Personalised nutrient dosing with on-going feedback loop

Abstract

A method for providing nutritional supplement information for a subject is proposed, including a sequence of steps in given order and repeated at least once after a time span of at least 2 days or one week for adapting the provided nutritional information: A) taking a sample from the subject; B) analyzing said sample to determine the nutritional status; C) based on the results calculation of nutritional supplements to improve the nutritional status; D) providing individualized nutritional supplement information. This sequence involves the prediction of at least one initial characteristics matrix and multiplication of this matrix weighted with factors, with an initial recommendation vector for the calculation of a target profile vector after a given first time interval from the profile vector as determined in step B), and in each following cycle adaptation by adapting at least one of the characteristics matrix and the weighting factors.

Claims

1. A method for providing nutritional supplements for a subject, said method including a plurality of cycles comprising the following steps in given order: A) taking a sample of at least one of blood, blood serum and blood plasma, from the subject; B) analyzing said sample to determine a nutritional status of the subject as a profile vector (x); C) based on results of step B), calculating required nutritional supplements to improve the nutritional status of the subject in the form of a recommendation vector (d); D) providing at least three individualized nutritional supplements to the subject on the basis of the recommendation vector (d); and repeating the cycle of A)-D) after a time span of at least two days for adapting the individualized nutritional supplements provided in step D) of a second cycle based on the development of the subject at least in the time interval between the cycles, wherein a first cycle step C) comprises a prediction of at least one initial absorption characteristics matrix (A.sub.k, B.sub.p(k), C.sub.k, D.sub.p(k)), and a multiplication of this at least one initial absorption characteristics matrix (A.sub.k, B.sub.p(k), C.sub.k, D.sub.p(k)), weighted with weighting factors (α, β, γ, δ), by an initial recommendation vector (d.sub.1) for a calculation of a target profile vector (x.sub.1.sup.†) after a given first nutritionally supplemented time interval from the profile vector (x.sub.0) as determined in step B), and wherein a difference between the previously calculated target profile vector (x.sub.n.sup.†) and the profile vector (x.sub.n) as determined in the analysis step B) is minimized by adapting and thereby personalizing at least one of the initial absorption characteristics matrix (A.sub.k, B.sub.p(k), D.sub.p(k)) and the weighting factors (α, β), and using the adapted absorption characteristics matrix (A.sub.k, B.sub.p(k), C.sub.k, D.sub.p(k)) and the weighting factors (α, β, γ, δ) for the calculation of a next required recommendation vector (d.sub.n+1) and for the prediction of a next target profile vector (x.sub.n+1.sup.†) after a given next nutritionally supplemented time interval, wherein a dimensionality of the recommendation vector (d) and the target profile vector (x.sub.n.sup.†) is at least 3.

2. The method according to claim 1, wherein the at least three nutritional supplements are selected from the following group: vitamins; vitaminoids; minerals; herbals; botanicals; amino acids; enzymes carbohydrate; fiber, fatty acids; proteins; peptides; terpene-based biological molecules, including carotenoids; steroid-based biological molecules, including hormones; pyrrol-based biological molecules, including tetrapyrrols; alkaloid-based biological molecules, including caffeine; prebiotics, probiotics, flavonoids, and antioxidants.

3. The method according to claim 1, wherein step A) is taking a sample of blood, from the subject by the subject, applying the sample to a solid substrate, and immobilizing and drying the sample on said substrate to form a dried sample.

4. The method according to claim 3, wherein said solid substrate is provided with identification information.

5. The method according to claim 3, wherein said solid substrate is provided with identification information in computer readable form, including a barcode, a QR code, or an OCR-readable text.

6. The method according to claim 1, wherein for step C) individual subject related input information is taken account of.

7. The method according to claim 1, wherein for a second or further cycle further individual subject related input information of the time interval between the cycle and the corresponding actual cycle is taken account of for step C).

8. The method according to claim 1, wherein the cycles are carried out at regular intervals over an extended time span of at least 1 week, with carrying out the cycle at least every two days, and either continuing at the same frequency or reducing the frequency after having reached a steady-state situation.

