset: EU-GEI (45%) (55%) (48%) (52%) (0.37-0.57) Features: 65 clinical Target: transition status N: 49 transition, 84 non-transition Model 1c: proteomic 45  4 75  9 91.8 89.3 90.6 0.97 83.3 94.9 8.6 0.1 Dataset: EU-GEI (92%)  (8%) (89%) (11%) (0.94-1.00) Features: 166 proteomic Target: transition status N: 49 transition, 84 non-transition Model 1d: top 10, 28  2 45  5 93.3 90.0 91.7 0.97 84.8 95.7 9.3 0.1 training (93%)  (7%) (90%) (10%) (0.93-1.00) Dataset: EU-GEI, London data withheld Features: 10 proteomic Target: transition status N: 30 transition, 50 non-transition Model 1d: top 10, test 16  3 30  4 84.2 88.2 86.2 0.93 80.0 90.9 7.2 0.2 Dataset: EU-GEI, (84%) (16%) (88%) (12%) (0.85-1.00) London data Features: 10 proteomic Target: transition status N: 19 transition, 34 non-transition Model 2: ELISA 31 13 54 28 70.5 65.9 68.2 0.76 52.5 80.6 2.1 0.4 Dataset: EU-GEI (70%) (30%) (66%) (34%) (0.67-0.85) Features: 9 ELISA Target: transition status N: 44 transition, 82 non-transition Model 3: functional 25 22 20 12 53.2 62.5 57.8 0.72 67.6 47.6 1.4 0.7 outcome (53%) (47%) (63%) (37%) (0.61-0.83) Dataset: EU-GEI Features: 65 clinical and 166 proteomic Target: functional outcome N: 47 poor functioning, 32 good functioning Model 4: replication 47  2 82  4 95.9 95.3 95.6 0.98 92.2 97.6 20.6 <0.1 Dataset: EU-GEI (96%)  (4%) (95%)  (5%) (0.95-1.00) replication Features: 65 clinical and 119 proteomic Target: transition status N: 49 transition, 86 non-transition Model 5: ALSPAC 38 17 49 17 69.1 74.2 71.7 0.76 69.1 74.2 2.7 0.4 PEs (69%) (31%) (74%) (26%) (0.67-0.85) Dataset: ALSPAC Features: 265 proteomic Target: PEs age 18 N: 55 PEs, 66 no PE AUC: area under the receiver-operating characteristic curve; EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions; ALSPAC: Avon Longitudinal Study of Parents and Children; PEs: psychotic experiences. Models 1a-d, 2 and 3 are adjusted for age, sex, body mass index and years in education and Model 4 is additionally adjusted for ethnicity and tobacco use. Model 5 is adjusted for sex, maternal social class at birth and body mass index at age 12.

TABLE-US-00003 TABLE 3 Ten percent highest-weighted features for support vector machine models (ranked according to mean feature weight for models selected in cross-validation inner loop) Model 1a: EU-GEI transition Model 3: EU-GEI functional outcome Model 5: ALSPAC psychotic experiences Mean Mean Mean Feature weight Feature weight Feature weight P01023 Alpha-2-macroglobulin −0.339 P01023 Alpha-2- −0.180 P04003 C4b-binding protein −0.228 P01871 Immunoglobulin heavy −0.242 macroglobulin alpha chain constant mu BPRS: suspiciousness 0.179 P27169 Serum paraoxonase/ −0.182 P04003 C4b-binding protein −0.154 P55058 Phospholipid −0.178 arylesterase 1 alpha chain transfer protein P07225 Vitamin K-dependent −0.150 P55058 Phospholipid −0.153 P01871 Immunoglobulin −0.167 protein S transfer protein heavy constant mu Q03591 Complement factor −0.147 P07357 Complement component 0.152 O43866 CDS antigen-like −0.164 H-related protein 1 8 alpha chain Q9UGM5 Fetuin-B 0.147 P61626 Lysozyme C −0.145 P07225 Vitamin K-dependent- −0.139 P19827 Inter-alpha-trypsin 0.133 P55103 Inhibin beta C chain 0.132 protein S inhibitor heavy chain H1 Q08380 Galectin-3-binding 0.131 O75636 Ficolin-3 −0.139 P14618 Pyruvate kinase −0.128 protein P02766 Transthyretin −0.138 SANS: impersistence at 0.122 P01871 Immunoglobulin heavy −0.128 P13671 Complement component 6 0.128 work or school constant mu P02774 Vitamin D binding protein 0.120 SANS: increased latency 0.115 P01019 Angiotensinogen −0.123 P43320 Beta-crystallin B2 0.120 of response P24593 Insulin-like growth 0.122 P02753 Retinol-binding protein 4 0.120 SANS: blocking 0.110 factor-binding protein 5 P23142 Fibulin-1 0.113 P17936 insulin-like growth 0.102 P00746 Complement factor D 0.120 P10909 Clusterin 0.112 factor-binding protein 3 P09871 Complement C1s −0.115 P19827 Inter-alpha-trypsin 0.112 P10909 Clusterin 0.099 subcomponent inhibitor heavy chain H1 SANS: grooming and hygiene 0.099 P02654 Apolipoprotein C-I −0.113 P05155 Plasma protease −0.112 Q08380 Galectin-3- 0.096 O75636 Ficolin-3 0.113 C1 inhibitor binding protein Q9NQ79 Cartilage acidic protein 1 0.112 P08697 Alpha-2-antiplasmin −0.111 P36955 Pigment epithelium- 0.095 P01023 Alpha-2-macroglobulin −0.109 MADRS: concentration difficulties −0.104 derived factor P10909 Clusterin −0.108 P02747 Complement C1q 0.103 P01042 Kininogen-1 0.092 P04275 von Willebrand factor −0.107 subcomponent subunit C P60174 Triosephosphate −0.092 P07358 Complement component 0.101 P07195 L-lactate dehydrogenase 0.103 isomerase C8 beta chain B chain BPRS: excitement 0.091 Q5T7F0 Neuropilin −0.100 Q16610 Extracellular matrix −0.102 P02656 Apolipoprotein C-III 0.091 Q9H4A9 Dipeptidase 2 −0.100 protein 1 SANS: sexual activity −0.090 P02679 Fibrinogen gamma chain −0.100 P02489 Alpha-crystallin A chain 0.101 MADRS: suicidal thoughts −0.090 P24592 Insulin-like growth factor- 0.099 Q76LX8 A disintegrin and −0.100 P04275 von Willebrand factor 0.089 binding protein 6 metalloproteinase with O95497 Pantetheinase 0.098 thrombospondin motifs 13 P04040 Catalase 0.098 H0Y755 Low affinity 0.096 immunoglobulin gamma Fc region receptor III-A Proteins are presented with their Uniprot accession number and corresponding protein name. EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions; ALSPAC: Avon Longitudinal Study of Parents and Children; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery-Asberg Depression Rating Scale; SANS: Scale for the Assessment of Negative Symptoms

