Method For Protein Kinase Activity Ranking

Abstract

The present invention provides a method of quantifying the activity of a protein modifying enzyme in a sample, comprising calculating the value K for said protein-modifying enzyme on the basis of the number of modified peptides in a sample that are substrates of said protein modifying enzyme, the intensity of the modified peptides, each modified peptide in the sample that is a substrate of said protein modifying enzyme, the total number of modified peptides in the sample, the intensity of the modified peptides and all of the modified peptides in the sample. A method of quantifying the activity of a protein modifying enzyme in a sample, comprising calculating the value SC for said protein-modifying enzyme on the basis of a reduction in proliferation using an inhibitor at an inhibitor concentration at which proliferation is measured and the “in vitro” IC50i of the inhibitor against a primary target is also provided. The invention further provides methods of identifying inhibitors with which to treat a patient, methods of treatment, a computer readable medium, a computer program product and devices for carrying out the methods.

Claims

1-39. (canceled)

40. A method of treating a patient in need thereof with an inhibitor of a protein modifying enzyme, comprising: (i) calculating the value K for each protein modifying enzyme in a sample taken from said patient according to $K=\frac{\underset{i=1}{\overset{m}{.Math.}}.Math.{\alpha }_i}{\underset{j=1}{\overset{l}{.Math.}}.Math.{\beta }_j},$ wherein: m=the number of modified peptides in the sample that are substrates of the protein-modifying enzyme; α=the intensity of the modified peptides i; i=each modified peptide in the sample that is a substrate of the protein-modifying enzyme; l=the total number of modified peptides in the sample; β=the intensity of the modified peptides j; and j=all of the modified peptides in the sample; (ii) identifying the protein modifying enzyme with the highest value K; (iii) selecting an inhibitor that targets the protein modifying enzyme with the highest value K; and (iv) administering said inhibitor to said patient.

41-42. (canceled)

43. A non-transitory computer readable medium comprising computer readable code operable, in use, to instruct a computer to perform a method of calculating the value K for a protein-modifying enzyme in a sample taken from a patient according to $K=\frac{\underset{i=1}{\overset{m}{.Math.}}.Math.{\alpha }_i}{\underset{j=1}{\overset{l}{.Math.}}.Math.{\beta }_j},$ wherein: m=the number of modified peptides in the sample that are substrates of the protein-modifying enzyme; α=the intensity of the modified peptides i; i=each modified peptide in the sample that is a substrate of the protein-modifying enzyme; l=the total number of modified peptides in the sample; β=the intensity of the modified peptides j; and j=all of the modified peptides in the sample.

44. (canceled)

45. A device comprising: a memory having computer executable code stored thereon arranged to carry out a method of calculating the value K for a protein-modifying enzyme in a sample taken from a patient according to $K=\frac{\underset{i=1}{\overset{m}{.Math.}}.Math.{\alpha }_i}{\underset{j=1}{\overset{l}{.Math.}}.Math.{\beta }_j},$ wherein: m=the number of modified peptides in the sample that are substrates of the protein-modifying enzyme; α=the intensity of the modified peptides i; i=each modified peptide in the sample that is a substrate of the protein-modifying enzyme; l=the total number of modified peptides in the sample; β=the intensity of the modified peptides j; and j=all of the modified peptides in the sample, and a processor arranged to execute the code stored in the memory.

46. The method according to claim 40, wherein the value K is calculated for said protein-modifying enzyme as follows: $K=\frac{\underset{i=1}{\overset{m}{.Math.}}.Math.{\alpha }_i}{\underset{j=1}{\overset{l}{.Math.}}.Math.{\beta }_j}.Math.{\left(\frac{m}{t}\right)}^{1/2},$ wherein t=the total number of known target modified peptides for said protein modifying enzyme.

47. The method according to claim 40, wherein the protein modifying enzyme is a protein kinase.

48. The method according to claim 40, further comprising identifying and/or quantifying modified peptides in a first sample and a second sample from the patient using mass spectrometry (MS).