9. The method according to claim 1, wherein step C) includes a step of manufacturing individualized nutritional supplements based on required nutritional supplements for the subject and step D) includes sending of these individualized nutritional supplements to the subject.

10. The method according to claim 1, wherein step B) includes a step of mass spectrometric analysis of the blood sample, further including a step of chromatographic separation or inductively coupled plasma (ICP), and determining minerals or protein markers characterizing the nutritional status of the subject based the step of mass spectrometric analysis of the blood sample, further including a step of chromatographic separation or inductively coupled plasma.

11. The method according to claim 1, wherein characteristic data specific to the subject is stored for each cycle of steps A)-D) in a database, and is used at least for the calculation in step C) and wherein the calculation in step C) includes taking account of the development of the characteristic data over time.

12. The method according to claim 1, wherein step D) further includes providing to the subject daily-recommended doses of nutrients, which are to be taken up additionally.

13. The method according to claim 1, wherein the calculation in step C) takes account of maximal or minimal recommended possible doses of the individual supplements.

14. The method according to claim 1, wherein step A) is taking a sample of blood by the subject, applying the sample to a solid substrate, provided with sample stabilization agents, and immobilizing and drying the sample on said substrate to form a dried sample, all these steps of step A) being carried out at home, and sending said sample to a place for carrying out step B).

15. The method according to claim 1, wherein step C) input information of the subject further comprises an algorithmic determination of personalized target marker values taking into account at least one of the following: sex, age, ethnic affinity, physical activity, mental activity, general status, nutritional information, climatic information, lifestyle information, molecular information, and genetic/genomic of the subject.

16. The method according to claim 1, wherein for a second or further cycle further input information of the subject of the time interval between the cycle and the corresponding actual cycle is taken account of for step C), in the form of at least one of the following information: use of the supplied nutritional advice or nutritional supplement, physical activity, mental activity, general status, general nutritional information, climatic information, lifestyle information.

17. The method according to claim 1, wherein the cycles carried out at regular intervals over an extended time span of at least 6 months, with carrying out the cycle at least every second week, or at least every month, and either continuing at the same frequency or reducing the frequency after having reached a steady-state situation.

18. The method according to claim 1, wherein the frequency of carrying out the cycles is determined by the results of the analysis in B) and the calculation in C) and wherein step D) further comprises providing information to the subject when a next sample needs to be taken.

19. The method according to claim 1, wherein step C) includes a step of manufacturing individualized nutritional supplements based on required nutritional supplements for the subject and step D) includes sending of these individualized nutritional supplements, together with instructions as to how to consume the supplements, to the subject.

20. The method according to claim 1, wherein step B) includes a step of mass spectrometric analysis of the sample taken in step A) and including a step of chromatographic separation or inductively coupled plasma (ICP), determining minerals and/or protein markers characterizing the nutritional status of the subject based the step of mass spectrometric analysis of the blood sample, further including a step of chromatographic separation or inductively coupled plasma, characterizing required needs for at least one of: vitamins, vitaminoids, minerals, herbals, botanicals, amino acids, enzymes, carbohydrate, fiber, fatty acids, proteins, peptides, terpene-based biological molecules, including carotenoids, steroid-based biological molecules, including hormones, pyrrol-based biological molecules, including tetrapyrrole, alkaloid-based biological molecules, including caffeine, prebiotics, probiotics, flavonoids, and antioxidants.

21. The method according to claim 1, wherein characteristic data specific to the subject in the form of analysis data of the sample taken in step A) and input data of the subject, are stored for each cycle of steps A)-D) in a database, and are used at least for the calculation in step C) and wherein the calculation in step C) includes taking account of a development of the characteristic data over time.

22. The method according to claim 1, wherein the calculation in step C) takes account of maximal or minimal recommended possible doses of the individual supplements and wherein the calculation includes calculating dose ranges for the individual supplements necessary to smoothly shift an actual nutrient profile towards an optimal range.