REFERENCES

[0270] 1. Larsen T K, Melle I, Auestad B, Haahr U, Joa I, Johannessen J O, Opjordsmoen S, Rund B R, Rossberg J I, Simonsen E, Vaglum P, Friis S, McGlashan T. Early detection of psychosis: positive effects on 5-year outcome. Psychol Med. 2011; 41(7):1461-1469. [0271] 2. Fusar-Poli P, Borgwardt S, Bechdolf A, Addington J, Riecher-Rössler A, Schultze-Lutter F, Keshavan M, Wood S, Ruhrmann S, Seidman Li, Valmaggia L, Cannon T, Velthorst E, De Haan L, Cornblatt B, Bonoldi I, Birchwood M, McGlashan T, Carpenter W, McGorry P, Klosterkötter J, McGuire P, Yung A. The psychosis high-risk state: a comprehensive state-of-the-art review. JAMA psychiatry. 2013; 70(1):107-120. [0272] 3. Schmidt S J, Schultze-Lutter F, Schimmelmann B G, Maric N P, Salokangas R K, Riecher-Rossler A, van der Gaag M, Meneghelli A, Nordentoft M, Marshall M, Morrison A, Raballo A, Klosterkotter J, Ruhrmann S. EPA guidance on the early intervention in clinical high risk states of psychoses. Eur Psychiatry. 2015; 30(3):388-404. [0273] 4. Schultze-Lutter F, Michel C, Schmidt Si, Schimmelmann B G, Maric N P, Salokangas R K, Riecher-Rossler A, van der Gaag M, Nordentoft M, Raballo A, Meneghelli A, Marshall M, Morrison A, Ruhrmann S, Klosterkotter J. EPA guidance on the early detection of clinical high risk states of psychoses. Eur Psychiatry. 2015; 30(3):405-416. [0274] 5. Fusar-Poli P, Bonoldi I, Yung A R, Borgwardt S, Kempton M J, Valmaggia L, Barale F, Caverzasi E, McGuire P. Predicting psychosis: meta-analysis of transition outcomes in individuals at high clinical risk. Arch Gen Psychiatry. 2012; 69(3):220-229. [0275] 6. Cannon T D, Yu C, Addington J, Bearden C E, Cadenhead K S, Cornblatt B A, Heinssen R, Jeffries C D, Mathalon D H, McGlashan T H, Perkins D O, Seidman U, Tsuang M T, Walker E F, Woods S W, Kattan M W. An Individualized Risk Calculator for Research in Prodromal Psychosis. Am J Psychiatry. 2016; 173(10):980-988. [0276] 7. Rutigliano G, Valmaggia L, Landi P, Frascarelli M, Cappucciati M, Sear V, Rocchetti M, De Micheli A, Jones C, Palombini E, McGuire P, Fusar-Poli P. Persistence or recurrence of non-psychotic comorbid mental disorders associated with 6-year poor functional outcomes in patients at ultra high risk for psychosis. J Affect Disord. 2016; 203:101-110. [0277] 8. Kelleher I, Keeley H, Corcoran P, Lynch F, Fitzpatrick C, Devlin N, Molloy C, Roddy S, Clarke M C, Harley M, Arseneault L, Wasserman C, Carli V, Sarchiapone M, Hoven C, Wasserman D, Cannon M. Clinicopathological significance of psychotic experiences in non-psychotic young people: evidence from four population-based studies. Br J Psychiatry. 2012; 201(1):26-32. [0278] 9. Fusar-Poli P, Rocchetti M, Sardella A, Avila A, Brandizzi M, Caverzasi E, Politi P, Ruhrmann S, McGuire P. Disorder, not just state of risk: meta-analysis of functioning and quality of life in people at high risk of psychosis. Br J Psychiatry. 2015; 207(3):198-206. [0279] 10. Addington J, Cornblatt B A, Cadenhead K S, Cannon T D, McGlashan T H, Perkins D O, Seidman U, Tsuang M T, Walker E F, Woods S W, Heinssen R. At clinical high risk for psychosis: outcome for nonconverters. Am J Psychiatry. 2011; 168(8):800-805. [0280] 11. Zammit S, Kounali D, Cannon M, David A S, Gunnell D, Heron J, Jones P B, Lewis S, Sullivan S, Wolke D, Lewis G. Psychotic experiences and psychotic disorders at age 18 in relation to psychotic experiences at age 12 in a longitudinal population-based cohort study. Am J Psychiatry. 2013; 170(7):742-750. [0281] 12. Healy C, Brannigan R, Dooley N, Coughlan H, Clarke M, Kelleher I, Cannon M. Childhood and adolescent psychotic experiences and risk of mental disorder: a systematic review and meta-analysis. Psychol Med. 2019; 49(10):1589-1599. [0282] 13. Yates K, Lang U, Cederlof M, Boland F, Taylor P, Cannon M, McNicholas F, DeVylder J, Kelleher I. Association of Psychotic Experiences With Subsequent Risk of Suicidal Ideation, Suicide Attempts, and Suicide Deaths: A Systematic Review and Meta-analysis of Longitudinal Population Studies. JAMA Psychiatry. 2019; 76(2):180-189. [0283] 14. Kelleher I, Wigman J T, Harley M, O'Hanlon E, Coughlan H, Rawdon C, Murphy J, Power E, Higgins N M, Cannon M. Psychotic experiences in the population: Association with functioning and mental distress. Schizophr Res. 2015; 165(1):9-14. [0284] 15. Healy C, Campbell D, Coughlan H, Clarke M, Kelleher I, Cannon M. Childhood psychotic experiences are associated with poorer global functioning throughout adolescence and into early adulthood. Acta Psychitr Scand. 2018; 138(1):26-34. [0285] 16. McGorry P, Keshavan M, Goldstone S, Amminger P, Allott K, Berk M, Lavoie S, Pantelis C, Yung A, Wood S, Hickie I. Biomarkers and clinical staging in psychiatry. World Psychiatry. 2014; 13(3):211-223. [0286] 17. Jeffries C D, Perkins D O, Fournier M, Do K Q, Cuenod M, Khadimallah I, Domenici E, Addington J, Bearden C E, Cadenhead K S, Cannon T D, Cornblatt B A, Mathalon D H, McGlashan T H, Seidman L I, Tsuang M, Walker E F, Woods S W. Networks of blood proteins in the neuroimmunology of schizophrenia. Transl Psychiatry. 2018; 8(1):112. [0287] 18. Upthegrove R, Manzanares-Teson N, Barnes N M. Cytokine function in medication-naive first episode psychosis: A systematic review and meta-analysis. Schizophrenia Research. 2014; 155(1-3):101-108. [0288] 19. Chan M K, Krebs M O, Cox D, Guest P C, Yolken R H, Rahmoune H, Rothermundt M, Steiner J, Leweke F M, van Beveren N J, Niebuhr D W, Weber N S, Cowan D N, Suarez-Pinilla P, Crespo-Facorro B, Mam-Lam-Fook C, Bourgin J, Wenstrup R J, Kaldate R R, Cooper J D, Bahn S. Development of a blood-based molecular biomarker test for identification of schizophrenia before disease onset. Transl Psychiatry. 2015; 5:e601. [0289] 20. Sabherwal S, English J A, Focking M, Cagney G, Cotter D R. Blood biomarker discovery in drug-free schizophrenia: the contribution of proteomics and multiplex immunoassays. Expert Rev Proteomics. 2016; 13(12):1141-1155. [0290] 21. Schmitt A, Martins-de-Souza D, Akbarian S, Cassoli J S, Ehrenreich H, Fischer A, Fonteh A, Gattaz W F, Gawlik M, Gerlach M, Grunblatt E, Halene T, Hasan A, Hashimoto K, Kim Y K, Kirchner S K, Kornhuber J, Kraus T F J, Malchow B, Nascimento J M, Rossner M, Schwarz M, Steiner J, Talib L, Thibaut F, Riederer P, Falkai P. Consensus paper of the WFSBP Task Force on Biological Markers: Criteria for biomarkers and endophenotypes of schizophrenia, part Ill: Molecular mechanisms. World J Biol Psychiatry. 2017; 18(5):330-356. [0291] 22. English J A, Lopez L M, O'Gorman A, Focking M, Hryniewiecka M, Scaife C, Sabherwal S, Wynne K, Dicker P, Rutten B P F, Lewis G, Zammit S, Cannon M, Cagney G, Cotter D R. Blood-Based Protein Changes in Childhood Are Associated With Increased Risk for Later Psychotic Disorder: Evidence From a Nested Case-Control Study of the ALSPAC Longitudinal Birth Cohort. Schizophr Bull. 2018; 44(4:297-306. [0292] 23. Ricking M, Sabherwal S, Cates H M, Scaife C, Dicker P, Hryniewiecka M, Wynne K, Rutten B P F, Lewis G, Cannon M, Nestler E J, Heurich M, Cagney G, Zammit S, Cotter D R. Complement pathway changes at age 12 are associated with psychotic experiences at age 18 in a longitudinal population-based study: evidence for a role of stress. Molecular Psychiatry. 2019. [0293] 24. Li Y, Zhou K, Zhang Z, Sun L, Yang J, Zhang M, Ji B, Tang K, Wei Z, He G, Gao L, Yang L, Wang P, Yang P, Feng G, He L, Wan C. Label-free quantitative proteomic analysis reveals dysfunction of complement pathway in peripheral blood of schizophrenia patients: evidence for the immune hypothesis of schizophrenia. Mol Biosyst. 2012; 8(10):2664-2671. [0294] 25. Sekar A, Bialas A R, de Rivera H, Davis A, Hammond T R, Kamitaki N, Tooley K, Presumey J, Baum M, Van Doren V, Genovese G, Rose S A, Handsaker R E, Daly M J, Carroll M C, Stevens B, McCarroll S A. Schizophrenia risk from complex variation of complement component 4. Nature. 2016; 530(7589):177-183. [0295] 26. Zhang C, Zhang D F, Wu Z G, Peng D H, Chen J, Ni J, Tang W, Xu L, Yao Y G, Fang Y R. Complement factor H and susceptibility to major depressive disorder in Han Chinese. Br J Psychiatry. 2016; 208(5):446-452. [0296] 27. Koutsouleris N, Kambeitz-llankovic L, Ruhrmann S, Rosen M, Ruef A, Dwyer D B, Paolini M, Chisholm K, Kambeitz J, Haidl T, Schmidt A, Gillam J, Schultze-Lutter F, Falkai P, Reiser M, Riecher-Rossler A, Upthegrove R, Hietala J, Salokangas R K R, Pantelis C, Meisenzahl E, Wood S J, Beque D, Brambilla P, Borgwardt S, Consortium P. Prediction Models of Functional Outcomes for Individuals in the Clinical High-Risk State for Psychosis or With Recent-Onset Depression: A Multimodal, Multisite Machine Learning Analysis. JAMA Psychiatry. 2018; 75(11):1156-1172. [0297] 28. Fusar-Poli P, Hijazi Z, Stahl D, Steyerberg E W. The science of prognosis in psychiatry: A review. JAMA Psychiatry. 2018; 75(12):1289-1297. [0298] 29. Hahn T, Nierenberg A A, Whitfield-Gabrieli S. Predictive analytics in mental health: applications, guidelines, challenges and perspectives. Molecular Psychiatry. 2016; 22:37. [0299] 30. Jongsma H E, Gayer-Anderson C, Lasalvia A, Quattrone D, Mule A, Szoke A, Selten J P, Turner C, Arango C, Tarricone I, Berardi D, Tortelli A, Llorca P M, de Haan L, Bobes J, Bernardo M, Sanjuan J, Santos J L, Arrojo M, Del-Ben C M, Menezes P R, Velthorst E, Murray R M, Rutten B P, Jones P B, van Os J, Morgan C, Kirkbride J B, European Network of National Schizophrenia Networks Studying Gene-Environment Interactions Work Package G. Treated Incidence of Psychotic Disorders in the Multinational E U-GEI Study. JAMA Psychiatry. 2018; 75(1):36-46. [0300] 31. European Network of National Networks studying Gene-Environment Interactions in S, van Os J, Rutten B P, Myin-Germeys I, Delespaul P, Viechtbauer W, van Zelst C, Bruggeman R, Reininghaus U, Morgan C, Murray R M, Di Forti M, McGuire P, Valmaggia L R, Kempton M J, Gayer-Anderson C, Hubbard K, Beards S, Stilo S A, Onyejiaka A, Bourque F, Modinos G, Tognin S, Calem M, O'Donovan M C, Owen M J, Holmans P, Williams N, Craddock N, Richards A, Humphreys I, Meyer-Lindenberg A, Leweke F M, Tost H, Akdeniz C, Rohleder C, Bumb J M, Schwarz E, Alptekin K, Ucok A, Saka M C, Atbasoglu E C, Guloksuz S, Gumus-Akay G, Cihan B, Karadag H, Soygur H, Cankurtaran E S, Ulusoy S, Akdede B, Binbay T, Ayer A, Noyan H, Karadayi G, Akturan E, Ulas H, Arango C, Parellada M, Bernardo M, Sanjuan J, Bobes J, Arrojo M, Santos J L, Cuadrado P, Rodriguez Solano J J, Carracedo A, Garcia Bernardo E, Roldan L, Lopez G, Cabrera B, Cruz S, Diaz Mesa E M, Pouso M, Jimenez E, Sanchez T, Rapado M, Gonzalez E, Martinez C, Sanchez E, Olmeda M S, de Haan L, Velthorst E, van der Gaag M, Selten J P, van Dam D, van der Ven E, van der Meer F, Messchaert E, Kraan T, Burger N, Leboyer M, Szoke A, Schurhoff F, Llorca P M, Jamain S, Tortelli A, Frijda F, Vilain J, Galliot A M, Baudin G, Ferchiou A, Richard J R, Bulzacka E, Charpeaud T, Tronche A M, De Hert M, van Winkel R, Decoster J, Derom C, Thiery E, Stefanis N C, Sachs G, Aschauer H, Lasser I, Winklbaur B, Schlogelhofer M, Riecher-Rossler A, Borgwardt S, Walter A, Harrisberger F, Smieskova R, Rapp C, Ittig S, Soguel-dit-Piquard F, Studerus E, Klosterkotter J, Ruhrmann S, Paruch J, Julkowski D, Hilboll D, Sham P C, Cherny S S, Chen E Y, Campbell D D, Li M, Romeo-Casabona C M, Emaldi Cirion A, Urruela Mora A, Jones P, Kirkbride J, Cannon M, Rujescu D, Tarricone I, Berardi D, Bonora E, Seri M, Marcacci T, Chiri L, Chierzi F, Storbini V, Braca M, Minenna M G, Donegani I, Fioritti A, La Barbera D, La Cascia C E, Mule A, Sideli L, Sartorio R, Ferraro L, Tripoli G, Seminerio F, Marinaro A M, McGorry P, Nelson B, Amminger G P, Pantelis C, Menezes P R, Del-Ben C M, Gallo Tenan S H, Shuhama R, Ruggeri M, Tosato S, Lasalvia A, Bonetto C, Ira E, Nordentoft M, Krebs M O, Barrantes-Vidal N, Cristobal P, Kwapil T R, Brietzke E, Bressan R A, Gadelha A, Maric N P, Andric S, Mihaljevic M, Mirjanic T. Identifying gene-environment interactions in schizophrenia: contemporary challenges for integrated, large-scale investigations. Schizophr Bull. 2014; 40(4):729-736. [0301] 32. Kraan T C, Velthorst E, Themmen M, Valmaggia L, Kempton M J, McGuire P, van Os J, Rutten B P F, Smit F, de Haan L, van der Gaag M, Study E-GHR. Child Maltreatment and Clinical Outcome in Individuals at Ultra-High Risk for Psychosis in the EU-GEI High Risk Study. Schizophr Bull. 2018; 44(3):584-592. [0302] 33. Yung A R, Yuen H P, McGorry P D, Phillips U, Kelly D, Dell'Olio M, Francey S M, Cosgrave E M, Killackey E, Stanford C, Godfrey K, Buckby J. Mapping the onset of psychosis: the Comprehensive Assessment of At-Risk Mental States. Aust N Z J Psychiatry. 2005; 39(11-12):964-971. [0303] 34. Aas I H. Global Assessment of Functioning (GAF): properties and frontier of current knowledge. Ann Gen Psychiatry. 2010; 9:20. [0304] 35. Goldman H H, Skodol A E, Lave T R. Revising axis V for DSM-IV: a review of measures of social functioning. Am J Psychiatry. 1992; 149(9):1148-1156. [0305] 36. Koutsouleris N, Kambeitz-llankovic L, Ruhrmann S, Rosen M, Ruef A, Dwyer D B, Paolini M, Chisholm K, Kambeitz J, Haidl T, Schmidt A, Gillam J, Schultze-Lutter F, Falkai P, Reiser M, Riecher-Rossler A, Upthegrove R, Hietala J, Salokangas R K R, Pantelis C, Meisenzahl E, Wood S J, Beque D, Brambilla P, Borgwardt S. Prediction Models of Functional Outcomes for Individuals in the Clinical High-Risk State for Psychosis or With Recent-Onset Depression: A Multimodal, Multisite Machine Learning Analysis. JAMA Psychiatry. 2018; 75(11):1156-1172. [0306] 37. Andreasen N C. The Scale for the Assessment of Negative Symptoms (SANS): conceptual and theoretical foundations. Br J Psychiatry Suppl. 1989(7):49-58. [0307] 38. Overall J E, Gorham D R. The Brief Psychiatric Rating Scale. Psychological Reports. 1962; 10(3):799-812. [0308] 39. Montgomery S A, Asberg M. A new depression scale designed to be sensitive to change. Br J Psychiatry. 1979; 134:382-389. [0309] 40. English J A, Fan Y, Focking M, Lopez L M, Hryniewiecka M, Wynne K, Dicker P, Matigian N, Cagney G, Mackay-Sim A, Cotter D R. Reduced protein synthesis in schizophrenia patient-derived olfactory cells. Transl Psychiatry. 2015; 5:e663. [0310] 41. Focking M, Opstelten R, Prickaerts J, Steinbusch H W, Dunn M J, van den Hove D L, Cotter D R. Proteomic investigation of the hippocampus in prenatally stressed mice implicates changes in membrane trafficking, cytoskeletal, and metabolic function. Dev Neurosci. 2014; 36(5):432-442. [0311] 42. Topol A, Zhu S, Hartley B J, English J, Hauberg M E, Tran N, Rittenhouse C A, Simone A, Ruderfer D M, Johnson J, Readhead B, Hadas Y, Gochman P A, Wang Y C, Shah H, Cagney G, Rapoport J, Gage F H, Dudley J T, Sklar P, Mattheisen M, Cotter D, Fang G, Brennand K J. Dysregulation of miRNA-9 in a Subset of Schizophrenia Patient-Derived Neural Progenitor Cells. Cell Rep. 2016; 15(5):1024-1036. [0312] 43. Boyd A, Golding J, Macleod J, Lawlor D A, Fraser A, Henderson J, Molloy L, Ness A, Ring S, Davey Smith G. Cohort Profile: the ‘children of the 90s’—the index offspring of the Avon Longitudinal Study of Parents and Children. Int J Epidemiol. 2013; 42(1):111-127. [0313] 44. Fraser A, Macdonald-Wallis C, Tilling K, Boyd A, Golding J, Davey Smith G, Henderson J, Macleod J, Molloy L, Ness A, Ring S, Nelson S M, Lawlor D A. Cohort Profile: the Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol. 2013; 42(1):97-110. [0314] 45. Cox J, Mann M. Quantitative, high-resolution proteomics for data-driven systems biology. Annu Rev Biochem. 2011; 80:273-299. [0315] 46. Cox J, Mann M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat Biotechnol. 2008; 26(12):1367-1372. [0316] 47. Lazar C. Impute LCMD. R package version 2.0. 2015. [0317] 48. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. 1995; Series B(57,1):289-300. [0318] 49. Gaonkar B, Davatzikos C. Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification. NeuroImage. 2013; 78:270-283. [0319] 50. Koutsouleris N, Kahn R S, Chekroud A M, Leucht S, Falkai P, Wobrock T, Derks E M, Fleischhacker W W, Hasan A. Multisite prediction of 4-week and 52-week treatment outcomes in patients with first-episode psychosis: a machine learning approach. The Lancet Psychiatry. 2016; 3(10):935-946. [0320] 51. Golland P, Fischl B. Permutation Tests for Classification: Towards Statistical Significance in Image-Based Studies. Paper presented at: Information Processing in Medical Imaging; 2003//, 2003; Berlin, Heidelberg. [0321] 52. Fan R-E, Chang K-W, Hsieh C-J, Wang X-R, Lin C-J. LIBLINEAR: A Library for Large Linear Classification. J. Mach. Learn. Res. 2008; 9:1871-1874. [0322] 53. Krawczyk B. Learning from imbalanced data: open challenges and future directions. Progress in Artificial Intelligence. 2016; 5(4):221-232. [0323] 54. Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-Cepas J, Simonovic M, Roth A, Santos A, Tsafou K P, Kuhn M, Bork P, Jensen U, von Mering C. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 2015; 43(Database issue):D447-452. [0324] 55. Radua J, Ramella-Cravaro V, Ioannidis J P A, Reichenberg A, Phiphopthatsanee N, Amir T, Yenn Thoo H, Oliver D, Davies C, Morgan C, McGuire P, Murray R M, Fusar-Poli P. What causes psychosis? An umbrella review of risk and protective factors. World Psychiatry. 2018; 17(1):49-66. [0325] 56. Studerus E, Ramyead A, Riecher-Rossler A. Prediction of transition to psychosis in patients with a clinical high risk for psychosis: a systematic review of methodology and reporting. Psychological Medicine. 2017; 47(7):1163-1178. [0326] 57. Malda A, Boonstra N, Barf H, de Jong S, Aleman A, Addington J, Pruessner M, Nieman D, de Haan L, Morrison A, Riecher-Rossler A, Studerus E, Ruhrmann S, Schultze-Lutter F, An SK, Koike S, Kasai K, Nelson B, McGorry P, Wood S, Lin A, Yung A Y, Kotlicka-Antczak M, Armando M, Vicari S, Katsura M, Matsumoto K, Durston S, Ziermans T, Wunderink L, Ising H, van der Gaag M, Fusar-Poli P, Pijnenborg G H M. Individualized Prediction of Transition to Psychosis in 1,676 Individuals at Clinical High Risk: Development and Validation of a Multivariable Prediction Model Based on Individual Patient Data Meta-Analysis. Front Psychiatry. 2019; 10:345. [0327] 58. Riecher-Rossler A, Studerus E. Prediction of conversion to psychosis in individuals with an at-risk mental state: a brief update on recent developments. Curr Opin Psychiatry. 2017; 30(3):209-219. [0328] 59. Mechelli A, Lin A, Wood S, McGorry P, Amminger P, Tognin S, McGuire P, Young J, Nelson B, Yung A. Using clinical information to make individualized prognostic predictions in people at ultra high risk for psychosis. Schizophr Res. 2017; 184:32-38. [0329] 60. Schmidt A, Cappucciati M, Radua J, Rutigliano G, Rocchetti M, Dell'Osso L, Politi P, Borgwardt S, Reilly T, Valmaggia L, McGuire P, Fusar-Poli P. Improving Prognostic Accuracy in Subjects at Clinical High Risk for Psychosis: Systematic Review of Predictive Models and Meta-analytical Sequential Testing Simulation. Schizophr Bull. 2017; 43(2):375-388. [0330] 61. Ruhrmann S, Schultze-Lutter F, Salokangas R K, Heinimaa M, Linszen D, Dingemans P, Birchwood M, Patterson P, Juckel G, Heinz A, Morrison A, Lewis S, von Reventlow H G, Klosterkotter J. Prediction of psychosis in adolescents and young adults at high risk: results from the prospective European prediction of psychosis study. Arch Gen Psychiatry. 2010; 67(3):241-251. [0331] 62. Koutsouleris N, Meisenzahl E M, Davatzikos C, Bottlender R, Frodl T, Scheuerecker J, Schmitt G, Zetzsche T, Decker P, Reiser M, Moller R J, Gaser C. Use of neuroanatomical pattern classification to identify subjects in at-risk mental states of psychosis and predict disease transition. Arch Gen Psychiatry. 2009; 66(7):700-712. [0332] 63. Koutsouleris N, Borgwardt S, Meisenzahl E M, Bottlender R, Moller H-J, Riecher-Rossler A. Disease Prediction in the At-Risk Mental State for Psychosis Using Neuroanatomical Biomarkers: Results From the FePsy Study. Schizophrenia Bulletin. 2011; 38(6):1234-1246. [0333] 64. Das T, Borgwardt S, Hauke D J, Harrisberger F, Lang U E, Riecher-Rossler A, Palaniyappan L, Schmidt A. Disorganized Gyrification Network Properties During the Transition to Psychosis. JAMA Psychiatry. 2018; 75(6):613-622. [0334] 65. Koutsouleris N, Davatzikos C, Bottlender R, Patschurek-Kliche K, Scheuerecker J, Decker P, Gaser C, Moller H-J, Meisenzahl E M. Early Recognition and Disease Prediction in the At-Risk Mental States for Psychosis Using Neurocognitive Pattern Classification. Schizophrenia Bulletin. 2011; 38(6):1200-1215. [0335] 66. Perkins D O, Jeffries C D, Addington J, Bearden C E, Cadenhead K S, Cannon T D, Cornblatt B A, Mathalon D H, McGlashan T H, Seidman U, Tsuang M T, Walker E F, Woods S W, Heinssen R. Towards a psychosis risk blood diagnostic for persons experiencing high-risk symptoms: preliminary results from the NAPLS project. Schizophr Bull. 2015; 41(2):419-428. [0336] 67. Ruhrmann S, Schultze-Lutter F, Schmidt S J, Kaiser N, Klosterkotter J. Prediction and prevention of psychosis: current progress and future tasks. Eur Arch Psychiatry Clin Neurosci. 2014; 264 Suppl 1:S9-16. [0337] 68. van der Gaag M, Smit F, Bechdolf A, French P, Linszen D H, Yung A R, McGorry P, Cuijpers P. Preventing a first episode of psychosis: meta-analysis of randomized controlled prevention trials of 12 month and longer-term follow-ups. Schizophr Res. 2013; 149(1-3):56-62. [0338] 69. Stafford M R, Jackson H, Mayo-Wilson E, Morrison A P, Kendall T. Early interventions to prevent psychosis: systematic review and meta-analysis. Bmj. 2013; 346:f185. [0339] 70. Hutton P, Taylor Pt Cognitive behavioural therapy for psychosis prevention: a systematic review and meta-analysis. Psychol Med. 2014; 44(3):449-468. [0340] 71. Preti A, Cella M. Randomized-controlled trials in people at ultra high risk of psychosis: a review of treatment effectiveness. Schizophr Res. 2010; 123(1):30-36. [0341] 72. Jaros J A, Martins-de-Souza D, Rahmoune H, Rothermundt M, Leweke F M, Guest P C, Bahn S. Protein phosphorylation patterns in serum from schizophrenia patients and healthy controls. Journal of proteomics. 2012; 76 Spec No.:43-55. [0342] 73. Yang Y, Wan C, Li H, Zhu H, La Y, Xi Z, Chen Y, Jiang L, Feng G, He L. Altered levels of acute phase proteins in the plasma of patients with schizophrenia. Anal Chem. 2006; 78(11):3571-3576. [0343] 74. Levin Y, Wang L, Schwarz E, Koethe D, Leweke F M, Bahn S. Global proteomic profiling reveals altered proteomic signature in schizophrenia serum. Mol Psychiatry. 2010; 15(11):1088-1100. [0344] 75. Hoirisch-Clapauch S, Amaral O B, Mezzasalma M A, Panizzutti R, Nardi A E. Dysfunction in the coagulation system and schizophrenia. Transl Psychiatry. 2016; 6:e704. [0345] 76. Kopczynska M, Zelek W, Touchard S, Gaughran F, Di Forti M, Mondelli V, Murray R, O'Donovan M C, Morgan B P. Complement system biomarkers in first episode psychosis. Schizophr Res. 2017; 17(pi is 50920-9964):30764-30768. [0346] 77. Boyajyan A, Khoyetsyan A, Chavushyan A. Alternative complement pathway in schizophrenia. Neurochem Res. 2010; 35(6):894-898. [0347] 78. Ricking M, Dicker P, Lopez L M, Cannon M, Schafer M R, McGorry P D, Smesny S, Cotter D R, [0348] Amminger G P. Differential expression of the inflammation marker IL12p40 in the at-risk mental state for psychosis: a predictor of transition to psychotic disorder? BMC psychiatry. 2016; 16(1):326. [0349] 79. Miller B J, Buckley P, Seabolt W, Mellor A, Kirkpatrick B. Meta-analysis of cytokine alterations in schizophrenia: clinical status and antipsychotic effects. Biol Psychiatry. 2011; 70(7):663-671. [0350] 80. Khandaker G M, Pearson R M, Zammit S, Lewis G, Jones P B. Association of serum interleukin 6 and C-reactive protein in childhood with depression and psychosis in young adult life: a population-based longitudinal study. JAMA Psychiatry. 2014; 71(10):1121-1128. [0351] 81. van Beveren N J, Schwarz E, Noll R, Guest P C, Meijer C, de Haan L, Bahn S. Evidence for disturbed insulin and growth hormone signaling as potential risk factors in the development of schizophrenia. Transl Psychiatry. 2014; 4:e430. [0352] 82. Schwarz E, van Beveren N J, Ramsey J, Leweke F M, Rothermundt M, Bogerts B, Steiner J, Guest P C, Bahn S. Identification of subgroups of schizophrenia patients with changes in either immune or growth factor and hormonal pathways. Schizophr Bull. 2014; 40(4):787-795. [0353] 83. Baumeister D, Russell A, Pariante C M, Mondelli V. Inflammatory biomarker profiles of mental disorders and their relation to clinical, social and lifestyle factors. Soc Psychiatry Psychiatr Epidemiol. 2014; 49(6):841-849. [0354] 84. Laskaris L, Zalesky A, Weickert C S, Di Biase M A, Chana G, Baune B T, Bousman C, Nelson B, McGorry P, Everall I, Pantelis C, Cropley V. Investigation of peripheral complement factors across stages of psychosis. Schizophr Res. 2018. [0355] 85. Rehman A A, Ahsan H, Khan F H. alpha-2-Macroglobulin: a physiological guardian. J Cell Physiol. 2013; 228(8):1665-1675. [0356] 86. Upthegrove R, Manzanares-Teson N, Barnes N M. Cytokine function in medication-naive first episode psychosis: a systematic review and meta-analysis. Schizophr Res. 2014; 155(1-3):101-108. [0357] 87. Borth W. Alpha 2-macroglobulin, a multifunctional binding protein with targeting characteristics. Faseb j. 1992; 6(15):3345-3353. [0358] 88. de Boer J P, Creasey A A, Chang A, Abbink J J, Roem D, Eerenberg A J, Hack C E, Taylor F B, Jr. Alpha-2-macroglobulin functions as an inhibitor of fibrinolytic, clotting, and neutrophilic proteinases in sepsis: studies using a baboon model. Infect Immun. 1993; 61(12):5035-5043. [0359] 89. Baker S K, Chen Z L, Norris E H, Revenko A S, MacLeod A R, Strickland S. Blood-derived plasminogen drives brain inflammation and plaque deposition in a mouse model of Alzheimer's disease. Proc Natl Acad Sci USA. 2018; 115(41):E9687-E9696. [0360] 90. Amara U, Flierl M A, Rittirsch D, Klos A, Chen H, Acker B, Bruckner U B, Nilsson B, Gebhard F, Lambris J D, Huber-Lang M. Molecular intercommunication between the complement and coagulation systems. J Immunol. 2010; 185(9):5628-5636. [0361] 91. Ryu J K, Petersen M A, Murray S G, Baeten K M, Meyer-Franke A, Chan J P, Vagena E, Bedard C, Machado M R, Rios Coronado P E, Prod′homme T, Charo I F, Lassmann H, Degen J L, Zamvil S S, Akassoglou K. Blood coagulation protein fibrinogen promotes autoimmunity and demyelination via chemokine release and antigen presentation. Nat Commun. 2015; 6:8164. [0362] 92. Pollak T A, Drndarski S, Stone J M, David A S, McGuire P, Abbott N J. The blood-brain barrier in psychosis. Lancet Psychiatry. 2018; 5(1):79-92. [0363] 93. Ryu J K, Rafalski V A, Meyer-Franke A, Adams R A, Poda S B, Rios Coronado P E, Pedersen L O, Menon V, Baeten K M, Sikorski S L, Bedard C, Hanspers K, Bardehle S, Mendiola A S, Davalos D, Machado M R, Chan J P, Plastira I, Petersen M A, Pfaff S J, Ang K K, Hallenbeck K K, Syme C, Hakozaki H, Ellisman M H, Swanson R A, Zamvil S S, Arkin M R, Zorn S H, Pico A R, Mucke L, Freedman S B, Stavenhagen J B, Nelson R B, Akassoglou K. Fibrin-targeting immunotherapy protects against neuroinflammation and neurodegeneration. Nat Immunol. 2018; 19(11):1212-1223. [0364] 94. Comes A L, Papiol S, Mueller T, Geyer P E, Mann M, Schulze T G. Proteomics for blood biomarker exploration of severe mental illness: pitfalls of the past and potential for the future. Translational Psychiatry. 2018; 8(1):160. [0365] 95. Deighton S, Neville A, Pusch D, Dobson K. Biomarkers of adverse childhood experiences: A scoping review. Psychiatry Res. 2018; 269:719-732. [0366] 96. Baumeister D, Akhtar R, Ciufolini S, Pariante C M, Mondelli V. Childhood trauma and adulthood inflammation: a meta-analysis of peripheral C-reactive protein, interleukin-6 and tumour necrosis factor-α. Molecular psychiatry. 2016; 21(5):642-649. [0367] 97. Rasmussen U H, Moffitt T E, Arseneault L, Danese A, Eugen-Olsen J, Fisher H L, Harrington H, Houts R, Matthews T, Sugden K, Williams B, Caspi A. Association of Adverse Experiences and Exposure to Violence in Childhood and Adolescence With Inflammatory Burden in Young People. JAMA Pediatrics. 2019:1-11.