49. The method according to claim 48, wherein identifying and/or quantifying modified peptides in a first sample and a second sample is carried out using a method comprising the following steps: a) obtaining peptides from a sample; b) adding reference modified peptides to the peptides obtained in step a) to produce a mixture of peptides and reference modified peptides; c) carrying out mass spectrometry (MS) on said mixture of peptides and reference modified peptides to obtain data relating to the peptides in the sample; and d) comparing the data relating to the peptides in the sample with data in a database of modified peptides using a computer programme; wherein the database of modified peptides is compiled by a method comprising: i) obtaining peptides from a sample; ii) enriching modified peptides from the peptides obtained in step i); iii) carrying out liquid chromatography-tandem mass spectrometry (LC-MS/MS) on the enriched modified peptides obtained in step ii); iv) comparing the modified peptides detected in step iii) to a known reference database in order to identify the modified peptides; and v) compiling data relating to the modified peptides identified in step iv into a database.

50. The method according to claim 49, wherein step b) further comprises enriching modified peptides from said mixture of peptides and reference modified peptides to produce a mixture of enriched modified peptides, and step c) comprises carrying out mass spectrometry (MS) on said mixture of enriched modified peptides to obtain data relating to the modified peptides in the sample.

51. The method according to claim 50, wherein the step of enriching modified peptides is carried out using chromatography.

52. The method according to claim 49, wherein the chromatography is selected from the group consisting of immobilized metal ion affinity chromatography (IMAC), titanium dioxide (TiO.sub.2) chromatography, and zirconium dioxide (ZrO.sub.2) chromatography.

53. The method according to claim 50, wherein the step of enriching modified peptides is carried out using antibody-based methods.

54. The method according to claim 49, wherein the data relating to the peptides in the sample comprises the mass to charge (m/z) ratio, charge (z), and relative retention time of the peptides.

55. The method according to claim 49, wherein said mass spectrometry (MS) in step c) is liquid chromatography-mass spectrometry (LC-MS).

56. The method according to claim 49, wherein step ii) is carried out using multidimensional chromatography.

57. The method according to claim 56, wherein the multidimensional chromatography is carried out using strong cation exchange high performance liquid chromatography (SCX-HPLC), immobilized metal ion affinity chromatography (IMAC), and titanium dioxide (TiO.sub.2) chromatography.

58. The method according to claim 56, wherein the multidimensional chromatography is carried out using anion exchange high performance liquid chromatography (SAX-HPLC), immobilized metal ion affinity chromatography (IMAC), and titanium dioxide (TiO.sub.2) chromatography.

59. The method according to claim 49, wherein step ii) is carried out using antibody-based methods.

60. The method according to claim 49, wherein step iv) is carried out using the MASCOT search engine.

61. The method according to claim 49, wherein the data relating to the modified peptides identified in step iv) is selected from the group consisting of identity of the modified peptide, mass to charge (m/z) ratio, charge (z), and relative retention time of the modified peptide.

62. The method according to claim 48, wherein the MS technique uses isotope labels for quantification.

Description

[0187] The present invention will now be further described by way of reference to the following Examples which are present for the purposes of illustration only. In the Examples, reference is made to a number of Figures in which:

[0188] FIG. 1. Reproducibility, linearity and accuracy of K-score determination._(a) The principle of Kinase Activity Ranking (KAR, which produces K-scores). The intensities of phosphorylated peptides containing sites commonly phosphorylated by given kinases are summed and normalised to the sum of all phosphopeptide intensities. The resulting values are further normalized to the number of substrates known and identified per kinase. (b) Experimental design to assess quantitative nature of the algorithm. (c, d), Reproducibility, and linearity of K-scores for three representative kinases. (e) Ranking of kinases based on K-scores as a function of pV treated cells. Tyrosine kinases (shown as triangle data points) clearly increase their K-score-based ranks as a function of pV treated extracts present in samples.

[0189] FIG. 2. K-scores as a function of EGF or IGF treatment. The phosphoproteomics data in Wilkes et al PNAS 112, 7719-7724 (2015) was processed for KAR analysis. The results illustrate that the outputs of KAR (K-scores) are a measure of kinase activity as these values show the expected kinetics of growth factor stimulation on kinase activity.