23. A non-transient computer-readable medium comprising a stored program, executable by a processor, including a processor of a mobile device including a smart phone, configured to perform the steps recited in claim 1.

24. A kit of parts for carrying out the method according to claim 1 comprising at least a non-transient computer-readable medium comprising a stored program executable by a processor, including a processor of a mobile device including a smart phone, configured to perform the steps recited in claim 1, a sample carrier, as well as at least one of the following items: a puncture device; a transfer device; alcohol wipes; a plaster; a gauze swab; a user manual; a specimen bag including a sachet with desiccant; an identity code, a barcode or QR code; preaddressed envelope adapted for the sample carrier.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Preferred embodiments of the invention are described in the following with reference to the drawings, which are for the purpose of illustrating the present preferred embodiments of the invention and not for the purpose of limiting the same. In the drawings,

(2) FIG. 1 shows a schematic general flow scheme of the proposed approach of an integrated platform and ongoing optimisation cycle;

(3) FIG. 2 shows another schematic flow scheme of the proposed approach and

(4) FIG. 3 shows a possible graphical representation of the results on a smart phone of a user, more specifically two sketched mockups of the app, showing plain measurement data (right) and measurement data aggregated by fitness category (left)

(5) FIG. 4 shows the simulated progress of an inner nutrient profile x of an individual k after 20 days of daily intake of a recommended dose “d=(0.3 mmol Calcifediol, 1.2 g EPA, 0.8 g DHA)”; the light grey stripe indicates the normal range of the according nutrient, the dark grey stripe the optimal range; the upper and lower bounds of the simulation's confidence band are depicted as the dotted line, assuming an increase of ±0.5%/day;

(6) FIG. 5 shows the simulated progress of an inner nutrient profile x of an individual k over three cycles of dosage adaption; the light grey stripe indicates the normal range of the according nutrient, the dark grey stripe the optimal range; the upper and lower bounds of the simulation's confidence band are depicted as the dotted line; the black horizontal bars at the end of each cycle indicate the presumed (but random) measurements and mark the start values of the subsequent cycles.

DESCRIPTION OF PREFERRED EMBODIMENTS

(7) FIG. 1 provides an overview of the key components of the proposed approach/platform. A convenient sample kit 12 for home use provides the user 1 with a tool to provide a blood sample 2 for laboratory analysis. The dried blood sample card 2 preferably comprises specific areas 3 for applying and drying droplets of blood thereon, and further the sample card includes identification information 4 which can be user specific but which can also be, in order to keep anonymity, just be characteristic for the card, the link between the user 1 and the corresponding analysis data determined from the sample card 2 only being established downstream in the process. High-throughput laboratory methods in a corresponding analysis and data processing facility 6 form a robust and low-cost diagnostic platform to analyse the molecular composition of the blood sample 3. The analytical results are captured in the digital biobank and interpreted by the recommendation algorithms of the platform intelligence. The output is communicated to the user e.g. via a smart phone 8 through an understandable user interface and actionable recommendations are provided. Dietary supplements 7 with personalised nutrient contents are produced and delivered to the subject 1. Users either act on the recommendations, take the supplements with varying degrees of compliance or do both. Afterwards, the described cycle starts anew.

(8) The impact of the user actions on the molecular composition of his blood describes a user-characteristic feedback. Based on this feedback a user-specific model is generated, comprising the user's demands and resorption characteristics. Through each cycle, this model is updated and rendered more precise via algorithmic approaches.

(9) Any of the three components sample (kit), laboratory analysis and actionable product/information are necessary to fuel and push the recommendation cycle. However, the sample kit for home testing can be as well replaced by a kit which is applied by medically trained persons, the laboratory analysis can be replaced by a point-of-care testing device, the personalised product can be some kind of actionable recommendation, which improves individual fitness and wellness.

(10) The individual elements are characterized as follows:

Sample Kit

(11) The sample kit enables the user to conveniently sample blood himself, normally at home or another place which does not have to be a point-of-care and does not involve medical personnel. So sampling takes place without the need of being medically trained or a third person who is.