Supplementary Material & Methods

Sample Preparation

Blood Collection

[0368] Plasma P100 tubes were used for blood collection and samples were stored on ice for a maximum of 90 minutes until processed, centrifuged and stored in a −80° C. freezer. The standard quality of the plasma samples was ensured by assessing the overall MS protein profile to facilitate the identification of outlier protein expression profiles.

Protein Depletion of Plasma Samples

[0369] To improve the dynamic range for proteomic analysis, 40 μl of plasma from each case in all samples was immunodepleted of the 14 most abundant proteins (α-1-antitrypsin, A1-acid glycoprotein, Serum Albumin, α2-macroglobulin, Apolipoprotein A-I, Apolipoptrotein A-II, Complement C3, Fibrinogen α/β/γ, Haptoglobin, IgA, IgG, IgM, Transthyretin, and Serotransferrin), using the Agilent Hu14 Affinity Removal System (MARS) coupled to a High Performance Liquid Chromatography (HPLC) system.sup.1. Protein depletion was undertaken according to the manufacturer's instructions and buffer exchange was performed with 50 mM ammonium bicarbonate using spin columns with a 10 kDA-molecular weight cut-off (Merck Millipore). Prior to sample preparation for mass spectrometry (MS), the protein concentration was determined using a Bradford Assay.sup.2, according to the manufacturer's (BioRad) instructions.

Sample Preparation for Mass Spectrometry

[0370] Protein digestion and peptide purification was performed as previously described.sup.3. For quality control (QC), an equal aliquot from each protein digest in the experiment was pooled into one sample for use as an internal QC. This QC standard was injected at the beginning of the MS study to condition the column, and after every ten injections throughout the experiment to monitor the MS performance.

Discovery Proteomic Analysis Using Data Dependent Acquisition (DDA)

[0371] All samples were injected on a Thermo Scientific Q Exactive mass spectrometer connected to a Dionex Ultimate 3000 (RSLCnano) chromatography system. Tryptic peptides (5 μl of digest) from each sample were loaded onto a fused silica emitter (75 μm ID, pulled using a laser puller (Sutter Instruments P2000), packed with UChrom C18 (1.8 μm) reverse phase media (nanoLCMS Solutions LCC) and was separated by an increasing acetonitrile gradient over 90 minutes at a flow rate of 250 nL/min. This QC standard was injected 3 times at the beginning of the MS study to condition the column, and after every ten injections throughout the experiment to monitor the MS performance. The mass spectrometer was operated in data dependent TopN 8 mode, with the following settings: mass range 300-1600 Th; resolution for MS1 scan 70000; A Vitamin D binding protein target 3e6; resolution for MS2 scan 17500; A Vitamin D binding protein target 2e4; charge exclusion unassigned, 1; dynamic exclusion 40 s.

Confirmatory Analyses by ELISA

[0372] To validate our findings, we assessed several human complement and coagulation proteins and apolipoproteins in the plasma samples of the same CHR-T and CHR-NT subjects who contributed to the proteomic study using enzyme-linked immunoassays (ELISA). Candidate proteins were chosen on the basis of machine learning results and differential expression as well as previous study results from our group (.sup.4-7). Specifically, we tested human α-2 macroglobulin (Abcam ab108888, 1:400), apolipoprotein E (ThermoFisher Scientific, EHAPOE, 1:2,000), Complement C1q (Abcam ab170246, 1:100,000), Complement C1r (Abcam ab170245, 1:40,000), Complement C4 binding protein (Abcam, ab222866, 1:40,000), Complement C8 (Abcam ab 137971, 1:10,000), complement factor H (Hykult Biotech HK342, 1:10,000), immunoglobulin M (Abcam, ab137982, 1:60,000), and plasminogen (Abcam ab108893, 1:20,000) in accordance with the manufacturer's instructions. Concentrations for unknown samples were interpolated using 4 parameter logistic curve fit in GraphPad Prism 8 software and means for each protein were compared using the t-test with unequal variances in Stata version 15.

Bioinformatics and Statistical Analysis

EU-GEI Clinical Variables

[0373] A full list of the baseline clinical variables included is provided in eTable 1.

Leave-Site-Out Cross-Validation

[0374] The data were first split into a number of folds in the ‘outer loop’ of cross-validation. To incorporate geographical generalisability, we split the data by study site. Several of the smaller sites from the EU-GEI study were combined to ensure a large enough transition sample was present at each site. Thus, Amsterdam and The Hague were combined, Vienna and Basel were combined, Copenhagen and Paris were combined, and Barcelona and Sao Paulo were combined. This resulted in 6 final sites which were folds in the outer cross-validation loop (% transition): London (35.2%), the Netherlands (12.5%), Switzerland/Austria (57.1%), Melbourne (35.7%), Denmark/France (45.8%) and Spain/Brazil (30.8%). For each cycle of cross-validation, data from each of the 6 sites were held out and the rest of the data moved into the ‘inner loop’ for training.

Repeated Nested Cross-Validation

[0375] Within the inner loop, we used 5 non-overlapping folds with iterative training-test cycles. Thus, training was applied to four-fifths of the data in the inner loop and then tested against the final one-fifth, with the five different inner loop folds as the test fold. Models were trained and tested within the inner loop using a range of regularisation parameter values (in 11 steps from 0.015625 to 16).

[0376] The optimal models thus derived were tested against the held-out site in the outer loop. This process was then repeated, with each site in the outer loop as the test site, to determine the overall optimal model and final predictive accuracy. For a detailed description of repeated nested cross-validation, see the Supplementary material of Koutsouleris et al.sup.8 and the Neurominer manual (available from https//www pronia.eu/neurominer/).

Confidence Intervals for Area Under the Curve

[0377] 95% confidence intervals for the area under the receiver-operating curve (AUC) for each model were calculated according to the method of Hanley & McNeil..sup.9

EU-GEI Replication Dataset

[0378] The replication dataset included 49 CHR-T participants (2 of whom were different from those in Models 1a-d) and 86 CHR-NT participants (all of whom were different from those in Models 1a-d). Characteristics of participants included in the replication dataset are compared to those not included in Table 9. Included participants were more likely to be male, but otherwise were comparable with non-included participants on baseline characteristics. Characteristics for CHR-T and CHR-NT participants included in the replication dataset are compared in Table 10. There was evidence of differences for ethnicity and tobacco use between the two groups, higher mean total SANS global and composite scores and total BPRS score in CHR-T.

Supplementary References

[0379] 1. Levin Y, Wang L, Schwarz E, Koethe D, Leweke F M, Balm S. Global proteomic profiling reveals altered proteomic signature in schizophrenia serum. Mol Psychiatry. 2010; 15(11):1088-1100. [0380] 2. Bradford M M. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem. 1976; 72:248-254. [0381] 3. English J A, Fan Y, Focking M, Lopez L M, Hryniewiecka M, Wynne K, Dicker P, Matigian N, Cagney G, Mackay-Sim A, Cotter D R. Reduced protein synthesis in schizophrenia patient-derived olfactory cells. Transl Psychiatry. 2015; 5:e663. [0382] 4. English J A, Lopez L M, O'Gorman A, Focking M, Hryniewiecka M, Scaife C, Sabherwal S, Wynne K, Dicker P, Rutten B P F, Lewis G, Zammit S, Cannon M, Cagney G, Cotter D R. Blood-Based Protein Changes in Childhood Are Associated With Increased Risk for Later Psychotic Disorder: Evidence From a Nested Case-Control Study of the ALSPAC Longitudinal Birth Cohort. Schizophr Bull. 2018; 44(2):297-306. [0383] 5. Ricking M, Sabherwal S, Cates H M, Scaife C, Dicker P, Hryniewiecka M, Wynne K, Rutten B P F, Lewis G, Cannon M, Nestler E J, Heurich M, Cagney G, Zammit S, Cotter D R. Complement pathway changes at age 12 are associated with psychotic experiences at age 18 in a longitudinal population-based study: evidence for a role of stress. Molecular Psychiatry. 2019. [0384] 6. Sabherwal S, English J A, Focking M, Cagney G, Cotter D R. Blood biomarker discovery in drug-five schizophrenia: the contribution of proteomics and multiplex immunoassays. Expert Rev Proteomics. 2016; 13(12):1141-1155. [0385] 7. Sabherwal S, Focking M, English J A, Fitzsimons S, Hryniewiecka M, Wynne K, Scaife C, Healy C, Cannon M, Belton O, Zammit S, Cagney G, Cotter D R. ApoE elevation is associated with the persistence of psychotic experiences from age 12 to age 18: Evidence from the ALSPAC birth cohort. Schizophr Res. 2019. [0386] 8. Koutsouleris N, Kahn R S, Chekroud A M, Leucht S, Falkai P, Wobrock T, Derks E M, Fleischhacker W W, Hasan A. Multisite prediction of 4-week and 52-week treatment outcomes in patients with first-episode psychosis: a machine learning approach Lancet Psychiatry. 2016; 3(10):935-946. [0387] 9. Hanley J A, McNeil B J. The meaning and use of the are under a receiver operating characteristic (ROC) curve. Radiology. 1982 April; 143(1):29-36.

TABLE-US-00004 TABLE 4 List of 65 baseline clinical variables included in EU-GEI support vector models GAF symptoms GAF disability SANS: unchanging facial expression SANS: decreased spontaneous movements SANS: paucity of expressive gestures SANS: poor eye contact SANS: affective nonresponsivity SANS: inappropriate affect SANS: lack of vocal inflections SANS: global rating of affective flattening SANS: poverty of speech SANS: poverty of speech content SANS: blocking SANS: increased latency of response SANS: global rating of alogia SANS: grooming and hygiene SANS: impersistence at work or school SANS: physical anergia SANS: global rating for avolition-apathy SANS: recreational interests and activities SANS: sexual activity SANS: ability to feel intimacy and closeness SANS: relationship with friends and peers SANS: global rating of anhedonia-asociality SANS: social inattentiveness SANS: inattentiveness during mental status testing SANS: global rating of attention Total SANS composite score Total SANS global score BPRS: somatic concern BPRS: anxiety BPRS: depression BPRS: suicidality BPRS: guilt BPRS: hostility BPRS: elevated mood BPRS: grandiosity BORS: suspiciousness BPRS: hallucinations BPRS: unusual thought content BPRS: bizarre behaviour BPRS: self-neglect BPRS: disorientation BPRS: conceptual disorganisation BPRS: blunted affect BPRS: emotional withdrawal BPRS: motor retardation BPRS: tension BPRS: uncooperativeness BPRS: excitement BPRS: distractibility BPRS: motor hyperactivity BPRS: mannerisms and posturing Total BPRS score MADRS: apparent sadness MADRS: reported sadness MADRS: inner tension MADRS: reduced sleep MADRS: reduced appetite MADRS: concentration difficulties MADRS: lassitude MADRS: inability to feel MADRS: pessimistic thoughts MADRS: suicidal thoughts Total MADRS score GAF: General Assessment of Functioning; SANS: Scale for the Assessment of Negative Symptoms; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery Asberg Depression Rating Scale

TABLE-US-00005 TABLE 5 Comparison of characteristics for participants included in original experiment (N = 133) from total EU-GEI CHR cohort (N = 344) Included, Not included, N = 133 N = 211 Missing (49 CHR-T, (16 CHR-T, data, n (%) 84 CHR-NT) 195 CHR-NT) t/χ.sup.2 p Baseline age 0 22.6 (4.5) 22.3 (5.2) 0.686 0.493 in years, mean (SD) Sex, n (%) 0 68 male (51.1%) 117 male (55.5%) 0.613 0.434 65 female (49.9%) 94 female (45.5%) Baseline 50 (14.5%) 24.4 (5.6) 23.7 (4.9) 1.190 0.235 BMI in kg/m.sup.2, mean (SD) Baseline 38 (11.0%) 14.3 (3.1) 14.4 (3.0) −0.318 0.751 years in education, mean (SD) Ethnicity, n 0 91 white (68.4%) 156 white (73.9%) 1.239 0.538 (%) 15 black (11.3%) 19 black (9.0%) 27 other (20.3%) 36 other (17.1%) Ever used 10 (2.9%) 101 yes (75.9%) 143 yes (67.8%) 2.326 0.127 cannabis, 29 no (21.8%) 61 no (28.9%) n (%) 3 not known (2.3%) 7 not known (3.3%) Baseline 95 (27.6%) 41 yes (30.8%) 47 yes (22.3%) 1.302 0.254 cannabis 63 no (47.4%) 98 no (46.4%) use, n (%) 29 not known (21.8%) 66 not known (31.3%) Baseline 38 (11.0%) 64 yes (48.1%) 97 yes (46.0%) 0.106 0.744 tobacco use, 55 no (41.4%) 90 no (42.7%) n (%) 14 not known (10.5%) 24 not known (11.4%) Baseline 12 (3.5%) 93 yes (69.9%) 141 yes (66.8%) 0.115 0.735 alcohol use, 37 no (27.8%) 61 no (28.9%) n (%) 3 not known (2.3%) 9 not known (4.3%) Baseline 86 (25.0%) 51 yes (38.3%) 78 yes (37.0%) 0 1.0 medication Antidepressant 29 Antidepressant 45 use, n (%) Antipsychotic 9 Antipsychotic 11 Hypnotic 7 Hypnotic 7 Other 6 Other 15 51 no (38.3%) 78 no (37.0%) 31 not known (23.3%) 55 not known (26.1%) Baseline 27 (7.8%) 54.7 (10.2) 55.4 (10.0) −0.634 0.526 GAF symptoms score, mean (SD) Baseline 12 (3.5%) 53.9 (11.7) 56.4 (12.5) −1.827 0.069 GAF disability score, mean (SD) Baseline 45 (13.1%) 18.0 (12.7) 14.2 (10.7) 2.782 0.006 SANS total composite score, mean (SD) Baseline 29 (8.4%) 6.1 (3.9) 5.0 (3.4) 2.528 0.012 SANS total global score, mean (SD) Baseline 25 (7.3%) 46.0 (10.9) 42.2 (9.6) 3.174 0.002 BPRS total score, mean (SD) Baseline 16 (4.7%) 19.6 (9.7) 18.4 (8.8) 1.155 0.249 MADRS total score, mean (SD) 2 year GAF 142 (41.3%) 54.6 (15.0) 63.0 (11.6) −4.083 <0.001 symptoms score, mean (SD) 2 year GAF 124 (36.0%) 56.9 (15.0) 63.6 (13.8) −3.333 0.001 disability score, mean (SD) 2 year GAF 124 (36.0%) 32 good (24.1%) 80 good (37.9%) 5.337 0.021 disability 47 poor (35.3%) 61 poor (28.9%) score, 54 not known (40.6%) 70 not known (33.2%) dichotomous outcome .sup.a .sup.a Poor functioning: GAF disability score ≤60; good functioning: GAF disability score >60 Tobacco use was defined as daily use for at least 1 month over the previous 12 months. Alcohol use was defined as at least 12 or more alcoholic beverages over the previous 12 months. Missing data excluded in hypothesis tests. EU-GEI: European Union Gene Environment Interaction study; CHR-T: clinical high risk, transitioned to psychosis; CHR-NT: clinical high risk, did not transition to psychosis; BMI: body mass index; GAF: General Assessment of Functioning; SANS: Scale for the Assessment of Negative Symptoms; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery Asberg Depression Rating Scale