[0190] FIG. 3. The K-score ranks kinase activities based on their contribution to cellular phosphorylation relative to each other. (a) K-scores of over 60 kinases across eight haematological cell-lines. Phosphoproteomics experiments were performed on the named cell-lines in four independent occasions as previously described and each was analysed twice by LC-MS/MS. Mean K-scores of the eight analytical replicates are shown. (b) Ranking of kinases based on their contribution to total cellular phosphorylation as determined by KAR analysis.

[0191] FIG. 4. Dose response curves of hematological cell lines treated with a panel of kinase inhibitors. The named cell lines were treated with the inhibitors shown and viability measured using a Guava assay after 72 h of treatment. Data points are mean±SD (n=4).

[0192] FIG. 5. Kinase activity Ranking models the contribution of kinase activities to cell viability in haematological cell-lines. (a) SC (sensitivity coefficient) of the named compounds across eight haematological cell-lines. (b) Frequency of ranks of kinase inhibitors based on their SC across cell-lines. (c) Association between kinase activity (as measured by KAR, FIG. 1) and SC within individual cell-lines. K-score values are shown as mean±SEM of four independent experiments, each measured in duplicate. r, pearson correlation coefficient; p values were derived from r values.

[0193] FIG. 6. Relationship between K-scores and sensitivity to kinase inhibitors across eight hematological cell lines. Linear regression analysis measured across 8 cell lines shows that Sensitivity Coefficient for 9 different drugs were significantly associated with the K-scores for their main target kinases.

[0194] FIG. 7. Modelling the contribuon of kinase activity to acute myeloid leukemia (AML) cell viability. (a) Distribution of K-scores in 45 primary AML biopsies. (b) Frequency of ranks of the named kinases as determined by K-score analysis. (c) Distribution of IC.sub.50s and SCs in 36 AML biopsies treated with the named kinase inhibitors; P38i, TAK775; CK2i, CX494; MEKi tramentinib (inhibits MAPK signalling); PAKi, PF-03758309. (d) Association between SC and K-scores in two representative AML patient samples and in 36 cases of primary AML (e). (f,g) Correlation between K-scores for MEK1 and JAK2 and cell viability as a function of MEKi and PKC412 treatment, respectively.

[0195] FIG. 8. Dose response curves of primary AML cells treated with a panel of kinase inhibitors. AML biopsies were obtained from the Barts Cancer Institute biobank with ethical consent, treated with the named compounds for 72 h and viability measures using Guava ViaCount assay. Data points are mean±SD (n=3).

[0196] FIG. 9. Complex drug response phenotypes are associated with differences in kinase activities. (a) Patterns of responses to five kinase inhibitors in primary AML cells. Sensitivity was measured at five inhibitor concentratons. The heatmap shows viability at 1 μM concentration (left) and a decrease in viability by >%50 is denoted by green boxes (right). A total of 13 groups based on responses were identified which could be further classified into five main response groups. (b-g) K-scores for the named kinases, or their ratios, across the response groups. Individual K-scores were normalised to the average across patients. Shown are the mean±SEM and individual values within response groups. P-values were calculated through an unpaired t-test against the responses groups 12 or 11 in (c), (d) and (f), and (g) respectively.

[0197] FIG. 10. Models of cell viability in a panel of primary AML biopsies. K-Scores and Sensitivity Confidents were obtained from phosphoproteomics and cell viability data, respectively, and compared in 36 different AML patient biopsies.

[0198] FIG. 11. Kinase activities associated with patterns of responses to kinase inhibitors. (a) K-scores were normalized to the average across the response groups shown in FIG. 4a of the main text (average of the row) and log transformed. (b) Comparison of K-scores between cells sensitive to at least 3 compounds (Groups 1-5) with those from cells resistant to all compounds (Group 13) showing kinase activities increased in sensitive cells (red fonts). (c) Comparison of K-scores between cells sensitive to MEK1i only (group 12) and those sensitive to CK2i only (Group 11). (d) Comparison of K-scores between cells sensitive to CK2i only (group 11) and those sensitive to PAKi only (Group 10). (e) Comparison of K-scores between cells sensitive to MEK1i only (group 12) and those sensitive to PAKi only (Group 10).