(12) The blood sampling using a dried blood sample fulfils three criteria:

(13) It is safe and easy to use. From the moment of unpacking throughout the different steps its use is clear and intuitive, and in addition to that the corresponding technology is well-established and approved.

(14) It allows sampling a defined and reproducible sample of blood.

(15) The blood sample is stabilized after drying, such that it can be sent with a conventional postal service and without the need of cooling. The change in concentration of analytes of interest in the sample does not exceed validated ranges over the course of days without cooling. The general analysability in as far as the present method for nutritional advice/supplement is not affected within this timespan.

(16) The two main parts of the kit 12 are a puncture device, i.e. a device that does the actual puncture to collect blood; and the sample carrier 2, i.e. the device that stabilizes the blood sample 3 and makes it easily transportable.

(17) Puncture Device:

(18) Within the scope of this approach, blood means preferably specifically capillary whole blood. Accordingly, a puncture device is typically a single-use lancet or a lancet pen in combination with disposable lancets. Additionally, a novel class of devices can be considered which emerged through recent developments, which puncture the skin through micro- or nano-tubes while drawing capillary blood through a vacuum into the device.

(19) A typical syringe used for venepuncture is possible as well.

(20) Sample Carriers:

(21) The collected blood is stabilized and sent to the laboratory in the form of dried whole blood. Possible carriers 3 for the dried blood 3 are either filter papers in arbitrary formats, typically as dried blood spot cards, or dried blood spot sticks. Possible products are available from sources such as Tasso Inc under the trade name HemoLink, TAP devices as available by Seventh Sense Biosystems, or micro sampling device is available from Neotyrex.

(22) To stabilize the blood by drying is optional but preferred. Accordingly, other sample carriers are possible as well, including capillary tubes, the typical blood collection tubes, as well as those puncture devices, which draw the blood into a reservoir before releasing it.

(23) Stabilization of the blood can be promoted by agents, which are contained by or coated on the sample carriers 2. Typically they affect coagulation, oxidation or other chemical, physical or biological effects, which destabilise the concentration, accessibility or general analysability of the target analytes in the blood over time.

(24) Depending on the sample carrier, a transfer device can be advisable or makes the process convenient and more reproducible. It allows transferring a defined volume of drawn or collected blood from the puncture device onto or into the sample carrier. This can be for example a precision or non-precision pipette. Capillaries or capillary arrangements which collect and distribute the sampled blood are also covered by this definition.

(25) Content of the Kit:

(26) The kit preferably contains at least one of the following items: Puncture device; Sample carrier 2; Transfer device (if applicable); Alcohol wipes; Plaster; Gauze swab; User manual; Specimen bag including a sachet with desiccant (if applicable); Identity code, barcode or QR code, identifying the particular kit; Preaddressed envelope suitable for the sample carrier; Any other components that are required by regulation in a given country.

(27) Not part of the kit itself but strongly connected to the kit is a software, typically a mobile app, which lets users order, register or track the kit and through which results from the blood analysis are communicated.

Laboratory Analysis

(28) Relevant Markers:

(29) The blood analysis, as one central component of the proposed approach, is carried out to unravel the molecular blood profile of the tested person 1. This is typically composed of markers characterizing the following substance classes: minerals, in particular trace quantity elements small molecules including terpene-based biological molecules, including carotenoids, steroid-based biological molecules, including hormones, pyrrol-based biological molecules, including tetrapyrrols, alkaloid-based biological molecules, including caffeine, prebiotics, probiotics, flavonoids, antioxidants amino acids carotenoids fatty acids steroids vitamins polypeptides and proteins.

(30) The panel of markers is not necessarily limited to the described substance classes. It may also include genetic markers, RNA-based markers or a blood cell count.

(31) The analytical methods for the detection are in most cases mass-spectrometry based and typically coupled to a preceding chromatography procedure. This has the advantage that with one analytical run, multiple markers can be measured simultaneously as long as they belong to a substance class, which is sufficiently homogenous to be detected with the same method and sufficiently heterogeneous to be uniquely distinguishable from each other.