TABLE-US-00006 TABLE 6 Results of ANCOVA and fold changes (CHR-T vs. CHR-NT) for proteins identified in EU- GEI baseline plasma samples (adjusted for age, sex, BMI and years in education) Mean LFQ, Ratio of Mean LFQ, non- means FDR transition transition (T vs Uniprot Name F p 5% group group NT] P01023 Alpha-2-macroglobulin 146 7.55E−23 <0.001 1.14E+09 3.49E+09 0.33 P01871 Immunoglobulin heavy 72.16 4.53E−14 <0.001  1.1E+08 2.71E+08 0.41 constant mu P07357 Complement component C8 44.25 7.76E−10 <0.001  2.5E+08 1.69E+08 1.48 alpha chain P02774 Vitamin D-binding protein 40.97 2.72E−09 <0.001  7.4E+09 5.17E+09 1.43 P02747 Complement C1q 36.52 1.56E−08 <0.001 2.06E+08 1.35E+08 1.53 subcomponent subunit C P00747 Plasminogen 31.39 1.25E−07 <0.001 4.09E+09 3.17E+09 1.29 P10909 Clusterin 29.74 2.48E−07 <0.001 1.64E+09 1.27E+09 1.29 P23142 Fibulin-1 28.56 4.06E−07 <0.001 83608657 55104835 1.52 P55058 Phospholipid transfer protein 19.09 2.57E−05 <0.001  7342621 10962754 0.67 P00736 Complement C1r 18.07 4.09E−05 <0.001  9.9E+08 7.81E+08 1.27 subcomponent P08603 Complement factor H 16.18 0.0001 0.001 4.43E+09 3.81E+09 1.16 P05156 Complement factor I 16.72 0.0001 0.001 1.42E+09 1.16E+09 1.23 O75882 Attractin 16.85 0.0001 0.001 1.17E+08 90521420 1.30 P03951 Coagulation factor XI 17.61 0.0001 0.001 44890311 32936650 1.36 P43320 Beta-crystallin B2 16.39 0.0001 0.001 1.23E+08 63601609 1.80 P04003 C4b-binding protein alpha 14.71 0.0002 0.002 6.96E+08 9.15E+08 0.76 chain P19827 Inter-alpha-trypsin inhibitor 15.05 0.0002 0.002 2.95E+09 2.49E+09 1.19 heavy chain H1 O75636 Ficolin-3 14.09 0.0003 0.003 1.48E+08 2.11E+08 0.70 P01860 Immunoglobulin heavy 13.19 0.0004 0.003 5.61E+08 8.04E+08 0.70 constant gamma 3 P15144 Aminopeptidase N 13.38 0.0004 0.003 10232560 14138254 0.72 P02489 Alpha-crystallin A chain 12.08 0.0007 0.006 1.81E+08 1.11E+08 1.63 P06396 Gelsolin 11.83 0.0008 0.006 1.92E+09 1.64E+09 1.17 Q14520 Hyaluronan-binding protein 2 11.87 0.0008 0.006 1.31E+08 1.08E+08 1.21 P05155 Plasma pretease C1 inhibitor 11.01 0.0012 0.008 1.66E+09 1.95E+09 0.85 P02766 Transthyretin 10.79 0.0013 0.009 21337361 35624572 0.60 P04217 Alpha-1B-glycoprotein 9.97 0.002 0.012 3.37E+09 4.04E+09 0.83 P02749 Beta-2-glycoprotein 1 9.93 0.002 0.012 5.76E+09 4.54E+09 1.27 P22891 Vitamin K-dependert protein Z 9.86 0.0021 0.012 24836850 36479890 0.68 P00751 Complement factor B 9.41 0.0026 0.015 4.72E+09 4.04E+09 1.17 P05546 Heparin cofactor 2 8.91 0.0034 0.019 2.04E+09 2.43E+09 0.84 P06276 Cholinesterase 8.71 0.0038 0.020  1.5E+08 1.26E+08 1.18 P51884 Lumican 8.4 0.0044 0.023 5.93E+08 4.74E+08 1.25 P02649 Apolipoprotein E 8.25 0.0048 0.024  9.1E+08  7.1E+08 1.28 Q76LX8 A disintegrin and 7.56 0.0068 0.033  9713735 11499010 0.84 metalloproteinase with thrombospondin motifs 13 Q06033 Inter-alpha-trypsin inhibitor 7.15 0.0085 0.040 2.04E+08 2.75E+08 0.74 heavy chain H3 P02656 Apolipoprotein C-III 6.59 0.0114 0.053  4.4E+08 3.22E+08 1.37 P05543 Thyroxine-binding globulin 6.43 0.0124 0.054 1.93E+08 1.64E+08 1.18 P02751 Fibronectin 6.43 0.0124 0.054 7.41E+09 5.71E+09 1.30 P00450 Ceruloplasmin 6.32 0.0132 0.056 5.67E+09 4.99E+09 1.14 Q04756 Hepatocyte growth factor 5.69 0.0185 0.077 1.08E+38 94519792 1.14 activator P05090 Apolipoprotein D 5.61 0.0193 0.078 32446284 27798720 1.17 Q08380 Galectin-3-binding protein 5.38 0.0219 0.087 9803809 12738574 0.77 P11226 Mannose-binding protein C 5.18 0.0246 0.092 56219776 41212566 1.36 P10643 Complement component C7 5.17 0.0247 0.092 5.43E+08 4.64E+08 1.17 P07225 Vitamin K-dependent protein S 5.12 0.0254 0.092  2.3E+08 2.64E+08 0.87 Q9BXR6 Complement factor H-related 5.1 0.0256 0.092 30517136 23408385 1.30 protein 5 P49747 Cartilage oligomeric matrix 4.94 0.028 0.099 21691930 18636716 1.16 protein P02675 Fibrinogen beta chain 4.79 0.0305 0.103 6.19E+08 5.39E+08 1.15 P02671 Fibrinogen alpha chain 4.77 0.0307 0.103 5.85E+08 5.24E+08 1.12 P04114 Apolipoprotein B-100 4.75 0.0311 0.103 1.92E+10 2.16E+10 0.89 P01042 Kininogen-1 4.69 0.0322 0.105  4.3E+09 3.94E+09 1.09 P43652 Afamin 4.49 0.0361 0.115 1.92E+09 1.72E+09 1.12 P05160 Coagulation factor XIII B chain 4.25 0.041 0.125 1.05E+08 91990562 1.14 Q9NZP8 Complement C1r 4.23 0.0417 0.125 79864565 62745563 1.27 subcomponent-like protein P02753 Retinol-binding protein 4 4.22 0.0421 0.125 7.38E+08 5.44E+08 1.36 P00742 Coagulation factor X 4.21 0.0423 0.125 1.66E+08 1.43E+08 1.11 P36980 Complement factor H-related 3.72 0.0559 0.163 84913609 65032425 1.31 protein 2 P07195 L-lactate dehydrogenase B 3.68 0.0572 0.164 29579331 25550208 2.16 chain P02748 Complement component C9 3.64 0.0585 0.165  1.3E+09 1.43E+09 0.91 P01024 Complement C3 3.56 0.0615 0.170 4.52E+08 3.99E+08 1.13 P01876 Immunoglobulin heavy 3.4 0.0675 0.184 1.84E+08 1.62E+08 1.14 constant alpha 1 P02743 Serum amyloid P-component 3.33 0.0705 0.189 8.62E+08  7.7E+08 2.22 P05452 Tetranectin 3.21 0.0756 0.199 5.07E+08 4.55E+08 1.21 P20742 Pregnancy zone protein 2.87 0.093 0.241 85801269 1.13E+08 0.76 P00740 Coagulation factor IX 2.83 0.0949 0.242 1.58E+08 1.41E+08 1.12 P01859 Immunoglobulin heavy 2.55 0.1125 0.283 73571215 65265721 1.13 constant gamma 2 P00748 Coagulation factor XII 2.49 0.117 0.290 6.19E+08 5.44E+06 1.14 P68871 Hemoglobin subunit beta 2.3 0.1315 0.317 1.92E+08 1.45E+08 1.33 P02787 Serotransferrin 2.3 0.1316 0.317 7.38E+08 7.06E+08 1.05 P01011 Alpha-1-antichymotrypsin 2.27 0.1347 0.319 5.65E+09 5.31E+09 1.06 P07737 Profilin-1 2.07 0.1526 0.357 42263756 56358056 0.75 Q92954 Proteoglycan 4 2.03 0.1565 0.361 47223124 53868662 0.88 P13671 Complement component C6 1.86 0.1745 0.397 8.74E+08 8.26E+08 1.06 P09871 Complement C1s 1.84 0.1776 0.398  7.8E+08 7.36E+08 1.06 subcomponent P07358 Complement component C8 1.77 0.1856 0.405 3.71E+08 4.06E+08 0.91 beta chain Q03591 Complement factor H-related 1.77 0.1856 0.405 81709939 68757368 1.19 protein 1 P02654 Apolipoprotein C-I 1.55 0.2147 0.463 1.31E+08  1.9E+08 0.69 P48740 Mannan-binding lectin serine 1.43 0.2336 0.497 35332742 32142164 1.10 protease 1 P02760 Protein AMBP [Cleaved into: 1.41 0.2371 0.498 1.29E+09 1.19E+09 1.08 Alpha-1-microglobulin P02790 Hemopexin 1.37 0.2439 0.503 9.58E+09 1.02E+10 0.94 P29622 Kallistatin 1.34 0.2484 0.503 6.97E+08 6.58E+08 1.06 P04004 Vitronectin 1.34 0.2486 0.503 3.24E+09 3.38E+09 0.96 P14618 Pyruvate kinase PKM 1.32 0.253 0.506 21521028 25635061 0.84 P22792 Carboxypeptidase N subunit 2 1.26 0.2639 0.522 2.59E+08 2.44E+08 1.06 P17936 Insulin-like growth factor- 1.21 0.2727 0.527 1.37E+08  1.5E+08 0.91 binding protein 3 Q96PD5 N-acetylmuramoyl-L-alanine 1.21 0.2734 0.527 5.23E+08 5.63E+08 0.93 amidase P01019 Angiotensinogen 1.19 0.2767 0.527 3.35E+09 3.75E+09 0.89 Q96XN2 Beta-Ala-His dipeptidase 1.17 0.2808 0.527 93683454 1.02E+08 0.92 P02652 Apolipoprotein A-II 1.15 0.2828 0.527 3.59E+08  3.7E+08 0.97 P27169 Serum 1.13 0.2894 0.534 72098714 65029166 1.21 paraoxonase/arylesterase 1 P22105 Tenascin-X 1.08 0.3001 0.547 31438477 27910277 1.13 P09172 Dopamine beta-hydroxylase 0.95 0.3314 0.598 28931848 26122837 1.11 P60174 Triosephosphate isomerase 0.89 0.3481 0.621  8691317 10128185 0.86 P15169 Carboxypeptidase N catalytic 0.8 0.3724 0.641 1.58E+08 1.48E+08 1.07 chain P08185 Corticosteroid-binding globulin 0.8 0.3725 0.641   2E+08  2.3E+08 0.87 P23528 Cofilin-1 0.8 0.3737 0.641 37675779 33910502 1.11 P80108 Phosphatidylinositol-glycan- 0.79 0.3746 0.641 2.47E+08 2.28E+08 1.08 specific phospholipase D P02775 Platelet basic protein 0.78 0.3801 0.644 1.93E+08 2.47E+08 0.78 P06681 Complement C2 0.72 0.3974 0.666  2.7E+08 2.87E+08 0.94 P00734 Prothrombin 0.67 0.413 0.674 4.54E+09 4.41E+09 1.03 P60709 Actin, cytoplasmic 1 0.67 0.4157 0.674 3.01E+08 2.93E+08 1.03 P35858 Insulin-like growth factor- 0.66 0.4173 0.674 6.57E+08 6.95E+08 0.94 binding protein complex acid labile subunit P02647 Apolipoprotein A-I 0.66 0.4185 0.674 8.83E+08 9.95E+08 0.89 P02763 Alpha-1-acid glycoprotein 1 0.63 0.4285 0.677 1.61E+08 1.63E+08 0.99 P04075 Fructose-bisphosphate 0.62 0.4332 0.677 39556865 41635811 0.95 aldolase A P18428 Lipopolysaccharide-binding 0.62 0.4335 0.677  1.3E+08 1.43E+08 0.91 protein P01834 Immunuglobulin kappa 0.61 0.4363 0.677 1.28E+08 1.31E+08 0.98 constant P04196 Histidine-rich glycoprotein 0.58 0.4474 0.688 1.51E+09  1.4E+09 1.08 Q9V6R7 IgGFc-binding protein 0.56 0.456 0.691 45729367 44506955 1.03 P08697 Alpha-2-antiplasmin 0.55 0.4594 0.691 1.13E+09 1.16E+09 0.97 Q14624 Inter-alpha-trypsin inhibitor 0.54 0.4618 0.691 4.43E+09 4.35E+09 1.02 heavy chain H4 P00738 Haptoglobin 0.53 0.468 0.694 4.45E+08 7.47E+08 0.60 O43866 CD5 antigen-like 0.5 0.4793 0.698 39577655 42728179 0.93 P0C0L5 Complement C4-B 0.5 0.4794 0.698 3.45E+08 3.07E+08 1.12 Q12913 Receptor-type tyrosine-protein 0.49 0.4863 0.702 8940384  8582967 1.04 phosphatase eta P04070 Vitamin K-dependent protein C 0.44 0.5103 0.730 41642694 47336839 0.88 P12814 Alpha-actinin-1 0.41 0.5251 0.739 33166338 40502857 0.82 P11021 Endoplasmic reticulum chaperone BiP 0.41 0.5252 0.739 10424759 10245287 1.02 P04275 von Willebrand factor 0.36 0.5505 0.768 96700211 89150500 1.08 P36955 Pigment epithelium-derived factor 0.33 0.5696 0.784 9.18E+08  8.8E+08 1.04 P03952 Plasma kallikrein 0.32 0.5714 0.784  5.4E+08 5.26E+08 1.03 P07360 Complement component C8 0.3 0.5837 0.793 3.22E+08 3.22E+08 1.00 gamma chain P02750 Leucine-rich alpha-2- 0.29 0.5908 0.793 8.16E+08 7.95E+08 1.03 glycoprotein Q15582 Transforming growth factor- 0.29 0.5922 0.793 22379758 23515537 0.95 beta-induced protein ig-h3 P49908 Selenoprotein P 0.27 0.6054 0.803 97906499 92437854 1.06 Q15610 Extracellular matrix protein 1 0.26 0.5124 0.803  2.7E+08 2.68E+08 1.01 P19825 Inter-alpha-typsin inhibtor 0.26 0.5141 0.803 3.18E+09 3.26E+09 0.97 heavy chain H2 P19320 Vascular cell adhesion protein 0.23 0.629 0.814 19514434 21040364 0.93 1 P43251 Biotinidase 0.23 0.6327 0.814  1.3E+08 1.42E+08 0.91 P01031 Complement C5 0.19 0.6677 0.814 1.85E+09 1.89E+09 0.98 Q86UX7 Fermitin family homolog 3 0.18 0.6701 0.849 76129860 87451916 0.87 P21333 Filamin-A 0.18 0.6763 0.850 38045207 40631822 0.94 P00488 Coagulation factor XIII A chain 0.13 0.7166 0.892 1.84E+08 1.91E+08 0.96 Q9UGM5 Fetuin-B 0.13 0.7198 0.892 1.65E+08 1.69E+08 0.98 P04278 Sex hormone-binding globulin 0.11 0.7403 0.910 93647907 1.08E+08 0.86 P30041 Peroxiredoxin-6 0.09 0.7667 0.927 17144828 17396439 0.99 P02745 Complement C1q 0.09 0.7703 0.927 7.05E+08 6.88E+08 1.02 subcomponent subunit B P02679 Fibrinogen gamma chain 0.08 0.772 0.927 4.42E+08 4.32E+08 1.02 Q9NPH3 Interleukin-1 receptor 0.08 0.7759 0.927 18337034 21353257 0.86 accessory protein P00338 L-lactate dehydrogenase A 0.07 0.7934 0.941  9971161  9783411 1.02 chain P07996 Thrombospondin-1 0.06 0.8058 0.944 92835344 1.01E+08 0.91 P25311 Zinc-alpha-2-glycoprotein 0.06 0.8074 0.944 1.72E+09  1.7E+09 1.01 Q9Y490 Talin-1 0.06 0.8149 0.946  1.4E+08 1.58E+08 0.89 P0CG06 Immunoglobulin lambda constant 3 0.05 0.8228 0.949 1.35E+08 1.54E+08 0.85 Q96IY4 Carboxypeptidase B2 0.05 0.8738 0.992 1.26E+08 1.31E+08 0.96 O00533 Neural cell adhesion molecule 0.02 0.8774 0.992 11480232 11161656 1.03 L1-like protein P08571 Monocyte differentiation 0.02 0.8899 0.992 1.27E+08 1.24E+08 1.03 antigen CD14 P02765 Alpha-2-HS-glycoprotein 0.02 0.8933 0.992 1.46E+09 1.44E+09 1.01 P01857 Immunoglobulin heavy 0.01 0.9034 0.992 1.56E+08 1.59E+08 0.98 constant gamma 1 P01008 Antithrombin-III 0.01 0.9108 0.992 5.16E+09  5.1E+09 1.01 P00915 Carbonic anhydrase 1 0.01 0.9123 0.992 54151372 51702715 1.05 Q6EMK4 Vasorin 0.01 0.9118 0.992 22619673 23533310 0.96 O00391 Sulfhydryl oxidase 1 0.01 0.9143 0.992 57356817 57950126 0.99 P22259 Coagulation factor V 0.01 0.9247 0.993 3.68E+08 3.68E+28 1.00 P07359 Platelet glycoprotein Ib alpha 0.01 0.9291 0.993 56061100 57453774 0.98 chain P06727 Apolipoprotein A-IV 0.01 0.9391 0.993 4.54E+09 4.49E+09 1.01 P26927 Hepatocyte growth factor-like 0 0.9457 0.993 1.21E+08 1.23E+08 0.98 protein P13796 Plastin-2 0 0.9503 0.993 34554768 34214663 1.01 P63104 14-3-3 protein zeta/delta 0 0.9538 0.993 28443038 28830798 0.99 P01623 Immunoglobulin kappa 0 0.9579 0.993 25042231 27111840 0.92 variable 3-20 P05154 Plasma serine protease 0 0.9629 0.993 1.55E+08 1.64E+08 0.95 inhibitor P0C0L4 Complement C4-A 0 0.9748 0.998 1.73E+10 1.71E+10 1.01 P22352 Glutathione peroxidase 3 0 0.9805 0.998 1.13E+08  1.2E+08 0.94 PL8206 Vinculin 0 0.9858 0.998 71329343 78664847 0.91 Q9UK55 Protein Z-dependent protease 0 0.9926 0.999 68425526 69087635 0.99 inhibitor P01009 Alpha-1-antitrypsin 0 0.9993 0.999 2.39E+08 2.77E+08 0.86 CHR-T: clinical high-risk participants who transitioned to first episode psychosis; CHR-NT: clinical high-risk participants who did not transition; FDR: false discovery rate

TABLE-US-00007 TABLE 7 Functional enrichment analyses: 6 KEGG pathways significantly enriched Fisher’s exact test p (adjusted Count in for false KEGG pathway gene set discovery rate) Complement and 13 of 78 2.23E−21 coagulation cascades Staphylococcus aureus infection 6 of 51 5.29E−09 Pertussis 4 of 74 6.38E−05 Cholesterol metabolism 3 of 48 0.00047 Systemic lupus erythematosus 3 of 94 0.0025 Prion diseases 2 of 33 0.0058

TABLE-US-00008 TABLE 8 Results of enzyme-linked immunoassay (ELISA) tests for CHR-T and CHR-NT in EU-GEI CHR-T CHR-NT mean (SD) mean (SD) t p Alpha-2-macroglobulin 1173.1 (459.1) 11501.7 (711.1) 3.2202 0.0016 (μg/ml) Apolipoprotein E 163751.3 (47433.8) 151740.6 (50903.5) −1.3449 0.1818 (ng/ml).sup.a Complement C1q 82811.7 (35347.1) 80204.6 (33535.8) −0.4153 0.6789 (ng/ml) Complement C1r 65008.9 (27901.6) 52803.9 (18481.6) −2.7099 0.0084 (μg/ml) Complement C4 495765.9 (222274.7) 482208.2 (192019.3) −0.3538 0.7243 binding protein (ng/ml) Complement C8 58233.5 (22885.2) 55706.0 (21938.2) −0.6196 0.5370 (ng/ml) Complement factor H 701713.1 (207717.4) 663292.1 (199397.1) −1.0112 0.3147 (ng/ml).sup.b Immunoglobulin M 1752941.0 (770671.5) 1941142.0 (936493.0) 1.2460 0.2153 (ng/ml) Plasminogen 206880.2 (73232.4) 176929.9 (62709.8) −2.3786 0.0196 (ng/ml) Data available for 48 CHR-T and 84 CHR-NT, except for .sup.aApolipoprotein E (46 CHR-T, 84 CHR-NT) and .sup.bComplement factor H (45 CHR-T, 82 CHR-NT). Means (and standard deviations) are presented and are compared using 2-sided t-test with unequal variances. CHR-T: clinical high-risk participants who transitioned to first episode psychosis; CHR-NT: clinical high-risk participants who did not transition; SD: standard deviation