EXAMPLE

[0199] Materials and Methods

[0200] Cell Culture

[0201] B-cell lymphoma and leukemia cell-lines were routinely maintained in RPMI-1640 medium supplemented with 10% fetal-bovine serum (FBS) and 100 U.Math.mL.sup.−1 penicillin/streptomycin (P/S). Cells were maintained at a confluency of 0.5-2.0×10.sup.6 cells.Math.mL.sup.−1. Stromal cells were grown in IMDM medium (supplemented with 10% FBS and 100 U.Math.mL.sup.−1 P/S) and maintained at a confluency of 2.0-30.0×10.sup.6 cells in 175 cm.sup.2 flasks. MS-5 conditioned IMDM medium was generated by growing stromal cells in IMDM for 3 days. All cells were kept at 37° C. in a humidified atmosphere at 5% CO.sub.2.

[0202] Primary AML Cells

[0203] All patients gave informed consent for the storage of their blood cells for research purposes. Each procedure was conducted in accordance with the East London and City Research Ethics Committee, as previously described (Miraki-Moud, F. et al. Blood 125, 4060-4068 (2015)). All studies comply with the rules of the Review Board and by the revised Helsinki protocol. Peripheral blood was extracted from AML patients at St Bartholomew's Hospital and mononuclear cells, isolated using Ficoll gradient followed by red cell lysis, were stored in liquid N.sub.2. Primary AML blasts were thawed following standard procedures: briefly, vials were defrosted at 37 C and exposed to 500 μg of DNAse (Sigma) for 5 minutes. 10 mL of PBS, supplemented with 2% FBS, was added and cell suspension was centrifuged at 1500 rpm for 5 min at 5° C. Cells were resuspended in MS-5 conditioned IMDM medium and filtered using a 70 μm strainer (Fisherbrand). Cell number and viability were determined by trypan blue staining using a Vi-CELL XR cell viability analyser (Beckman Coulter).

[0204] Cell Lysis and Protein Digestion

[0205] For each cell-line, 4 independent biological replicates were performed: 10×10.sup.6 cells were seeded at 0.5×10.sup.6 cells-mL.sup.−1 and left overnight. For each AML primary sample, 10×10.sup.6 cells were seeded at 1×10.sup.6 cells.Math.mL.sup.−1 and left in the incubator for 2 h. Cells were subsequently harvested by centrifugation, washed twice with ice-cold phosphate-buffered saline—supplemented with 1 mM Na.sub.3VO.sub.4 and 1 mM NaF—and lysed in 0.2 mL of ice-cold urea lysis buffer (8M urea in 20 mM HEPES (pH 8.0), supplemented with 1 mM Na.sub.3VO.sub.4, 1 mM NaF, 1 mM Na.sub.2H.sub.2P.sub.2O.sub.7 and 1 mM β-glycerol phosphate). Lysates were further homogenized by sonication and any insoluble material removed by centrifugation. Protein concentration was estimated via the bicinchoninic acid (BCA) assay. After normalizing each condition to a common protein concentration (0.5 μg.Math.μL.sup.−1), each sample was reduced and alkylated by sequential incubation with 10 mM dithiothreitol and 16.6 mM iodoacetamide for 30 min at room temperature, in the dark. For protein digestion, the urea concentration was reduced to 2M through the addition of 20 mM HEPES (pH 8.0). Immobilized tosyl-lysine chloromethyl ketone (TLCK)-trypsin was then added, and samples incubated overnight at 37 C. Trypsin beads were removed by centrifugation and the resultant peptide solutions were desalted using OASIS HLB 1 cc solid phase extraction cartridges as described previously (Montoya, A. et al. Methods 54, 370-378 (2011)).

[0206] Phosphopeptide Enrichment

[0207] Phosphorylated peptides were enriched using TiO.sub.2(GL Sciences) as previously described (Wilkes, E. H. et al. Proceedings of the National Academy of Sciences of the United States of America 112, 7719-7724 (2015)). The resulting phosphopeptide solutions were snap-frozen, dried with a SpeedVac, and stored at −80° C. until further use.