(32) The analytical procedure is identifying the minimal set of analytical methods, which is able to fully qualify and quantify all analytes at the same time with optimised time and cost efficiency.

(33) Automation:

(34) The general laboratory process comprises the following steps: Split the sample into as many sub samples, as different extraction protocols exist. Extract the analytes from the samples with appropriate extraction protocols. Perform chemical modifications like oxidations or derivatisations of those extracts where necessary, in order to make the analytes separable or detectable. Perform the instrumental analysis with the sample. Perform a raw data analysis of the instrumental signal data yielding the full quantities of the defined markers.

(35) Those steps are maximally automated by lab robotics, meshing the different steps on the split samples optimally, such that a total turnover time of a single sample can be minimized and time intervals of critical steps are kept constant across samples to ensure maximal reproducibility.

(36) Depending on the particular method different steps (e.g. extraction and derivatization) can be also directly combined in the same process step, if suitable.

(37) While being processed through the automation, the samples are always uniquely identifiable by the identity code, which is part of the according sample kit.

(38) Full quantification of the markers is performed by quantifications with external standard calibrations.

(39) Technologies:

(40) LC-QQQ-MS: Most low molecular weight molecules and proteinogenic markers are identified by liquid chromatography coupled to a triple-quadrupole mass-spectrometer (MS). Different analytical runs are tailored towards the physical properties of the subsets of the markers, i.e. proteinogenic extracts and water-soluble and fat-soluble compounds. Qualification and quantification of the analytes is typically performed in MRM (multi-reaction-monitoring) mode in the MS. ICP-MS: This method is performed to identify the elemental composition of the samples, i.e. analyse the trace and quantity elements. This is typically done, after the samples underwent a stringent oxidative extraction procedure. GC-FID/GC-Q-MS: This method is used for low molecular weight molecules with a high volatility. It can be applied to identify and quantify fatty acids, which usually undergo an esterification to extract and analyse also the fatty acid composition of the membrane bound triglycerides.

(41) The here described technology, however, is a means to an end to perform the molecular blood analysis. It is from a state-of-the-art perspective a very cost-effective approach to assess a large and flexibly growing set of markers from low blood volumes. As technology evolves over time and the categories of relevant markers expand, particular analytical methods could change or even be replaced by different measurement principles.

(42) Data Analysis:

(43) Instrumental raw data analysis is performed fully automated, including steps such as identification of the analyte spectra, integration of the chromatographic peak areas, performing the calibration of the external standards, quantifying the markers against the calibrations. Outcome of the entire analytical procedure is a table assigning to each marker the original concentration in the sample.

IT Platform

(44) Digital Round-Trip:

(45) The journey of one sample is at any step reflected by the digital platform underlying the entire process. The corresponding digital roundtrip is illustrated in FIG. 2.

(46) It starts with the sample kit 12 represented by its identity code. When the user 1 uses the sample kit he registers it through a mobile app and connects the kit, and accordingly his blood sample with his personal identity. However, knowledge of the kit identity code and ownership of the app are sufficient to claim the identity over time. This allows using the system in a completely anonymous fashion.

(47) In the laboratory 9 the sample is analyzed, which is formally a digitalization of the blood sample properties. The results in terms of determined quantities are exchanged with the digital platform 13, enriching the information existing for the particular sample and user.

(48) Based on the measured values and possibly existing profile information of the user, the recommended daily nutrient doses are calculated by algorithms in the platform intelligence 10. This subcomponent of the system implements the feedback-loop-based learning procedures and integrates the personalised user and peer-group specific nutrient demands and resorption models as explained as well further below.

(49) The calculated doses are pushed to the production site 11, where the personalised supplements are produced.

(50) While the user 1 is consuming his personalised supplements 7, he documents through the mobile application any events related to his general wellness or deviations from regular supplement intake protocol. This information is stored with his profile at the platform intelligence 10, to close the feedback-loop and strive for further personalisation.