TABLE-US-00009 TABLE 9 Comparison of characteristics for participants included in replication experiment (N = 135) from total EU-GEI CHR cohort (N = 344) Included, N = 135 Not included, N = 209 Missing (49 CHR-T, (CHR-T 16, data, n (%) 86 CHR-NT) CHR-NT 193) t/χ.sup.2 p Baseline age in 0 22.2 (4.8) 22.5 (5.0) −0.570 0.569 years, mean (SD) Sex, n (%) 0 82 male (60.7%) 103 male (49.3%) 4.322 0.037 53 female (39.3%) 106 female (50.7%) Baseline BMI in 50 (14.5%) 23.7 (4.1) 24.2 (5.8) −0.936 0.350 kg/m.sup.2, mean (SD) Baseline years 38 (11.0%) 14.3 (2.9) 14.5 (3.2) −0.574 0.566 in education, mean (SD) Ethnicity, n (%) 0 93 white (68.9%) 154 white (73.7%) 1.030 0.597 14 black (10.4%) 20 black (9.6%) 28 other (20.7%) 35 other (16.7%) Ever used 10 (2.9%) 101 yes (74.8%) 143 yes (68.4%) 1.328 0.249 cannabis, n (%) 31 no (23.0%) 59 no (28.2%) 3 not known (2.2%) 7 not known (3.3%) Baseline 95 (27.6%) 43 yes (31.9%) 45 yes (21.5%) 3.155 0.076 cannabis use, n 60 no (44.4%) 101 no (48.3%) (%) 32 not known (23.7%) 63 not known (30.1%) Baseline 38 (11.0%) 72 yes (53.3%) 89 yes (42.6%) 2.893 0.089 tobacco use, n 51 no (37.8%) 94 no (45.0%) (%) 12 not known (8.9%) 26 not known (12.4%) Baseline alcohol 12 (3.5%) 36 yes (26.7%) 137 yes (65.6%) 0.640 0.424 use, n (%) 97 no (71.9%) 62 no (29.7%) 2 not known (1.5%) 10 not known (4.8%) Baseline 86 (25.0%) 52 yes (38.5%) 77 yes (36.8%) 0.065 0.799 medication use, Antidepressant 30 Antidepressant 44 n (%) Antipsychotic 10 Antipsychotic 10 Hypnotic 3 Hypnotic 11 Other 9 Other 12 50 no (37.0%) 79 no (37.8%) 33 not known (24.4%) 53 not known (25.4%) Baseline GAF 27 (7.8%) 54.4 (10.2) 55.6 (10.0) −1.103 0.271 symptoms score, mean (SD) Baseline GAF 12 (3.5%) 55.5 (13.7) 55.4 (11.3) 0.006 0.996 disability score, mean (SD) Baseline SANS 45 (13.1%) 17.1 (12.7) 14.6 (10.8) 1.784 0.075 total composite score, mean (SD) Baseline SANS 29 (8.4%) 5.6 (3.8) 5.4 (3.5) 0.510 0.610 total global score, mean (SD) Baseline BPRS 25 (7.3%) 44.9 (11.2) 42.9 (9.6) 1.650 0.100 total score, mean (SD) Baseline 16 (4.7%) 18.8 (9.5) 18.9 (8.9) −0.126 0.900 MADRS total score, mean (SD) 2 year GAF 142 (41.3%) 54.6 (15.0) 63.0 (11.6) −4.083 <0.001 symptoms score, mean (SD) 2 year GAF 124 (36.0%) 56.9 (15.0) 63.6 (13.8) −3.333 0.001 disability score, mean (SD) 2 year GAF 124 (36.0%) 40 good (29.6%) 72 good (34.4%) 3.002 0.083 disability score, 51 poor (37.8%) 57 poor (27.2%) dichotomous 44 not known (32.6%) 80 not known (38.3%) outcome .sup.a .sup.a Poor functioning: GAF disability score ≤60; good functioning: GAF disability score >60 Tobacco use was defined as daily use for at least 1 month over the previous 12 months. Alcohol use was defined as at least 12 or more alcoholic beverages over the previous 12 months. Missing data excluded in hypothesis tests. EU-GEI: European Union Gene Environment Interaction study; CHR-T: clinical high risk, transitioned to psychosis; CHR-NT: clinical high risk, did not transition to psychosis; BMI: body mass index; GAF: General Assessment of Functioning; SANS: Scale for the Assessment of Negative Symptoms; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery Asberg Depression Rating Scale

TABLE-US-00010 TABLE 10 Descriptive statistics for EU-GEI replication CHR-T and CHR-NT groups Missing data, n CHR-T CHR-NT (%) N = 49 N = 86 t/χ.sup.2 p Baseline age 0 22.0 (4.7) 22.3 (4.9) t = −0.339 0.735 in years, mean (SD) Sex, n (%) 0 26 male (53%) 56 male (65%) χ.sup.2 = 1.902 0.168 23 female (47%) 30 female (35%) Baseline 21 24.5 (4.5) 23.2 (3.8) t = 1.722 0.088 BMI in (15.6%) kg/m.sup.2, mean (SD) Baseline 12 (8.8%) 14.0 (3.1) 14.3 (2.6) t = −0.573 0.568 years in education, mean (SD) Ethnicity, n 0 31 white (63.3%) 62 white (72.1%) χ.sup.2 = 8.549 0.014 (%) 10 black (20.4%) 4 black (4.7%) 8 other (16.3%) 20 other (23.3%) Ever used 3 (2.2%) 35 yes (71.4%) 66 yes (76.7%) χ.sup.2 = 0.170 0.680 cannabis, n 12 no (24.5%) 19 no (22.1%) (%) 2 not known (4.1%) 1 not known (1.2%) Baseline 32 14 yes (28.6%) 29 yes (33.7%) χ.sup.2 = 0.186 0.666 cannabis (23.7%) 22 no (44.9%) 38 no (44.2%) use, n (%) 13 not known (26.5%) 19 not known (22.1%) Baseline 12 (8.9%) 19 yes (38.8%) 53 yes (61.6%) χ.sup.2 = 4.647 0.031 tobacco use, 23 no (46.9%) 28 no (32.6%) n (%) 7 not known (14.3%) 5 not known (5.8%) Baseline 2 (1.5%) 34 yes (69.4%) 63 yes (73.3%) χ.sup.2 = 0.168 0.682 alcohol use, 14 no (28.6%) 22 no (25.6%) n (%) 1 not known (2.0%) 1 not known (1.2%) Baseline 33 20 yes (40.8%) 32 yes (37.2%) χ.sup.2 = 0.285 0.593 medication (24.4%) Antidepressant 12 Antidepressant 18 use, n (%) Antipsychotic 6 Antipsychotic 4 Hypnotic 1 Hypnotic 2 Other 1 Other 8 20 no (40.8%) 30 no (34.9%) 9 not known (18.4%) 24 not known (27.9%) Baseline 4 (3.0%) 53.0 (10.1) 55.1 (10.2) t = −1.089 0.278 GAF symptoms score, mean (SD) Baseline 4 (3.0%) 53.0 (12.6) 56.8 (14.1) t = −1.531 0.128 GAF disability score, mean (SD) Baseline 15 20.9 (14.1) 14.9 (11.4) t = 2.389 0.019 SANS total (11.1%) composite score, mean (SD) Baseline 10 (7.4%) 6.6 (4.1) 5.0 (3.5) t = 2.252 0.026 SANS total global score, mean (SD) Baseline 10 (7.4%) 48.1 (11.2) 43.1 (10.8) t = 2.456 0.015 BPRS total score, mean (SD) Baseline 6 (4.4%) 19.9 (10.2) 18.1 (9.1) t = 1.004 0.317 MADRS total score, mean (SD) 2 year GAF 49 43.6 (14.1) 63.5 (10.6) t = −7.281 <0.001 symptoms (36.3%) score, mean (SD) .sup.a 2 year GAF 44 45.3 (9.5) 65.1 (13.9) t = −7.969 <0.001 disability (32.6%) score, mean (SD) .sup.b 2 year GAF 44 28 poor functioning 23 poor functioning χ.sup.2 = 25.261 <0.001 disability (32.6%) (57.1%) (26.7%) score, 2 good functioning 38 good functioning dichotomous (4.1%) (44.2%) outcome .sup.c 19 not known (38.8%) 25 not known (29.1%) .sup.a Data available for 86 of 135 participants (CHR-NT n = 27, CHR-T n = 59) .sup.b Data available for 91 of 135 participants (CHR-NT n = 30, CHR-T n = 61) .sup.c Poor functioning: GAF disability score ≤60; good functioning: GAF disability score >60 Tobacco use was defined as daily use for at least 1 month over the previous 12 months. Alcohol use was defined as at least 12 or more alcoholic beverages over the previous 12 months. Missing data excluded in hypothesis tests. EU-GEI: European Union Gene Environment Interaction study; CHR-T: clinical high risk, transitioned to psychosis; CHR-NT: clinical high risk, did not transition to psychosis; BMI: body mass index; GAF: General Assessment of Functioning; SANS: Scale for the Assessment of Negative Symptoms; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery Asberg Depression Rating Scale

TABLE-US-00011 TABLE 11 Results of ANCOVA and fold changes (CHR-T vs. CHR-NT) for proteins identified in EU-GEI replication dataset (adjusted for age, sex, BMI, years in education, tobacco use and ethnicity) Mean LFQ, Mean LFQ, non- transition transition Mean 5% FDR- group group ratio T Uniprot no. Protein name F p adjusted p (n = 49) (n = 86) vs NT P01023 Alpha-2-macroglobulin 264.41 9.81E−33 1.17E−30 1.48E+10 6.14E+10 0.24 A0A075B6N9 Immunoglobulin heavy 109.92 7.11E−19 4.23E−17 1.19E+09 3.58E+09 0.33 constant mu P00747 Plasminogen 78.47 6.43E−15 2.55E−13 3.68E+10 2.83E+10 1.30 P10909 Clusterin 73341 3.17E−14 9.44E−13 1.28E+10 9.43E+09 1.36 P01011 Alpha-1-antichymotrypsin 68.88 1.38E−13 3.28E−12 3.57E+10 2.58E+10 1.38 G3XAW2 Complement factor I 65.48 4.24E−13  8.4E−12 9.37E+09 7.38E+09 1.27 A0A075B6L0 Immunoglobulin lambda 58.29 4.90E−12 8.33E−11 1.24E+09 2.22E+09 0.56 constant 3 A0A075B6N8 Immunogloblin heavy 56.09 1.06E−11 1.42E−10 4.02E+09 7.71E+09 0.52 constant gamma 3 P08603 Complement factor H (H 56.05 1.08E−11 1.42E−10 4.92E+10 4.18E+10 1.18 factor 1) P13671 Complement component C6 55.01 1.55E−11  1.7E−10 6.58E+09 4.74E+09 1.39 P04004 Vitronectin 54.97 1.57E−11  1.7E−10 2.17E+10 1.63E+10 1.33 P22792 Carboxypeptidase N subunit 52.23 4.18E−11 4.15E−10 3.69E+09 2.85E+09 1.29 2 P07360 Complement component C8 49.11 1.30E−10 1.19E−09 4.35E+09  3.4E+09 1.28 gamma chain P08571 Monocyte differentation 47.31 2.52E−10 2.14E−09 5.51E+08 3.95E+08 1.39 antigen CD14 P01031 Complement C5 44.69 6.71E−10 5.32E−29 1.36E+10 1.14E+10 1.19 B7ZXJ8 ITIH4 protein 43.28 1.14E−09 8.51E−29 3.08E+10 2.55E+10 1.21 P07359 Platelet glycoprotein Ib 35.1 2.79E−08 1.96E−07 2.84E+08 2.03E+08 1.40 alpha chain P15169 Carboxypeptidase N catalytic 33.9 4.53E−08 2.99E−07 1.75E+09 1.32E+09 1.33 chain (CPN) P36955 Pigment epithelium-derived 32.23 8.97E−08 5.62E−07 6.33E+09  5.2E+09 1.22 factor P07357 Complement component C8 31.63 1.15E−07 6.82E−07 3.98E+09 3.18E+09 1.25 alpha chain P27169 Serum 30.66 1.71E−07 9.69E−07 9.48E+08 7.27E+08 1.30 paraoxonase/arylesterase 1 Q96IY4 Carboxypeptidase B2 29.85 2.39E−07 1.29E−06 9.03E+08 6.39E+08 1.41 Q16610 Extracellular matrix protein 1 26.68 9.12E−07 4.72E−06 2.42E+09 1.75E+09 1.38 P04003 C4b-binding protein alpha 26.26 1.09E−06 5.41E−06  5.2E+09 6.76E+09 0.77 chain P12259 Coagulation factor V 24.43 2.40E−06 1.14E−05 1.12E+09 8.61E+03 1.30 P06727 Apolipoprotein A-IV 23.96 2.95E−06 1.35E−05 4.86E+10  4.2E+10 1.16 P80108 Phosphatidylinositol-glycan- 22.81 4.87E−06 2.15E−05 1.25E+09 9.86E+08 1.27 specific phospholipase D (PI- G PLD) P05155 Plasma protease C1 Inhibitor 22.67 5.19E−06  2.2E−05 1.61E+10  1.3E+10 1.23 P19827 Inter-alpha-trypsin inhibitor 20.64 1.28E−05 5.25E−05 2.21E+10 1.92E+10 1.15 heavy chain H1 P0C0L4 Complement C4-A (Acidic 19.53 2.11E−05 8.37E−05 3.94E+08 6.91E+08 0.57 complement C4) B4E1Z4 cDNA FLI55673 18.98 2.71E−05 0.000104  3.8E+10 3.36E+10 1.13 P51884 Lumican 18.9 2.81E−05 0.000104 3.37E+09 2.68E+09 1.26 P00736 Complement C1r 18.64 3.17E−05 0.000114 5.36E+09 4.75E+09 1.13 subcomponent O43866 CD5 antigen-like 18.47 3.42E−05 0.00012  2.39E+08 3.66E+08 0.65 P19823 Inter-alpha-trypsin inhibitor 15.64 0.0001 0.000313 4.08E+10 3.68E+10 1.11 heavy chain H2 P06396 Gelsolin 16.15 0.0001 0.000313 1.72E+10 1.49E+10 1.15 P02765 Alpha-2-HS-glycoprotein 17.11 0.0001 0.000313 2.82E+10 2.44E+10 1.16 Q04756 Hepatocyte growth factor 16.22 0.0001 0.000313 5.53E+08 4.65E+03 1.13 activator O00391 Sulfhydryl oxidase 1 14.82 0.0002 0.00061  3.36E+08 2.86E+03 1.18 P02748 Complement component C9 13.25 0.0004 0.00119  8.05E+09 6.74E+09 1.19 P00738 Haptoglobin 11.99 0.0007 0.001983 2.02E+09 4.13E+09 0.49 P02790 Hemopexin 11.97 0.0007 0.001983 7.36E+10 6.69E+10 1.10 P08697 Alpha-2-antiplasmm 11.36 0.001 0.002705 1.05E+10 9.47E+09 1.10 P23142 Fibulin-1 11.37 0.001 0.002705 1.15E+09 8.44E−08 1.37 P05452 Tetranectin (TN) 11.15 0.0011 0.002909 3.22E+09 2.79E+09 1.15 P06681 Complement C2 10.96 0.0012 0.003104 1.18E+09 9.97E+08 1.18 P02751 Fibronectin 10.62 0.0014 0.003545 5.96E+10  4.3E+10 1.39 P07558 Complement component C8 10.24 0.0017 0.004129 2.98E+03 2.53E+09 1.18 beta chain P02649 Apolipoprotein E 10.26 0.0017 0.004129  7.3E+09 6.04E+09 1.21 P01834 Immunoglobulin kappa 10.09 0.0019 0.004433 1.96E+03 2.65E+09 0.74 constant O75882 Attractin 10.07 0.0019 0.004433 1.55E+03 1.31E+09 1.19 P00734 Prothrombin 9.98 0.002 0.004577 3.74E+10 3.43E+10 1.03 P17936 Insulin-like growth factor- 9.46 0.0026 0.005838 8.49E+08 6.51E+03 1.30 binding protein 3 P00746 Complement factor D 9.15 0.003 0.006611 2.18E+08 1.77E+08 1.23 P02675 Fibrinogen beta chain 8.48 0.0043 0.005304 4.98E+09 5.74E+09 0.87 P00488 Coagulation factor XIII A 8.32 0.0046 0.005775 1.03E+03 8.21E+08 1.26 chain P43652 Afamin 8.04 0.0053 0.011065 1.65E+10  1.5E+10 1.10- P35858 Insulin-like growth factor- 7.33 0.0077 0 015758 6.06E+09 5.47E+09 1.11 binding protein complex acid labile subunit (ALS) P05543 T4-binding globulin 7.3 0.0079 0.015934 1.13E+09  8.9E+08 1.27 P10643 Complement component C7 7.09 0.0088 0.017453 2.99E+09 2.31E+09 1.25 P02787 Serotransferrin 6.92 0.0096 0.018728 6.56E+09  7.9E+09 0.83 P26927 Hepatocyte growth factor- 6.8 0.0102 0 019267 4.68E+08 3.79E+08 1.24 like protein (Macrophage stimulatory protein) P18428 Lipopolysaccharide-binding 6.81 0.0102 0.019267 5.69E+08 4.48E+08 1.27 protein (LBP) P02760 Protein AMBP 6.63 0.0112 0.020825 1.16E+10  1.1E+10 1.05 P00450 Ceruloplasmin 6.38 0.0128 0.023434 4.58E+10 4.92E+10 0.93 P05154 Plasma serine protease 6.25 0.0137 0.024702 7.96E+08 6.93E+08 1.15 inhibitor P02766 Transthyretin 6.14 0.0145 0.025754 4.93E+08 6.05E+08 0.82 Q14520-2 Hyaluronan-binding protein 6.09 0.0149 0.026042 1.38E+09 1.15E+09 1.20 2 P02679-2 Fibrinogen gamma chain 6.07 0.0151 0.026042 7.28E+09 8.08E+09 0.90 P00740 Coagulation factor IX 5.32 0.0164 0.02788  8.19E+08 7.41E+08 1.10 Q96XN2 Beta-Ala-His dipeptidase 5.82 0.0173 0.023956 8.09E+08 6.35E+08 1.27 P01019 Angiotensinogen 5.42 0.0215 0.035535 1.29E+10 1.15E+10 1.12 P04114 Apolipoprotein B-100 5.29 0.0231 0.037656 9.92E+13  9.3E+10 1.07 P02647 Apolipoprotein A-I 5.03 0.0266 0.042776 6.21E+09 6.98E+09 0.89 P05546 Heparin cofactor 2 4.81 0.0302 0.047917 1.36E+10 1.29E+10 1.06 P01009 Alpha-1-antitryosin 4.77 0.0307 0.04807  1.48E+09 1.86E+09 0.79 P04275 von Willebrand factor 4.71 0.0315 0.0493  4.87E+08 3.55E+08 1.37 P02743 Serum amyloid P-component 4.68 0.0324 0.049431  6.6E+03 5.42E+09 1.22 (SAP) P01857 Immunoglobulin heavy 4.63 0.0334 0.050311  2.3E+09 2.02E+09 1.14 constant gamma 1 P01008 Antithrombin-III 4.27 0.0408 0.06069  2.16E+13 2.38E+10 0.91 P25311 Zinc-alpha-2-glyprotein 4.08 0.0456 0.066953 1.41E+10 1.39E+10 1.01 P05160 Coagulation factor XIII B chain 3.4 0.0675 0.097957 1.21E+09 1.14E+09 1.06 P22891 Vitamin K-dependent protein 3.13 0.0795 0.113982  2.6E+08 2.35E+08 1.10 Z P02746 Complement C1q 2.85 0.0936 0.1326  3.71E+09 3.17E+09 1.17 subcomponent subunit B P0C0L5 Complement C4-B (Basic 2.72 0.1014 0.14196  9.89E+10 9.03E+10 1.09 complement C4) Q96PD5 N-acetylmuramoyl-L-alanine 2.63 0.1075 0.14875  5.13E+09 4.87E+09 1.05 amidase P01876 Immunoglobulin heavy 2.54 0.1135 0.154566 1.97E+09 2.73E+09 0.72 constant alpha 1 P02763 Alpha-1-acid glycoprotein 1 2.51 0.1154 0.154566 8.56E+08 1.02E+09 0.84 P01042 Kininogen-1 2.51 0.1156 0.154566 2.64E+10 2.55E+10 1.03 K7ERI9 Apolipoprotein C-I 2.17 0.1433 0.189474 1.06E+09 1.26E+09 0.84 Q06033 Inter-alpha-trypsin inhibitor 2.07 0.1524 0.199292 1.79E+09 1.59E+09 1.13 heavy chain H3 P02750 Leucine-rich alpha-2- 2.01 0.1589 0.205534 5.42E+09 5.03E+09 1.08 glycoprotein D6RF35 Vitamin D-binding protein 1.89 0.1721 0.220214 4.27E+10 4.19E+10 1.02 P00748 Coagulation factor XII 1.85 0.1762 0.223062  4.2E+09 4.54E+09 0.93 P09871 Complement C1q 1.7 0.1951 0.244388 4.23E+09 4.58E+09 0.92 subcomponent P07996 Thrombospondin-1 1.11 0.2932 0.363446 4.62E+08 4.19E+08 1.10- Q9NZP8 Complement C1r 1.06 0.3051 0.374298 3.33E+08 2.99E+08 1.12 subcomponent-like protein P02747 Complement C1q 1.02 0.3148 0.382257 3.96E+09 3.78E+09 1.05 subcomponent subunit C P04278 Sex hormone-binding 0.64 0.4244 0.507178 7.31E+08 6.27E+08 1.17 globulin P29622 Kallistatin (Kallikrein inhibitor) 0.64 0.4262 0.507178  3.1E+09 3.04E+09 1.02 P03952 Plasma kallikrein 0.58 0.4476 0.52737   4.4E+09 4.06E+09 1.08 P43251 Biotinidase (Biotinase) 0.55 0.4607 0.537483 4.38E+08 4.78E+08 0.92 P01024 Complement C3 0.45 0.5025 0.577493 2.13E+09 2.11E+09 1.01 P04196 Histidine-rich glycoprotein 0.45 0.5047 0.577493 2.15E+10 1.13E+10 1.02 P02656 Apolipoprotein C-III 0.43 0.5145 0.5831  2.95E+09  2.7E+09 1.09 O75636 Ficolin-3 0.36 0.5475 0.614646 1.74E+09 1.58E+09 1.10 P27918 Properdin 0.3 0.5819 0.64716  7.22E+08 6.85E+08 1.05 P00742 Coagulation factor X 0.27 0.607 0.668824 1.04E+09 9.82E+08 1.06 P02749 Beta-2-glycoprotein 1 0.23 0.6357 0.694021 1.96E+10 1.95E+10 1.01 P06276 Acylcholine acylhydrolase 0.2 0.6588 0.712702 5.34E+08 5.29E+08 1.01 P08185 Corticosteroid-binding 0.13 0.717 0.768676 1.72E+09 1.71E+09 1.00 globulin P02671 Fibrinogen alpha chain 0.08 0.7845 0.833531 6.62E+09 6.82E+09 0.97 P03951 Coagulation factor XI 0.05 0.8228 0.86275  2.95E+08 2.63E+08 1.12 P19652 Alpha-1-acid glycoprotein 2 0.05 0.8265 0.86275  1.44E+08 1.48E+08 0.97 (AGP 2) Q9UK55 Protein Z-dependent protease inhibitor 0.04 0.8393 0.868493 5.16E+08 5.13E+08 1.01 P07225 Vitamin K-dependent protein 0.03 0.8521 0.874137  2.7E+09  2.7E+09 1.03 S Q5VY30 Retinol-binding protein 0.02 0.8787 0.893721   3E+09 3.08E+09 0.97 P68871 Hemogloblin subunit beta 0.02 0.9022 0.9079  1.09E+09 1.03E+09 1.06 P22352 Glutathione peroxidase 5 0.01 0.9079 0.9079  3.78E+08 3.68E+08 1.03 (GPx-3) CHR-T: clinical high-risk participants who transitioned to first episode psychosis; CHR-NT: clinical high-risk participants who did not transition; FDR: false discovery rate