[0208] LC-MS/MS Phosphoproteomics Analysis

[0209] For cell-line samples, each biological replicate was analyzed twice by LC-MS/MS as follows: phosphopeptide pellets were re-suspended in 14 μL of 0.1% TFA and 4.0 μL per technical replicate was injected into a Waters NanoACQUITY UPLC system (Waters, Manchester, UK) coupled online to an LTQ-Orbitrap-XL mass spectrometer (Thermo Fisher Scientific). The samples were separated on a 100 minute linear gradient between 5 and 35% ACN on an ACQUITY BEH130 C.sub.18 UPLC column (15 cm×75 μm, 1.7 μm, 130 Å) at a flow rate of 300 nL.Math.min.sup.−1. The top five most intense multiply charged ions in each MS.sup.1 scan were selected for collision-induced dissociation fragmentation (with multistage activation enabled). The resolution of the MS.sup.1 was set to 30,000 FWHM.

[0210] For the primary AML samples, each technical replicate (two per sample) consisted of a 3.0 μL injection into a Dionex Ultimate nRSLC system (Thermo Fisher Scientific) coupled online to a Q-Exactive Plus (QEP) mass spectrometer (Thermo Fisher Scientific). The samples were separated on a 120 min linear gradient between 3 and 30% ACN on an Acclaim PepMap C.sub.18 RSLC column (25 cm×75 μm, 2 μm, 100 Å) at a flow rate of 300 nL.Math.min.sup.−1. The top twenty most intense multiply charged ions present in each MS.sup.1 scan were selected for higher-energy collision-induced dissociation (HCD). The resolution of the MS.sup.1 scans was set to 70,000 FWHM.

[0211] Phosphopeptide Identification and Quantification

[0212] Peptide identification was performed by matching deisotoped, MS/MS data to the Uniprot Swissprot human protein databases (September 2014 release, containing 20,233 entries), utilizing the Mascot server version 2.4. Mascot Distiller was used to generate peak lists in the mascot generic format. Mass tolerances were set to 10 ppm and 600 mmu (XL)/25 mmu (QEP) for the precursor and fragment ions respectively. For the phosphoproteomics experiments, the allowed variable modifications were: phospho-Ser, phospho-Thr, phospho-Tyr, pyro-Glu (N-terminal), and oxidation-Met. The identified phosphopeptides from each of the samples were collated and curated using in-house scripts. Unique phosphopeptides ions with expectancy <0.05 were then included in the subsequent analyses. Mascot decoy database searches showed that with these settings produce a false discovery rate of ˜1%. Peptide quantification was performed as described before by our group (Montoya et al. (2011), supra; Casado, P. & Cutillas, P. R. Molecular & Cellular Proteomics: MCP 10, M110 003079 (2011); Cutillas, P. R. & Vanhaesebroeck, B. Molecular & Cellular Proteomics 6, 1560-1573 (2007)) and others (Tsou, C. C. et al. Molecular & Cellular Proteomics: MCP 9, 131-144 (2010); Mann, B. et al. Rapid Communications in Mass Spectrometry: RCM 22, 3823-3834 (2008)). Briefly, Pescal software (written in Python v2.7) was then used to obtain peak areas of extracted ion chromatograms of each of the phosphopeptide ions in the database, across all of the samples being compared. The retention times of each phosphopeptide ion, in each sample, were predicted through alignment of common phosphopeptides' retention times using an in-house linear modelling algorithm. Chromatographic peaks obtained from extracted ion chromatograms for each phosphopeptide in each sample were then integrated and the peak areas recorded. The mass-to-charge (m/z) and retention time (t.sub.R) tolerances were set to 7 ppm and 1.5 min, respectively.

[0213] Kinase-Substrate Enrichment Analysis and Kinase Activity Ranking Technical replicates were averaged and peak areas for each phosphopeptide ion were then normalized to the sum of peptide intensities for each sample. Kinase-substrate matching was performed on these data as reported before (Casado, P. et al. Science Signaling 6, rs6 (2013)) using a VBA script against the PhosphoSitePlus database (downloaded in July 2014). Kinase Activity Ranking (KAR) was calculated for kinase K using the equation below (where m=the number of phosphorylation sites in the dataset matched to kinase K; α=the normalized intensity of the phosphorylation site i; l=the total number of phosphorylation sites in the dataset regardless of any kinase-substrate association; β=the normalized intensity of the phosphorylation site j; t=the total number of known target phosphorylation sites in the PhosphoSitePlus database for kinase K). Data were visualized either using Microsoft Excel 2007/2010 or within the R statistical computing environment (v3.0.0) using a combination of the reshape2 and ggplot2 packages.

[00012]$K=\frac{\underset{i=1}{\overset{m}{.Math.}}.Math.{\alpha }_i}{\underset{j=1}{\overset{l}{.Math.}}.Math.{\beta }_j}.{\left(\frac{m}{t}\right)}^{1/2}.Math.{10}^6$

[0214] Pervanadate Treatment

[0215] P31/Fuj cells were exposed to 1 mM sodium pervanadate or left untreated during 30 min (Sodium pervanadate was prepared by mixing 30% H.sub.2O.sub.2 and 100 mM Na.sub.3VO.sub.4 pH 8.0 at 1:100 ratio during 15 min). Cells were then harvested and lysed as outlined above. After homogenization and protein quantification, treated and untreated cell lysates were mixed to a final protein concentration of 1.0 μg.Math.μL.sup.−1. The proportions used were 0%, 25%, 50%, 75% and 100% of pervanadate treated extracts with 100%, 75%, 50%, 25% and 0% of untreated extracts. Protein mixtures were subsequently subjected to trypsin digestion and phosphopeptide enrichment as described above.

[0216] EGF and IGF Treatment

[0217] KAR results were obtained from a meta-analysis of Supplementary Dataset 2 in Wilkes et al. (2015), supra. Briefly, MCF-7 cells were starved for 24 h, and subsequently treated with 100 ng-mL.sup.−1 EGF or IGF-1 for 0, 5, 10, 30 or 60 min and processed for MS analysis as described in Wilkes et al. (2015), supra. K-scores were calculated as described above.

[0218] Viability Analysis and Sensitivity Coefficient Cell-lines were seeded in 96 well plates (10,000 cells-well-), left overnight and treated with vehicle, or 1 to 1000 nM of AZD-5438 (CDK2i;), GF-109203X (PKCαi; Tocris), PF-3758309 (PAKi; Calbiochem), Trametinib (MEKi; Selleckchem), MK-2206 (AKTi; Selleckchem), KU-0063794 (mTORi; Chemdea) or TAK 715 (P38αi;). Cells were also treated with 0.01 to 10 μM of PKC-412 (PKC/Flt3i; Tocris) or 0.1 to 10 μM of TBB (CK2i; Sigma). After 72 h, cells were stained with Guava ViaCount reagent (Millipore) as indicated by the manufacturer and cell number and viability was measured using a Guava EasyCyte Plus instrument. AML primary cells were thawed as described above, resuspended in MS-5 conditioned IMDM medium, seeded in 96 well plates (20,000 cells-well-1) and treated with vehicle or 1 to 10000 nM of PF-3758309 (PAKi), PKC412 (Flt3/PKCi), CX4945 (CK2i; Selleckchem), Trametinib (MEKi) and TAK 715 (P38i). After 72 h, cells were stained with Guava ViaCount reagent and cell number and viability was measured. All drugs were solubilized in DMSO and all measurements were performed in triplicate. Flow cytometry data were analyzed using CytoSoft (v2.5.7). IC.sub.50 values were calculated using Graphpad PRISM (v5.03). The sensitivity coefficient (SC) was calculated using the equation below (where P.sub.Ci=reduction in proliferation at C.sub.i, IC.sub.50=“in vitro” IC.sub.50 against primary target, and C.sub.i=inhibitor concentration at which proliferation is measured). Data were visualized using Microsoft Excel 2007/2010 or within the R statistical computing environment (v3.0.0), using a combination of the reshape2 and ggplot2 packages.

[00013]$S.Math.C=-{\log }_2\left(\frac{P_{C.Math.i}}{C_i}.Math.{\mathrm{IC}}_{5.Math.0.Math.i}\right)$

[0219] Results

[0220] Linearity and Reproducibility of Signaling Quantification Using Kinase Activity Ranking