TABLE-US-00012 TABLE 12 Ten percent highest-weighted features for support vector machine models in EU-GEI replication dataset (Model 4) according to mean feature weight for models selected in cross-validation inner loop Model 4: Replication model (119 proteomic and 65 clinical features) Mean Feature weight Alpha-2-macroglobulin −0.285 Carboxypeptidase N subunit 2 0.214 Complement C1s subcomponent −0.187 Immunoglobulin heavy constant mu −0.173 Alpha 1 anti-chymotrypsin 0.170 Plasminogen 0.163 Zinc alpha-2-glycoprotein −0.162 Clusterin 0.162 C4b binding protein alpha chain −0.157 Monocyte differentiation antigen CD14 0.155 Extracellular matrix protein 1 0.142 Attractin 0.128 Complement component 6 0.126 Complement factor 1 0.119 BPRS: bizarre behaviour 0.112 Immunoglobulin lambda constant 3 −0.111 Ceruloplasmin −0.109 Antithrombin III −0.109

TABLE-US-00013 TABLE 13 Descriptive statistics for ALSPAC cases and controls Cases Controls N = 55 N = 66 t/χ.sup.2 p Sex, n (%) 22 male (40.0%) 39 male (59.1%) χ.sup.2 = 4.374 0.036 33 female (60.0%) 27 female (40.9%) BMI age 12 18.1 (2.8) 17.7 (2.5) t = 0.749 0.455 in kg/m.sup.2, mean (SD) Ethnicity 50 white (90.9%) 63 white (95.5%) χ.sup.2 = 0.729 0.202 5 other/not known 3 other/not known (9.1%) (4.5%) Maternal 40 non-manual 44 non-manual χ.sup.2 = 0.005 0.946 social (72.7%) (66.7%) class 7 manual (12.7%) 8 manual (12.1%) 8 not known 14 not known (14.6%) (21.2%) Cases: participants with no PEs age 12 and definite PEs age 18; Controls: participants with no PEs age 12 and no PEs age 18 PEs: psychotic experiences; BMI: body mass index Missing data excluded in hypothesis tests.

TABLE-US-00014 TABLE 14 Results of ANCOVA and fold changes (definite PEs at 18 vs. no PEs at 18) for proteins identified in ALSPAC age 12 plasma samples (adjusted for sex, BMI and maternal social class) Fold change Uniprot Adjusted p (PE vs. no No. Protein name F p (FDR 5%) PE) P04003 C4b-binding protein alpha chain 24.59 2.44E−06 0.000647 0.77 P27169 Serum paraoxonase/arylesterase 1 17.78 4.94E−05 0.006544 0.80 P01871 Immunoglobulin heavy constant mu 16.04 0.0001 0.006625 0.78 P55103 Inhibin beta C chain 16.68 0.0001 0.006625 1.31 P10909 Clusterin 12.71 0.0005 0.0265 0.92 P01591 Immunoglobulin J chain 9.69 0.0023 0.1007 0.70 P01860 Immunoglobulin heavy constant gamma 3 9.25 0.0029 0.1007 0.80 P0DOY3 Immunoglobulin lambda constant 3 8.9 0.0035 0.1007 0.81 P07225 Vitamin K-dependent protein S 8.83 0.0036 0.1007 0.90 Q03591 Complement factor H-related protein 1 8.74 0.0038 0.1007 0.79 P01023 Alpha-2-macroglobulin 7.92 0.0057 0.137318 0.85 P01623 Immunoglobulin kappa variable 3-20 7.49 0.0072 0.159 0.81 P24593 Insulin-like growth factor-binding protein 5 7.34 0.0078 0.159 1.26 P01019 Angiotensinogen 6.74 0.0106 0.194333 0.91 P26038 Moesin 6.68 0.011 0.194333 1.16 P04040 Catalase 6.53 0.0119 0.197094 1.36 P12109 Collagen alpha-1 6.36 0.013 0.19875 1.16 P08571 Monocyte differentiation antigen CD14 6.29 0.0135 0.19875 0.91 B9A064 Immunoglobulin lambda-like polypeptide 5 6.03 0.0155 0.210675 0.85 P09871 Complement C1s subcomponent 5.99 0.0159 0.210675 0.93 P08697 Alpha-2-antiplasmin 5.5 0.0207 0.249542 0.94 P15151 Poliovirus receptor 5.44 0.0214 0.249542 1.16 P01717 Immunoglobulin lambda variable 3-25 5.34 0.0226 0.249542 0.85 Q12884 Prolyl endopeptidase FAP 5.34 0.0226 0.249542 1.24 P00746 Complement factor D 5.16 0.0249 0.255364 1.12 P01615 Immunoglobulin kappa variable 2D-28 5.1 0.0258 0.255364 0.84 P02671 Fibrinogen alpha chain 5.05 0.0266 0.255364 0.89 P23142 Fibulin-1 4.99 0.0275 0.255364 1.36 P01834 Immunoglobulin kappa constant 4.95 0.028 0.255364 0.90 P24592 Insulin-like growth factor-binding protein 6 4.85 0.0297 0.255364 1.15 P02679 Fibrinogen gamma chain 4.77 0.031 0.255364 0.89 P02675 Fibrinogen beta chain 4.76 0.0311 0.255364 0.90 P03951 Coagulation factor XI 4.72 0.0318 0.255364 0.91 P80748 Immunoglobulin lambda variable 3-21 4.63 0.0334 0.260324 0.85 P55290 Cadherin-13 4.43 0.0374 0.283171 0.88 P02652 Apolipoprotein A-II 4.37 0.0388 0.285611 0.85 P07358 Complement component C8 beta chain 4.12 0.0446 0.306075 1.08 D6RAR4 Hepatocyte growth factor activator 4.11 0.045 0.306075 0.89 P15144 Aminopeptidase N 4.07 0.046 0.306075 1.06 Q99878 Histone H2A type 1-J 4.06 0.0462 0.306075 1.37 P08294 Extracellular superoxide dismutase [Cu—Zn] 3.91 0.0505 0.326402 1.13 P02750 Leucine-rich alpha-2-glycoprotein 3.83 0.0528 0.333143 0.92 P61626 Lysozyme C 3.61 0.06 0.368644 0.91 P12111 Collagen alpha-3 3.55 0.0621 0.368644 1.09 P00915 Carbonic anhydrase 1 3.53 0.0626 0.368644 1.24 P14151 L-selectin 3.35 0.07 0.403261 1.08 O43866 CD5 antigen-like 3.26 0.0737 0.415543 0.86 P02654 Apolipoprotein C-I 3.14 0.0791 0.421962 0.91 P01024 Complement C3 3.09 0.0814 0.421962 0.94 P29622 Kallistatin 3.08 0.082 0.421962 1.11 P02647 Apolipoprotein A-I 3.07 0.0823 0.421962 0.86 Q14624 Inter-alpha-trypsin inhibitor heavy chain H4 3.06 0.0828 0.421962 0.91 O00187 Mannan-binding lectin serine protease 2 2.92 0.0901 0.440382 1.07 P16403 Histone H1.2 2.91 0.0907 0.440382 1.19 P06276 Cholinesterase 2.9 0.0914 0.440382 1.03 Q9NQ79 Cartilage acidic protein 1 2.84 0.0946 0.443526 1.08 Q9H4A9 Dipeptidase 2 2.83 0.0954 0.443526 0.84 H0Y755 Low affinity immunoglobulin gamma Fc 2.74 0.1008 0.460552 1.11 region receptor III-A Q9UBQ6 Exostosin-like 2 2.5 0.1164 0.514185 1.08 Q9UGM5 Fetuin-B 2.48 0.1178 0.514185 0.94 O75636 Ficolin-3 2.47 0.1191 0.514185 1.07 O95497 Pantetheinase 2.45 0.1203 0.514185 1.15 P43251 Biotinidase 2.38 0.1254 0.52197 1.06 Q08380 Galectin-3-binding protein 2.36 0.1276 0.52197 1.08 P63261 Actin, cytoplasmic 2 2.34 0.1289 0.52197 1.10 Q9Y5Y7 Lymphatic vessel endothelial hyaluronic 2.33 0.13 0.52197 1.09 acid receptor l P00747 Plasminogen 2.26 0.1358 0.536145 1.07 P22792 Carboxypeptidase N subunit 2 2.21 0.1395 0.536145 0.95 P01033 Metalloproteinase inhibitor 1 2.21 0.1396 0.536145 0.87 Q9ULI3 Protein HEG homolog 1 2.11 0.1488 0.563314 1.13 Q96PD5 N-acetylmuramoyl-L-alanine amidase 2.06 0.1539 0.56376 1.09 P05019 Insulin-like growth factor I 2.06 0.154 0.56376 1.24 P35858 Insulin-like growth factor-binding protein 2.05 0.1553 0.56376 1.10 complex acid labile subunit P02753 Retinol-binding protein 4 2.01 0.1592 0.570108 1.13 P02452 Collagen alpha-1 (I) chain 1.9 0.1706 0.592292 1.18 P39060 Collagen alpha-1 (XVIII) chain 1.9 0.1707 0.592292 1.06 P02747 Complement C1q subcomponent subunit C 1.89 0.1721 0.592292 0.96 P04004 Vitronectin 1.78 0.1853 0.629545 1.06 P07942 Laminin subunit beta-1 1.75 0.1887 0.632981 1.01 P02745 Complement C1q subcomponent subunit A 1.7 0.1944 0.64395 0.93 P07359 Platelet glycoprotein Ib alpha chain 1.65 0.201 0.653451 1.03 P07737 Profilin-1 1.65 0.2022 0.653451 1.17 P00751 Complement factor B 1.58 0.2116 0.67559 0.97 H0YD13 CD44 antigen 1.53 0.2185 0.681773 1.09 P48740 Mannan-binding lectin serine protease 1 1.53 0.2192 0.681773 0.91 P07357 Complement component C8 alpha chain 1.5 0.2227 0.681773 1.05 P08603 Complement factor H 1.48 0.2263 0.681773 1.04 Q16706 Alpha-mannosidase 2 1.48 0.2264 0.681773 0.92 P20851 C4b-binding protein beta chain 1.46 0.229 0.681854 0.89 Q12913 Receptor-type tyrosine-protein phosphatase eta 1.43 0.2341 0.686962 0.93 Q15582 Transforming growth factor-beta-induced 1.42 0.2359 0.686962 1.04 protein ig-h3 Q6UXB8 Peptidase inhibitor 16 1.39 0.2407 0.693321 1.09 P69905 Hemoglobin subunit alpha 1.31 0.2553 0.705108 3.63 F5GZZ9 Scavenger receptor cysteine-rich type 1 protein 1.29 0.2586 0.705108 1.20 M130 P13591 Neural cell adhesion molecule 1 1.27 0.2619 0.705108 1.03 Q15113 Procollagen C-endopeptidase enhancer 1 1.26 0.2633 0.705108 1.07 P26927 Hepatocyte growth factor-like protein 1.24 0.2671 0.705108 1.03 Q12841 Follistatin-related protein 1 1.24 0.2675 0.705108 1.04 Q15848 Adiponectin 1.24 0.2687 0.705108 0.95 P27487 Dipeptidyl peptidase 4 1.23 0.2706 0.705108 1.06 Q86U17 Serpin A11 1.22 0.2711 0.705108 1.11 Q9UNW1 Multiple inositol polyphosphate phosphatase 1 1.22 0.2714 0.705108 0.93 P43652 Afamin 1.2 0.2758 0.709583 1.07 P51884 Lumican 1.18 0.2798 0.712952 1.04 P13671 Complement component Complement 1.15 0.2868 0.719463 0.99 component 6 P04278 Sex hormone-binding globulin 1.14 0.2883 0.719463 0.89 P03950 Angiogenin 1.13 0.2905 0.719463 0.96 E9PBC5 Plasma kallikrein 1.1 0.2973 0.729486 0.96 P04275 von Willebrand factor 1.07 0.303 0.730918 0.93 P02748 Complement component C9 1.07 0.3034 0.730918 0.97 P08185 Corticosteroid-binding globulin 1.02 0.3139 0.749401 1.02 P14625 Endoplasmin 1 0.3189 0.75454 1.05 O75882 Attractin 0.97 0.3256 0.763575 1.02 P68871 Hemoglobin subunit beta 0.96 0.33 0.767105 2.75 P11021 Endoplasmic reticulum chaperone BiP 0.94 0.3342 0.770113 0.96 P02787 Serotransferrin 0.9 0.3459 0.786167 0.97 P18206 Vinculin 0.89 0.3471 0.786167 1.08 Q5T7F0 Neuropilin 0.87 0.3521 0.790733 0.96 P07360 Complement component C8 gamma chain 0.83 0.3655 0.813929 1.04 P55058 Phospholipid transfer protein 0.78 0.3783 0.826519 1.05 P01765 Immunoglobulin heavy variable 3-23 0.77 0.3806 0.826519 0.89 P13727 Bone marrow proteoglycan 0.74 0.39 0.826519 1.09 P12259 Coagulation factor V 0.73 0.3962 0.826519 0.97 P00742 Coagulation factor X 0.72 0.397 0.826519 1.03 P18065 Insulin-like growth factor-binding protein 2 0.7 0.4043 0.826519 0.92 P54289 Voltage-dependent calcium channel subunit 0.7 0.406 0.826519 1.03 alpha-2/delta-1 P13473 Lysosome-associated membrane glycoprotein 2 0.7 0.4061 0.826519 0.92 P32119 Peroxiredoxin-2 0.69 0.4084 0.826519 1.02 P04196 Histidine-rich glycoprotein 0.69 0.4087 0.826519 1.05 P02649 Apolipoprotein E 0.69 0.4092 0.826519 0.96 P02749 Beta-2-glycoprotein 1 0.68 0.4107 0.826519 0.98 P02655 Apolipoprotein C-II 0.68 0.4117 0.826519 1.06 P00734 Prothrombin 0.65 0.4218 0.840429 0.97 Q9UNN8 Endothelial protein C receptor 0.64 0.4267 0.843847 1.10 P02743 Serum amyloid P-component 0.62 0.4319 0.847804 0.98 P08709 Coagulation factor VII 0.59 0.4429 0.859138 1.02 P02746 Complement C1q subcomponent subunit B 0.59 0.4448 0.859138 1.05 P04180 Phosphatidylcholine-sterol acyltransferase 0.58 0.4474 0.859138 0.95 P24821 Tenascin 0.55 0.4606 0.875636 1.05 P01042 Kininogen-1 0.54 0.4626 0.875636 0.98 P23470 Receptor-type tyrosine-protein phosphatase 0.53 0.468 0.879133 1.00 gamma P07195 L-lactate dehydrogenase B chain 0.52 0.4712 0.879133 0.95 Q13822 Ectonucleotide pyrophosphatase/ 0.51 0.4744 0.879133 1.02 phosphodiesterase family member 2 P49908 Selenoprotein P 0.5 0.481 0.885174 1.00 P04114 Apolipoprotein B-100 0.49 0.4845 0.885466 0.99 P02766 Transthyretin 0.47 0.4941 0.890905 0.95 P27918 Properdin 0.47 0.4942 0.890905 1.05 H0Y897 Target of Nesh-SH3 0.44 0.5101 0.902403 1.06 P01344 Insulin-like growth factor II 0.44 0.5104 0.902403 0.97 P01031 Complement C5 0.42 0.5163 0.902403 1.01 P02776 Platelet factor 4 0.41 0.5214 0.902403 0.94 P06727 Apolipoprotein A-IV 0.41 0.5222 0.902403 1.01 P05155 Plasma protease C1 inhibitor 0.41 0.5226 0.902403 0.96 P15169 Carboxypeptidase N catalytic chain 0.41 0.5251 0.902403 0.97 Q9BXR6 Complement factor H-related protein 5 0.39 0.5319 0.902403 1.00 P02656 Apolipoprotein C-III 0.39 0.5337 0.902403 0.95 Q16270 Insulin-like growth factor-binding protein 7 0.39 0.5361 0.902403 0.96 095445 Apolipoprotein M 0.37 0.5426 0.902403 0.98 P00488 Coagulation factor XIII A chain 0.37 0.5441 0.902403 1.01 Q76LX8 A disintegrin and metalloproteinase with 0.36 0.5492 0.902403 0.97 thrombospondin motifs 13 Q8IUL8 Cartilage intermediate layer protein 2 0.35 0.5553 0.902403 1.07 P61769 Beta-2-micro globulin 0.34 0.5628 0.902403 1.05 Q99983 Osteomodulin 0.33 0.5661 0.902403 1.04 P54802 Alpha-N-acetylglucosaminidase 0.32 0.5723 0.902403 1.11 P11226 Mannose-binding protein C 0.32 0.5724 0.902403 0.98 P05543 Thyroxine-binding globulin 0.32 0.5731 0.902403 0.99 P10721 Mast/stem cell growth factor receptor Kit 0.32 0.5743 0.902403 1.01 Q14520 Hyaluronan-binding protein 2 0.31 0.5759 0.902403 0.98 P01861 Immunoglobulin heavy constant gamma 4 0.31 0.5783 0.902403 0.82 P00748 Coagulation factor XII 0.31 0.5789 0.902403 0.98 P09172 Dopamine beta-hydroxylase 0.29 0.5883 0.908087 0.97 Q16610 Extracellular matrix protein 1 0.29 0.5894 0.908087 0.98 P36980 Complement factor H-related protein 2 0.28 0.5952 0.90842 0.91 Q9UHG3 Prenylcysteine oxidase 1 0.28 0.5982 0.90842 1.02 P01008 Antithrombin-III 0.28 0.5999 0.90842 0.98 P01009 Alpha-1-antitrypsin 0.24 0.6231 0.935734 0.99 P33151 Cadherin-5 0.24 0.625 0.935734 0.96 P17936 Insulin-like growth factor-binding protein 3 0.23 0.6329 0.941564 1.06 P01876 Immunoglobulin heavy constant alpha 1 0.23 0.636 0.941564 0.96 P22352 Glutathione peroxidase 3 0.22 0.6413 0.943898 1.00 P01034 Cystatin-C 0.21 0.6447 0.943898 1.03 P28827 Receptor-type tyrosine-protein phosphatase mu 0.2 0.6555 0.95038 0.98 P36955 Pigment epithelium-derived factor 0.2 0.6563 0.95038 1.00 P59666 Neutrophil defensin 3 0.19 0.6605 0.951264 0.99 P18428 Lipopolysaccharide-binding protein 0.18 0.6698 0.959443 1.05 P43121 Cell surface glycoprotein MUC18 0.17 0.6784 0.966538 0.98 P19652 Alpha-1-acid glycoprotein 2 0.15 0.7034 0.979948 1.13 P04075 Fructose-bisphosphate aldolase A 0.14 0.7057 0.979948 0.97 Q6EMK4 Vasorin 0.14 0.7064 0.979948 0.99 Q14515 SPARC-like protein 1 0.14 0.7085 0.979948 1.05 P05109 Protein S100-A8 0.14 0.7099 0.979948 1.01 P05546 Heparin cofactor 2 0.14 0.71 0.979948 1.03 P49747 Cartilage oligomeric matrix protein 0.13 0.7175 0.981183 1.00 Q9NPY3 Complement component C1q receptor 0.13 0.7183 0.981183 1.01 Q9Y4L1 Hypoxia up-regulated protein 1 0.12 0.7318 0.992682 1.02 P02774 Vitamin D-binding protein 0.11 0.746 0.992682 0.97 P00736 Complement C1r subcomponent 0.1 0.7467 0.992682 0.99 P19827 Inter-alpha-trypsin inhibitor heavy chain H1 0.1 0.749 0.992682 0.98 P06702 Protein S100-A9 0.1 0.7506 0.992682 1.00 O00391 Sulfhydryl oxidase 1 0.1 0.7547 0.992682 1.01 Q92820 Gamma-glutamyl hydrolase 0.09 0.759 0.992682 1.03 P80108 Phosphatidylinositol-glycan-specific 0.09 0.7649 0.992682 1.01 phospholipase D Q9NPH3 Interleukin-1 receptor accessory protein 0.09 0.7649 0.992682 1.06 P01011 Alpha-1-antichymotrypsin 0.08 0.7712 0.992682 0.98 P35542 Serum amyloid A-4 protein 0.08 0.7743 0.992682 1.03 P02763 Alpha-1-acid glycoprotein 1 0.08 0.7752 0.992682 0.90 P20742 Pregnancy zone protein 0.08 0.7787 0.992682 1.05 P02775 Platelet basic protein 0.08 0.7815 0.992682 0.99 O95479 GDH/6PGL endoplasmic bifunctional protein 0.08 0.7835 0.992682 0.99 P05362 Intercellular adhesion molecule 1 0.07 0.7875 0.992682 0.95 O14791 Apolipoprotein L1 0.07 0.7904 0.992682 1.06 P33908 Mannosyl-oligosaccharide 1,2-alpha- 0.07 0.7947 0.993375 0.98 mannosidase IA P22891 Vitamin K-dependent protein Z 0.06 0.801 0.993379 0.91 Q13740 CD166 antigen 0.06 0.8022 0.993379 0.97 P22105 Tenascin-X 0.05 0.816 0.9976 1.00 P05156 Complement factor I 0.05 0.8205 0.9976 1.01 Q96KN2 Beta-Ala-His dipeptidase 0.05 0.8245 0.9976 1.03 P10643 Complement component C7 0.04 0.8343 0.9976 1.00 O00533 Neural cell adhesion molecule L1-like protein 0.04 0.8359 0.9976 0.97 Q9BWP8 Collectin-11 0.04 0.8409 0.9976 0.98 Q99969 Retinoic acid receptor responder protein 2 0.04 0.8425 0.9976 0.99 P04070 Vitamin K-dependent protein C 0.04 0.8443 0.9976 1.02 P04066 Tissue alpha-L-fucosidase 0.04 0.8513 0.9976 0.94 P55056 Apolipoprotein C-IV 0.03 0.8585 0.9976 1.12 P0C0L5 Complement C4-B 0.03 0.8646 0.9976 0.99 P00450 Ceruloplasmin 0.03 0.8653 0.9976 0.99 P02760 Protein AMBP 0.03 0.8721 0.9976 1.00 P19320 Vascular cell adhesion protein 1 0.02 0.884 0.9976 0.98 P22692 Insulin-like growth factor-binding protein 4 0.02 0.8848 0.9976 0.59 P02751 Fibronectin 0.02 0.8998 0.9976 0.97 P08519 Apolipoprotein 0.01 0.9049 0.9976 1.20 P12830 Cadherin-1 0.01 0.9146 0.9976 0.94 Q96IY4 Carboxypeptidase B2 0.01 0.919 0.9976 1.01 P17813 Endoglin 0.01 0.9194 0.9976 0.97 P35443 Thrombospondin-4 0.01 0.9285 0.9976 1.11 P98160 Basement membrane-specific heparan sulfate 0.01 0.9306 0.9976 0.98 proteoglycan core protein P00738 Haptoglobin 0.01 0.9311 0.9976 0.98 P19823 Inter-alpha-trypsin inhibitor heavy chain H2 0.01 0.9331 0.9976 1.00 P02790 Hemopexin 0.01 0.9342 0.9976 1.00 Q06033 Inter-alpha-trypsin inhibitor heavy chain H3 0.01 0.9358 0.9976 0.97 Q07954 Prolow-density lipoprotein receptor-related 0.01 0.9388 0.9976 0.98 protein 1 P0C0L4 Complement C4-A 0.01 0.939 0.9976 1.03 P00740 Coagulation factor IX 0 0.9441 0.9976 1.01 P07333 Macrophage colony-stimulating factor 1 0 0.9442 0.9976 0.98 receptor Q92954 Proteoglycan 4 0 0.9448 0.9976 1.01 Q6YHK3 CD109 antigen 0 0.9453 0.9976 0.97 Q99784 Noelin 0 0.949 0.9976 1.00 Q01459 Di-N-acetylchitobiase 0 0.9502 0.9976 0.99 Q9Y6R7 IgGFc-binding protein 0 0.9507 0.9976 1.06 Q9NZP8 Complement C1r subcomponent-like protein 0 0.9535 0.9976 1.01 P04217 Alpha-1B-glycoprotein 0 0.955 0.9976 1.00 K7EMN2 6-phosphogluconate dehydrogenase, 0 0.959 0.9976 0.95 decarboxylating G3V2W1 Protein Z-dependent protease inhibitor 0 0.9607 0.9976 1.00 P12955 Xaa-Pro dipeptidase 0 0.961 0.9976 0.94 P06681 Complement C2 0 0.9688 0.9976 1.02 B7ZKJ8 ITIH4 protein 0 0.9701 0.9976 0.99 P05154 Plasma serine protease inhibitor 0 0.9707 0.9976 1.00 P13796 Plastin-2 0 0.9799 0.9976 0.98 P09486 SPARC 0 0.9809 0.9976 0.95 P02765 Alpha-2-HS-glycoprotein 0 0.9812 0.9976 1.01 P05160 Coagulation factor XIII B chain 0 0.9826 0.9976 1.00 P54108 Cysteine-rich secretory protein 3 0 0.9883 0.9976 0.98 Q16853 Membrane primary amine oxidase 0 0.9943 0.9976 0.98 P05090 Apolipoprotein D 0 0.9949 0.9976 0.92 Q10588 ADP-ribosyl cyclase/cyclic ADP-ribose 0 0.9976 0.9976 0.97 hydrolase 2 PEs: psychotic experiences; FDR: false discovery rate; ALSPAC: Avon Longitudinal Study of Parents and Children

TABLE-US-00015 TABLE 15 Fold changes in poor functioning vs. good functioning Ratio of means Mean LFQ, Mean LFQ, (poor Uniprot poor good vs No. Protein name functioning functioning good) P63104 14-3-3 protein zeta/delta 27324340 37927971 0.72 Q76LX8 A disintegrin and metalloproteinase 10308467 11282881 0.91 with thrombospondin motifs 13 P60709 Actin, cytoplasmic 1 2.91E+08 3.96E+08 0.73 P43652 Afamin 1.86E+09 1.71E+09 1.09 P02763 Alpha-1-acid glycoprotein 1 1.66E+08 1.86E+08 0.89 P01011 Alpha-1-antichymotrypsin 5.35E+09  5.3E+09 1.01 P01009 Alpha-1-antitrypsin 2.63E+08 3.19E+08 0.83 P04217 Alpha-1B-glycoprotein 3.68E+09 3.46E+09 1.06 P08697 Alpha-2-antiplasmin 1.18E+09 1.17E+09 1.01 P02765 Alpha-2-HS-glycoprotein 1.45E+09 1.37E+09 1.06 P01023 Alpha-2-macroglobulin 1.74E+09 3.47E+09 0.50 P12814 Alpha-actinin-1 34823635 52957479 0.66 P02489 Alpha-crystallin A chain 1.46E+08 1.26E+08 1.15 P15144 Aminopeptidase N 13227156 12992373 1.02 P01019 Angiotensinogen 3.93E+09 3.46E+09 1.14 P01008 Antithrombin-III 5.28E+09  4.9E+09 1.08 P02647 Apolipoprotein A-I 9.24E+08 1.18E+09 0.78 P02652 Apolipoprotein A-II 3.14E+08  4.2E+08 0.75 P06727 Apolipoprotein A-IV 4.79E+09 4.46E+09 1.07 P04114 Apolipoprotein B-100 2.04E+10   2E+10 1.02 P02654 Apolipoprotein C-I 1.65E+08 1.69E+08 0.97 P02656 Apolipoprotein C-III 3.97E+08 2.74E+08 1.45 P05090 Apolipoprotein D 31119194 28321812 1.10 P02649 Apolipoprotein E 7.21E+08 7.01E+08 1.03 O75882 Attractin 1.02E+08 86797031 1.18 P02749 Beta-2-glycoprotein 1 5.14E+09 4.43E+09 1.16 Q96KN2 Beta-Ala-His dipeptidase   1E+08   1E+08 1.00 P43320 Beta-crystallin B2 93275660 75149812 1.24 P43251 Biotinidase 1.46E+08 1.37E+08 1.07 P04003 C4b-binding protein alpha chain 7.05E+08 7.41E+08 0.95 P00915 Carbonic anhydrase 1 55227827 65693374 0.84 Q96IY4 Carboxypeptidase B2 1.37E+08  1.2E+08 1.14 P15169 Carboxypeptidase N catalytic chain 1.43E+08 1.41E+08 1.02 P22792 Carboxypeptidase N subunit 2 2.47E+08 2.47E+08 1.00 P49747 Cartilage oligomeric matrix protein 20260311 19701309 1.03 O43866 CD5 antigen-like 39540069 42070317 0.94 P00450 Ceruloplasmin 5.54E+09 4.79E+09 1.16 P06276 Cholinesterase 1.33E+08 1.17E+08 1.13 P10909 Clusterin  1.5E+09 1.22E+09 1.22 P00740 Coagulation factor IX 1.48E+08 1.28E+08 1.16 P12259 Coagulation factor V 3.65E+08 3.55E+08 1.03 P00742 Coagulation factor X 1.65E+08 1.44E+08 1.15 P03951 Coagulation factor XI 36617287 32359907 1.13 P00748 Coagulation factor XII 5.97E+08 5.25E+08 1.14 P00488 Coagulation factor XIII A chain 1.86E+08 2.08E+08 0.89 P05160 Coagulation factor XIII B chain 1.02E+08 87470935 1.16 P23528 Cofilin-1 33766862 46939930 0.72 P02746 Complement C1q subcomponent 6.61E+08 6.82E+08 0.97 subunit B P02747 Complement C1q subcomponent 1.73E+08 1.37E+08 1.26 subunit C P00736 Complement C1r subcomponent 8.71E+08 7.53E+08 1.16 Q9NZP8 Complement C1r subcomponent-like 71274570 66217400 1.08 protein P09871 Complement C1s subcomponent 7.53E+08 7.07E+08 1.07 P06681 Complement C2  2.8E+08 2.87E+08 0.98 P01024 Complement C3 4.07E+08 4.76E+08 0.85 P0C0L4 Complement C4-A 1.71E+10  1.6E+10 1.07 P0C0L5 Complement C4-B 3.54E+08 3.09E+08 1.14 P01031 Complement C5 1.88E+09 1.74E+09 1.08 P13671 Complement component C6 8.14E+08 7.85E+08 1.04 P10643 Complement component C7 5.17E+08 4.59E+08 1.12 P07357 Complement component C8 alpha chain 2.12E+08 1.77E+08 1.20 P07358 Complement component C8 beta chain  3.8E+08 3.84E+08 0.99 P07360 Complement component C8 gamma 3.21E+08 3.22E+08 1.00 chain P02748 Complement component C9 1.29E+09 1.26E+09 1.02 P00751 Complement factor B 4.24E+09 3.93E+09 1.08 P08603 Complement factor H 4.18E+09 3.61E+09 1.16 Q03591 Complement factor H-related protein 1 80560638 67834507 1.19 P36980 Complement factor H-related protein 2 80956562 69473412 1.17 Q9BXR6 Complement factor H-related protein 5 26041456 20054874 1.30 P05156 Complement factor I 1.33E+09 1.11E+09 1.20 P08185 Corticosteroid-binding globulin 2.33E+08 2.27E+08 1.02 P09172 Dopamine beta-hydroxylase 29293024 25464467 1.15 P11021 Endoplasmic reticulum chaperone BiP 10076120 11107986 0.91 Q16610 Extracellular matrix protein 1 2.67E+08 2.48E+08 1.08 Q86UX7 Fermitin family homolog 3 75769339 1.13E+08 0.67 Q9UGM5 Fetuin-B 1.96E+08 1.35E+08 1.45 P02671 Fibrinogen alpha chain 4.95E+08 5.38E+08 0.92 P02675 Fibrinogen beta chain 5.56E+08 5.27E+08 1.05 P02679 Fibrinogen gamma chain  4.2E+08 4.52E+08 0.93 P02751 Fibronectin 6.03E+09 5.25E+09 1.15 P23142 Fibulin-1 70803799 59337625 1.19 O75636 Ficolin-3 2.01E+08 1.83E+08 1.10 P21333 Filamin-A 37834622 61900410 0.61 P04075 Fructose-bisphosphate aldolase A 35345594 52888573 0.67 Q08380 Galectin-3-binding protein 12756921 10979333 1.16 P06396 Gelsolin 1.83E+09 1.65E+09 1.11 P22352 Glutathione peroxidase 3 1.11E+08 1.24E+08 0.89 P00738 Haptoglobin   5E+08 5.68E+08 0.88 P68871 Hemoglobin subunit beta 1.89E+08 1.45E+08 1.30 P02790 Hemopexin 1.02E+10 9.65E+09 1.06 P05546 Heparin cofactor 2 2.37E+09 2.23E+09 1.06 Q04756 Hepatocyte growth factor activator 95828468 93649078 1.02 P26927 Hepatocyte growth factor-like protein  1.3E+08 1.15E+08 1.13 P04196 Histidine-rich glycoprotein 1.57E+09  1.3E+09 1.21 Q14520 Hyaluronan-binding protein 2 1.24E+08 1.06E+08 1.17 Q9Y6R7 IgGFc-binding protein 44637880 40960162 1.09 P01876 Immunoglobulin heavy constant alpha 1 1.69E+08 1.89E+08 0.89 P01857 Immunoglobulin heavy constant gamma 1.47E+08   2E+08 0.74 1 P01859 Immunoglobulin heavy constant gamma 62720442 80586495 0.78 2 P01860 Immunoglobulin heavy constant gamma 5.72E+08  8.1E+08 0.71 3 P01871 Immunoglobulin heavy constant mu 1.64E+08 2.88E+08 0.57 P01834 Immunoglobulin kappa constant 1.15E+08 1.54E+08 0.75 P01623 Immunoglobulin kappa variable 3-20 23132218 31380021 0.74 P0CG06 Immunoglobulin lambda constant 2  1.1E+08 1.89E+08 0.58 P17936 Insulin-like growth factor-binding 1.48E+08 1.26E+08 1.17 protein 3 P35858 Insulin-like growth factor-binding 7.02E+08 6.49E+08 1.08 protein complex acid labile subunit P19827 Inter-alpha-trypsin inhibitor heavy chain 2.87E+09 2.29E+09 1.25 H1 P19823 Inter-alpha-trypsin inhibitor heavy chain 3.27E+09 2.97E+09 1.10 H2 Q06033 Inter-alpha-trypsin inhibitor heavy chain  2.2E+08 2.45E+08 0.90 H3 Q14624 Inter-alpha-trypsin inhibitor heavy chain 4.48E+09 4.33E+09 1.03 H4 Q9NPH3 Interleukin-1 receptor accessory protein 20168003 23175142 0.87 P29622 Kallistatin 7.05E+08 6.29E+08 1.12 P01042 Kininogen-1  4.4E+09 3.63E+09 1.21 P02750 Leucine-rich alpha-2-glycoprotein 7.31E+08 7.76E+08 0.94 P18428 Lipopolysaccharide-binding protein 1.38E+08 1.35E+08 1.02 P00338 L-lactate dehydrogenase A chain 8952806 11233067 0.80 P07195 L-lactate dehydrogenase B chain 26817779 29679402 0.90 P51884 Lumican 5.61E+08 4.64E+08 1.21 P48740 Mannan-binding lectin serine protease 1 32715093 31407783 1.04 P11226 Mannose-binding protein C 53244859 42113257 1.26 P08571 Monocyte differentiation antigen CD14 1.28E+08 1.29E+08 1.00 Q96PD5 N-acetylmuramoyl-L-alanine amidase 5.66E+08 5.24E+08 1.08 O00533 Neural cell adhesion molecule L1-like 11603131 10077883 1.15 protein P30041 Peroxiredoxin-6 16785662 21645949 0.78 P80108 Phosphatidy linositol-glycan-specific 2.45E+08 2.07E+08 1.18 phospholipase D P55058 Phospholipid transfer protein 8536523 12589594 0.68 P36955 Pigment epithelium-derived factor 9.44E+08  8.2E+08 1.15 P03952 Plasma kallikrein 5.52E+08 4.87E+08 1.13 P05155 Plasma protease C1 inhibitor 1.77E+09 1.82E+09 0.97 P05154 Plasma serine protease inhibitor 1.68E+08 1.55E+08 1.09 P00747 Plasminogen 3.82E+09 3.11E+09 1.23 P13796 Plastin-2 35697546 34051982 1.05 P02775 Platelet basic protein 1.92E+08 3.07E+08 0.63 P07359 Platelet glycoprotein Ib alpha chain 58568238 61100167 0.96 P20742 Pregnancy zone protein 1.62E+08 47918310 3.39 P07737 Profilin-1 44916658 69435692 0.65 P02760 Protein AMBP 1.23E+09  1.2E+09 1.02 Q9UK55 Protein Z-dependent protease inhibitor 71636100 63032835 1.14 Q92954 Proteoglycan 4 46278448 45413513 1.02 P00734 Prothrombin  4.6E+09 4.17E+09 1.10 P14618 Pyruvate kinase PKM 19533899 33469384 0.58 Q12913 Receptor-type tyrosine-protein 8648297 8875812 0.97 phosphatase eta P02753 Retinol-binding protein 4 6.12E+08 5.07E+08 1.21 P49908 Selenoprotein P 89654111 94473461 0.95 P02787 Serotransferrin 7.14E+08 7.72E+08 0.92 P02743 Serum amyloid P-component 7.99E+08 7.47E+08 1.07 P27169 Serum paraoxonase/arylesterase 1 72238374 66744222 1.08 P04278 Sex hormone-binding globulin 1.26E+08 94259609 1.33 O00391 Sulfhydryl oxidase 1 59043964 58513437 1.01 Q9Y490 Talin-1 1.33E+08 2.24E+08 0.60 P22105 Tenascin-X 30161188 28254344 1.07 P05452 Tetranectin 4.82E+08 4.58E+08 1.05 P07996 Thrombospondin-1 92902684 1.24E+08 0.75 P05543 Thyroxine-binding globulin  1.9E+08 1.59E+08 1.19 Q15582 Transforming growth factor-beta- 25159916 22518702 1.12 induced protein ig-h3 P02766 Transthyretin 26501841 42941787 0.62 P60174 Triosephosphate isomerase 8978713 12621377 0.71 P19320 Vascular cell adhesion protein 1 19702232 19258760 1.20 Q6EMK4 Vasorin 23990562 25021406 0.96 P18206 Vinculin 66442583  1.1E+08 0.60 P02774 Vitamin D-binding protein 6.68E+09 5.45E+09 1.23 P04070 Vitamin K-dependent protein C 48342064 42493971 1.14 P07225 Vitamin K-dependent protein S 2.41E+08 2.43E+08 0.99 P22891 Vitamin K-dependent protein Z 32900507 29571834 1.11 P04004 Vitronectin 3.44E+09 3.18E+09 1.08 P04275 von Willebrand factor 91148073 85190545 1.07 P25311 Zinc-alpha-2-glycoprotein 1.71E+09 1.74E+09 0.98