[0221] We first investigated the reproducibility and quantitative nature of Kinase Activity Ranking (KAR, which produces K-scores) as a measure of net kinase activity. To this end, the DHL6 cell line was treated with sodium pervanadate (pV) and after lysis mixed with lysates from untreated cells at different proportions (FIG. 1b). To test the reproducibility of the analysis, we performed this experiment on three independent occasions and each was analyzed in analytical triplicate. As expected, pV, a tyrosine phosphatase inhibitor, induced an increase in the K-Score of tyrosine kinases; which consequently ranked higher in samples as a function of pV-treated cells (FIG. 1c-e). Thus these data show that K-scores are quantitative readouts of kinase-substrate groups and suggest that these values may be used to rank kinases based on their activation status in a reproducible manner. We also performed a meta-analysis of published phosphoprotemics data obtained from time-course experiments of cells treated with EGF or IGF (Wilkes et al. (2015), supra). We observed that K-scores for serine/threonine and tyrosine kinases changed upon treatment with these growth factors with the expected kinetics (FIG. 2) indicating that KAR outputs (K-scores) truly reflect the expected kinase activities.

[0222] Kinase Activity Ranking (KAR) Models the Contribution of Kinase Activities to Cell Viability in Haematological Cell-Lines

[0223] We then analyzed eight hematological cancer cell-lines and ranked >100 kinases based on their activation relative to each other. FIG. 3a shows the 60 kinases with greater K-scores, with CDK1, CDK2, ERK1, PAK1 and CK2 ranking the highest (FIG. 3b).

[0224] To investigate whether the K-scores reflected the contribution of kinases to cell viability, we reasoned that if kinases with high K-scores were contributing more to cell survival than those with lower K-scores, a correlation should exist between the K-scores of individual kinases and the impact of their inhibition on cell viability. We therefore measured cell viability of our eight cell-line panel as a function of treatment with nine kinase inhibitors (dose-response curves are shown in FIG. 4). In order to titrate enzyme inhibition in an acute manner, we opted to use pharmacological inhibitors (as opposed to genetic means). However, the effect of small molecule inhibitors on cell behavior is dependent on both the contribution of the target to the pathway flux as well as the affinity of the compound to the target (which is reflected by the in vitro IC.sub.50). To account for this, we normalized the drug-induced inhibition of cell viability by the reported in vitro IC.sub.50 of the compound against its known targets. We termed this value the “Sensitivity Coefficient” (SC), which was calculated as defined herein. By doing this, differences in inhibitor potencies were normalized, and kinase inhibitors with disparate in vitro IC.sub.50 values could be ranked against each other based on the contribution of their targets to cell viability.

TABLE-US-00001 TABLE 1 A panel of haematological cell-lines were treated with the named compounds and their viability measured after 48 hours using the Guava Viacount assay. Ref Name Targets CDK2i AZD5438 CDK1, CDK2, CDK9, GSK3B PKCαi GF109203X PKCA, CDK2 MEK1i GSK1120212 MEK1 mTORCi KU-0063794 mTORCi, mTORC2 PAKi PF03758309 PAK1, PAK4, AMPK PKC/Flt3 PKC412 PKCs, Flt3, Kit P38αi TAK715 P38A CK2i TBB CK2A1 Akti MK2206 Akt1, Akt2

[0225] Ranking the kinase inhibitors shown in Table 1 based on their SCs demonstrated that CDK2i and PAKi were ranked the highest, and overall, the ranking frequencies mirrored those obtained through KAR (FIG. 5a,b). It should be noted that the PKCα inhibitor (PKCαi) also inhibits CDK2 with high potency. A more direct comparison revealed a strong association between K-scores of specific kinases with the SCs against their inhibitor (FIG. 5c). Overall, KAR accurately modeled viability in the eight cell-lines tested, as assessed by linear regression analysis (Pearson r values ranging from 0.49 to 0.90 with a mean of 0.76, P=0.016), and overall, the model showed a statistically significant relationship between the two metrics (r=0.67; P=1.0×10.sup.−10; 4 FIG. 6). These data show that KAR accurately modeled the contribution of kinases to cell viability in in vitro cell-line cultures.

[0226] Kinases Activity Ranking (KAR) Models the Contribution of Kinase Activities to Cell Viability in AML Primary Cells.

[0227] To determine whether the approach may be able to identify regulatory kinases in an independent set of cells, we measured the phosphoproteomes of 45 primary AML biopsies, enriched with cases of normal karyotype, an intermediate risk marker for the disease. KAR of the resulting dataset (FIG. 7a) indicated that, as with the experiments in cell-lines, CDKs, ERK1, CK2A1 and PAK1 were frequently ranked highly in these primary cancer cells (FIG. 7b). In contrast with the cell-lines, however, ATR and PKACA also ranked highly in a number of patients.