Example 2

[0388] Another set of models were developed without incorporating correction for covariates. Here, instead, the previous covariates (age, sex, BMI and years in education) were included as clinical features in and of themselves in the models. Methods and results are explained in detail below.

Methods

[0389] As above, we used Neurominer v.1.0 (https://www.pronia.eu/neurominer/) for MatLab 2018a (MathWorks Inc.) to develop support vector machine (SVM) models. For all models, hyper-parameters were optimised in repeated nested cross-validation (see eMethods) and area under the receiver-operating characteristic curve (AUC) was the performance evaluation criterion. Random-label permutation analysis with 1000 permutations was used to verify against a null distribution and derive p-values for statistical significance. Missing clinical data were replaced using the mean (for continuous) or modal value (for categorical variables). Continuous clinical variables were converted to z-scores and winsorised within ±3z.

Models 1a-c: Predicting Transition Using Clinical and Proteomic Data

[0390] First, we developed a model predicting transition outcome based on the clinical and proteomic data in combination (Model 1a). The included clinical features are listed in Table 16 below. We incorporated geographical generalisability by using a leave-site-out cross-validation approach (see eMethods) as recommended for data from multisite consortia. We used the LIBLINEAR program with L2 regularisation to attenuate over-fitting whereby weightings of non-predictive features are minimised, but not reduced to zero (thus more closely modelling the biological effects of functionally inter-related proteins). The hyperplane was weighted (increasing the misclassification penalty in the minority class) which reduces the risk of bias in unbalanced group sizes.

[0391] Next, to assess the relative contribution of clinical and proteomic data, we developed models using the same cross-validation and training framework but based on clinical features (Model 1b) and proteomic features (Model 1c) separately.

Models 2a-b: Parsimonious Model

[0392] We sought to generate a parsimonious model based on the 10 highest-weighted proteomic predictors from a London test site, and the model was validated.

[0393] To derive the 10 highest-weighted proteins, we generated an L2-regularised SVM model (Model 2a) using proteomic data from all sites except London (CHR-T n=30, CHR-NT n=50). A reduced model was then developed based solely on data for these 10 proteins in the non-London dataset (Model 2b), before being tested in the London data (CHR-T n=19, CHR-NT n=34). Both models used leave-site-out cross-validation.

Model 3: Replication

[0394] Due to differences in protein identifications, it was not possible to apply the above models in the replication dataset. We instead sought to replicate our initial findings by conducting a second discovery analysis, generating a new L2-regularised SVM model (with leave-site-out cross-validation) for prediction of transition status based on the clinical and proteomic data in the replication dataset.

Results

Model 1a: Predicting Transition Using Clinical and Proteomic Data

[0395] An SVM model predicted transition status based on 65 clinical and 166 proteomic features (Model 1a) with excellent performance (AUC 0.95, p<0.001). Performance metrics are presented in Table 17. FIG. 11a shows mean algorithm scores and predicted outcomes stratified by site. The receiver-operating characteristic curve is shown in FIG. 11b. Table 18 lists the 10% highest-weighted features based on mean feature weighting across all models selected in the inner loop.

Model 1b: Clinical Data

[0396] The clinical model (Model 11b) demonstrated poor predictive performance (AUC 0.48, p=0.628; Table 17, FIG. 12).

Model 1c: Proteomic Data

[0397] The proteomic model (Model 1c) demonstrated excellent predictive performance (AUC 0.96, p<0.001; Table 17, FIG. 13).

Models 2a-b: Parsimonious Model

[0398] The AUC for the model based on proteomic data from all sites except London (Model 2a) was 0.94, p<0.001 (Table 17, FIG. 14). The 10 highest-weighted features were: alpha-2-macroglobulin (A2M), immunoglobulin heavy constant mu (IGHM), C4b-binding protein alpha chain (C4BPA), vitamin K-dependent protein S, fibulin-1, transthyretin, N-acetylmuramoyl-L-alanine amidase, vitamin D-binding protein, clusterin and complement component 6 (C6).

[0399] A reduced model based solely on these 10 features was then trained using data from all sites except London (Model 2b) with AUC 0.99, p<0.001 (Table 17, FIG. 15). This model predicted transition status in the London data (CHR-T n=19, CHR-NT n=34) with AUC 0.92 (Table 17).

Model 3: Replication

[0400] This model demonstrated excellent performance for prediction of transition outcome in the replication dataset (AUC 0.98, p<0.001; Table 17, FIG. 16). The highest-weighted 10% of features are shown in Table 18 Proteins among the highest-weighted 10% of features in both Model 1a and Model 3 (and weighted in similar directions) included: A2M, IGHM, C4BPA, plasminogen and complement component 6 (C6).

TABLE-US-00016 TABLE 16 List of 69 baseline EU-GEI clinical features included in Model 1a, Model 1b and Model 3 Age Sex Body mass index Years in education GAF symptoms GAF disability SANS: unchanging facial expression SANS: decreased spontaneous movements SANS: paucity of expressive gestures SANS: poor eye contact SANS: affective nonresponsivity SANS: inappropriate affect SANS: lack of vocal inflections SANS: global rating of affective flattening SANS: poverty of speech SANS: poverty of speech content SANS: blocking SANS: increased latency of response SANS: global rating of alogia SANS: grooming and hygiene SANS: impersistence at work or school SANS: physical anergia SANS: global rating for avolition-apathy SANS: recreational interests and activities SANS: sexual activity SANS: ability to feel intimacy and closeness SANS: relationship with friends and peers SANS: global rating of anhedonia-asociality SANS: social inattentiveness SANS: inattentiveness during mental status testing SANS: global rating of attention Total SANS composite score Total SANS global score BPRS: somatic concern BPRS: anxiety BPRS: depression BPRS: suicidality BPRS: guilt BPRS: hostility BPRS: elevated mood BPRS: grandiosity BORS: suspiciousness BPRS: hallucinations BPRS: unusual thought content BPRS: bizarre behaviour BPRS: self-neglect BPRS: disorientation BPRS: conceptual disorganisation BPRS: blunted affect BPRS: emotional withdrawal BPRS: motor retardation BPRS: tension BPRS: uncooperativeness BPRS: excitement BPRS: distractibility BPRS: motor hyperactivity BPRS: mannerisms and posturing Total BPRS score MADRS: apparent sadness MADRS: reported sadness MADRS: inner tension MADRS: reduced sleep MADRS: reduced appetite MADRS: concentration difficulties MADRS: lassitude MADRS: inability to feel MADRS: pessimistic thoughts MADRS: suicidal thoughts Total MADRS score GAF: General Assessment of Functioning; SANS: Scale for the Assessment of Negative Symptoms; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery Asberg Depression Rating Scale; EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions.

TABLE-US-00017 TABLE 17 Performance metrics for unadjusted support vector machine models Bal- True False True False anced AUC Positive Negative Positive Negative posi- nega- nega- posi- Sensi- Speci- accu- (95% predic- predic- likeli- likeli- tives, tives, tives, tives, tivity, ficity, racy, confidence tive tive hood hood Model description n (%) n (%) n (%) n (%) % % % interval) value, % value, % ratio ratio Model 1a: clinical 48 1 68 16 98.0 81.0 89.5 0.95 75.0 98.6 5.1 <0.1 and proteomic (98%) (2%) (81%) (19%) (0.91-0.99) Dataset: EU-GEI initial experiment, all sites Features: 69 clinical and 166 proteomic Target: transition status N: 49 transition, 84 non-transition Model 1b: clinical 23 26  45 39 46.9 53.6 50.3 0.48 37.1 63.4 1.0 1.0 Dataset: EU-GEI (47%) (53%)  (54%) (46%) (0.38-0.58) initial experiment, all sites Features: 69 clinical Target: transition status N: 49 transition, 84 non-transition Model 1c: proteomic 49 0 71 13 100.0 84.5 92.3 0.96 79.0 100.0 6.5 <0.1 Dataset: EU-GEI (100%)  (0%) (85%) (15%) (0.92-1.00) initial experiment, all sites Features: 166 proteomic Target: transition status N: 49 transition, 84 non-transition Model 2a: proteomic 28 2 40 10 93.3 80.0 86.7 0.94 73.7 95.2 4.7 0.1 (non-London) (93%) (7%) (80%) (20%) (0.88-1.00) Dataset: EU-GEI initial experiment, all sites except London Features: 166 proteomic Target: transition status N: 30 transition, 50 non-transition Model 2b: top 10, 30 0 41  9 100.0 82.0 91.0 0.99 76.9 100.0 5.6 <0.1 training data (100%)  (0%) (82%) (18%) (0.96-1.00) Dataset: EU-GEI initial experiment, all sites except London Features: 10 proteomic Target: transition status N: 30 transition, 50 non-transition Model 2b: top 10, test 18 1 30  4 94.7 88.2 91.5 0.92 81.8 96.8 8.1 0.1 data (95%) (5%) (88%) (12%) (0.83-1.00) Dataset: EU-GEI initial experiment, London site Features: 10 proteomic Target: transition status N: 19 transition, 34 non-transition Model 3: replication 48 1 77  9 98.0 89.5 93.7 0.98 84.2 98.7 9.4 <0.1 Dataset: EU-GEI (98%) (2%) (90%) (10%) (0.95-1.00) replication experiment, all sites Features: 69 clinical and 119 proteomic Target: transition status N: 49 transition, 86 non-transition Model 4: ALSPAC 40 15  47 19 72.7 71.2 72.0 0.74 67.8 75.8 2.5 0.4 PEs (73%) (27%)  (71%) (29%) (0.65-0.83) Dataset: ALSPAC Features: 265 proteomic Target: PEs age 18 N: 55 PEs, 66 no PE Model S1: ELISA 33 11  51 31 75.0 62.2 68.6 0.76 51.6 82.3 2.0 0.4 Dataset: EU-GEI (75%) (25%)  (62%) (38%) (0.67-0.85) initial experiment, all sites Features: 9 ELISA Target: transition status N: 44 transition, 82 non-transition Model S2: functional 27 20  22 10 57.4 68.8 63.1 0.74 73.0 52.4 1.8 0.6 outcome (57%) (43%)  (69%) (31%) (0.63-0.85) Dataset: EU-GEI initial experiment, all sites Features: 69 clinical and 166 proteomic Target: functional outcome N: 47 poor functioning (GAF ≤ 60); 32 good functioning (GAD > 60) AUC: area under the receiver-operating characteristic curve; EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions; ALSPAC: Avon Longitudinal Study of Parents and Children; PEs: psychotic experiences. Models 1a-d, 2 and 3 are adjusted for age, sex, body mass index and years in education and Model 4 is additionally adjusted for ethnicity and tobacco use. Model 5 is adjusted for sex, maternal social class at birth and body mass index at age 12.

TABLE-US-00018 TABLE 18 Ten percent highest-weighted features for Model 1a, Model 3 and Model 4 (ranked according to mean feature weight for models selected in cross-validation inner loop) Model 1a: EU-GEI clinical and proteomic data, initial experiment, Model 3: EU-GEI clinical and proteomic all sites data, replication experiment, all sites Model 4: ALSPAC proteomic data Mean Mean Mean Feature weight Feature weight Feature weight P01023 Alpha-2- −0.330 P01023 Alpha-2-macroglobulin −0.286 P04003 C4b-binding protein −0.227 macroglobulin alpha chain P01871 Immunoglobulin −0.256 P22792 Carboxypeptidase N 0.210 P27169 Serum −0.180 heavy constant mu subunit 2 paraoxonase/arylesterase 1 P04003 C4b-binding −0.161 P01871 Immunoglobulin heavy −0.193 Q03591 Complement factor H- −0.152 protein alpha chain constant mu related protein 1 P07357 Complement 0.158 P09871 Complement C1s −0.181 P07225 Vitamin K-dependent −0.145 component 8 alpha chain subcomponent protein S P55058 Phospholipid −0.146 P01011 Alpha-1-antichymotrypsin 0.168 P61626 Lysozyme C −0.142 transfer protein O75636 Ficolin-3 −0.145 P00747 Plasminogen 0.163 P55103 Inhibin beta C chain 0.139 P02774 Vitamin D 0.135 P08571 Monocyte differentiation 0.161 Q08380 Galectin-3-binding 0.132 binding protein antigen CD14 protein P07225 Vitamin K- −0.132 P10909 Clusterin 0.158 P24593 Insulin-like growth 0.122 dependent protein S factor-binding protein 5 P43320 Beta-crystallin 0.132 Q16610 Extracellular matrix 0.157 P00746 Complement factor D 0.120 B2 protein 1 P02766 Transthyretin −0.130 G3XAM2 Complement factor I 0.140 P01019 Angiotensinogen −0.118 P23142 Fibuln-1 0.125 P04003 C4b binding protein alpha −0.140 P01871 Immunoglobulin heavy −0.116 chain constant mu P10909 Clusterin 0.121 P13671 Complement component 6 0.132 O75636 Ficolin-3 0.115 P05155 Plasma protease −0.114 P25311 Zinc alpha-2-glycoprotein −0.131 Q9H4A9 Dipeptidase 2 −0.115 C1 inhibitor Sex −0.111 P07359 Platelet glycoprotein Ib 0.126 P01023 Alpha-2-macroglobulin −0.113 alpha chain P00747 Plasminogen 0.111 P01031 Complement C5 0.125 P04275 von Willebrand factor −0.111 P13671 Complement 0.111 O75882 Attractin 0.123 Q9NQ79 Cartilage acidic protein 0.107 component 6 1 P02747 Complement C1q 0.109 P0DOY3 Immunoglobulin lambda −0.120 P24592 Insulin-like growth 0.106 subcomponent subunit C constant 3 factor-binding protein 6 P02753 Retinol-binding 0.109 P15169 Carboxypeptidase N 0.115 P09871 Complement C1s −0.105 protein 4 catalytic chain (CPN) subcomponent Q76LX8 A disintegrin −0.108 P10909 Clusterin −0.105 and metalloproteinase with thrombospondin motifs 13 P08697 Alpha-2- −0.106 O95497 Pantetheinase 0.105 antiplasmin P19827 Inter-alpha- 0.105 P02654 Apolipoprotein C-I −0.099 trypsin inhibitor heavy chain H1 MADRS: concentration −0.104 P02679 Fibrinogen gamma chain −0.099 difficulties P02489 Alpha-crystallin 0.101 P07358 Complement component 0.097 A chain C8 beta chain Q5T7F0 Neuropilin −0.097 P04040 Catalase 0.094 P43251 Biotinidase 0.094 Proteins are presented with their Uniprot accession number and corresponding protein name. EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions; ALSPAC: Avon Longitudinal Study of Parents and Children; BPRS: Brief Psychiatric Rating Scale; MADRS: Montgomery-Asberg Depression Rating Scale; SANS: Scale for the Assessment of Negative Symptoms

TABLE-US-00019 TABLE 19 Ten percent highest-weighted features for Model S2 (support vector machine model predicting functional outcome at 24 months in EU-GEI) Mean Feature weight BPRS: suspiciousness 0.197 P01023 Alpha-2-macroglobulin −0.191 P55058 Phospholipid transfer protein −0.186 P01871 Immunoglobulin heavy constant −0.182 mu Q9UGM5 Fetuin-B 0.148 O43866 CD5 antigen-like −0.145 P14618 Pyruvate kinase −0.133 P19827 Inter-alpha-trypsin inhibitor heavy 0.129 chain H1 SANS: blocking 0.118 SANS: increased latency of response 0.111 P10909 Clusterin 0.107 P08603 Complement factor H 0.104 P36955 Pigment epithelium-derived factor 0.103 MADRS: suicidal thoughts −0.103 Impersistence at work or school 0.102 P17936 Insulin-like growth factor-binding 0.102 protein 3 Age −0.101 P04196 Histidine-rich glycoprotein 0.099 Q08380 Galectin-3-binding protein 0.097 SANS: grooming and hygiene 0.097 SANS: ability to feel intimacy and 0.095 closeness SANS: sexual activity −0.093 P11226 Mannose-binding protein C 0.090 Features are ranked according to mean feature weight for models selected in the cross-validation inner loop. EU-GEI: European Network of National Schizophrenia Networks Studying Gene-Environment Interactions; BPRS: Brief Psychiatric Rating Scale; SANS: Scale for Assessment of Negative Symptoms; MADRS: Montgomery-Asberg Depression Rating Scale.