[0228] To investigate whether KAR reflected the contribution of their respective kinases to viability in primary AML samples, the cells were treated with inhibitors against P38A, CK2, MEK1 (to inhibit ERK signaling), PAK and PKC412 (which inhibits several kinases including the receptor tyrosine kinase Flt3, whose gene is often altered in AML). We chose to test these compounds because, while the involvement of CDKs in AML is well documented, the contribution of ERK1, PAK and CKs to AML biology is less well understood. PKC/Flt3i and P38αi served as negative controls as the K-scores of their kinase targets were found to be low in most cases. We obtained dose-response curves in 36 primary AML samples (FIG. 8), and IC.sub.50 and SC values were computed (FIG. 7c). On average, primary AML cells were more sensitive to CK2i, PAK1i and MEK1i than to PKC/Flt3i and P38αi (FIG. 7c), consistent with the high K-scores for CK2A1, PAK1 and ERK1 relative to those for PKC/Flt3i and P38αi targets (PKCs/tyrosine kinases and P38A, respectively).

[0229] As with the cell-line data (FIG. 5), the predicted kinase activities reflected the sensitivity of cells to the inhibitors in primary AML blasts. Examples of data for two patient samples are shown in FIG. 7d and the data for the other patients are shown in FIG. 10. Linear regression analysis models showed that in 31 out of 36 biopsies K-scores and SCs were correlated with r>0.6 (FIG. 7e) and with an overall mean r=0.84. Interestingly, the MEK1 K-score was associated with the responses to the MEK1 inhibitor (FIG. 7f), irrespective of the mutation status of NRas or Kras (these having been previously linked to responses to MEK inhibitors in AML). Similarly, the JAK2 K-score (note that JAK2 acts downstream of Flt3) was weakly associated with sensitivity to PKC412 (which inhibits Flt3) across samples, and that this was a better predictor of sensitivity than Flt3 mutation status (FIG. 7g). Overall, these data indicate that models that use K-scores to quantify the contribution of kinases to the signaling output are able to predict the impact of inhibiting a given kinase on cell viability in cell-lines (FIG. 5) and primary cells (FIG. 7 d,e and FIG. 10).

[0230] Differences in Kinase Activities are Associated with the Complexity of Drug Response Phenotypes

[0231] Instead of being resistant or sensitive to single compounds, the primary AML cells showed complex patterns of responses (FIG. 9a). For example, imposing a threshold of a 50% reduction in viability at 1 μM treatment, 10/36 cases were sensitive to at least three inhibitors (groups 1 to 5 in FIG. 9a), whereas 13/36 cases were resistant to all of the compounds (group 13). Five and three cases were sensitive to CKi only (group 11) or to MEK1i only (group 12), respectively. These phenotypes could not be rationalized by considering only the contribution of the target kinase. Therefore, we investigated patterns of kinase activities that may explain the observed heterogeneity in responses. While cytogenetic or FAB (French-American-British) classifications were not different across the response groups, there were marked differences in K-scores (FIG. 11a). For example, tyrosine kinases and PLK1 were increased in sensitive cells (groups 1-5) relative to resistant cells (group 13) and there was an inverse correlation between activities in cells sensitive to MEK1i only or to CK2i only (FIG. 11b-e). In MEK1i-sensitive cells (group 12), RAF K-scores were increased, whereas those for ATR were decreased relative to patient samples with other patterns of inhibition (FIG. 9b,c). Thus, the ratios of RAF to ATR K-scores were significantly greater in this group than in any other response group (FIG. 9d). Similarly, while the CK2A1 K-score was unexpectedly not greater in group 11 (cells sensitive to CK2i only, FIG. 9e), the PKCA K-score was significantly decreased in this group of samples (FIG. 9f) and the CK2A1:PKCA ratio was greater in CK2i sensitive cells relative to the other groups (FIG. 9g). Thus, in addition to activation of the target kinase, exclusive response to a given compound required an absence of activation of pro-survival pathways acting in parallel to the target.