(51) API and Data Security:

(52) The digital platform 13 communicates with the data exchanging entities throughout a RESTful API (Representational State Transfer Application Programming Interface). This API is the single and central entry point to the digital platform.

(53) Consuming from the Eco System:

(54) The personalization of the recommended supplement doses is based on the measured blood values. However, the feedback loop is only fully closed if additional information concerning the user's lifestyle etc. is known. The most basic information is the regularity with which the user took the recommended supplements in the interval between individual measurement sequences. In addition to this, any information related to the user's general status, fitness and nutrition is valuable and allows painting a more complete picture. Also other genetic (like particular gene sequencing data) or phenotypic information (e.g. microbiome data, just to name one) can be integrated.

(55) As a large list of services exists which produce large amounts of this kind of data, and make them available throughout APIs, the user is free to authorize the digital platform, to integrate data managed by these services into his data records stored in the digital platform.

(56) Producing for the Eco System:

(57) In the same manner as other services provide data to the digital ecosystem of general status, lifestyle and nutrition tracking, the proposed service potentially allows for letting the user share his data with other services through the API.

Recommendation System

(58) Goal of the Recommendation System:

(59) One platform output is to give a user, who exposes unbalanced nutrient profiles, actionable recommendations that let him balance his nutrient profile in a desirably short period of time. As a consequence, the recommendation system needs to derive from the nutritional profile the daily-recommended doses of nutrients, which are to be taken up additionally or other convenient lifestyle recommendations.

(60) The recommendations are preferably constrained by two criteria: 1. According to scientifically accepted guidelines narrow optimal ranges exist for the individual's nutrient profile. Complying with the recommendations should smoothly shift the actual nutrient profile towards the optimal range, i.e. with a measureable improvement and without overshooting effects. 2. The actual recommendations never exceed accepted maximal or minimal recommended doses.
Calculating Daily Doses:

(61) Let r be a recommendation function and d be the nutrient recommendation profile; a vector with d.sub.i being a quantitative recommendation. This is in its simplest form a daily dose for a particular nutrient. Let's assume for the following, that d comprises only nutrient recommendations. Let x be the inner nutrient profile; a vector comprising the measured nutrient profile with concentration x.sub.j for nutrient j in whole capillary blood. Note that any nutrient i as part of the recommendation profile is not necessarily the same compound as the components j of the inner nutrient profile. For an individual k the actual recommendation profile is calculated by
d=r.sub.k(x,y,α,A.sub.k,β,B.sub.p(k),γ,C.sub.k,δ,D.sub.p(k),ω.sub.k)

(62) A.sub.k and B.sub.p(k) are functions or matrices describing how the concentration of an inner nutrient component x.sub.i is effected by a particular recommendation d.sub.i. A.sub.k is personalised for an individual k while B.sub.p(k) is personalised for k's peer group p(k). α and β are coefficients which assign different weights to A.sub.k and B.sub.p(k) with α+β=1 and α>0 and β>0.

(63) The influence of k's external influences on his inner nutrient profile is also taken into account, by the external influence profile vector y where coefficients y.sub.m characterise particular and quantifiable external influence categories. These categories may include—without being limited to—fitness data, nutritional intakes and patterns and behaviours, as well as seasonal and regional aspects, so generally speaking climatic aspects. C.sub.k and D.sub.p(k) are functions or matrices describing how the concentration of an inner nutrient component x.sub.j is effected by a particular external influence characteristic y.sub.m. C.sub.k is personalised for an individual k while D.sub.p(k) is personalised for k's peer group p(k). γ and δ are coefficients which assign different weights to C.sub.k and D.sub.p(k) with γ+δ=1 and γ>0 and δ>0. If y is a function over time between two measurements, d may also vary over the same time period.

(64) Finally, ω.sub.k describes the optimal inner nutrient concentration profile of an individual k according to scientifically accepted guidelines. k's inner nutrient profile is considered to be in balance if the following equation holds:
ω.sub.k−x=0

(65) It is noteworthy that none of A.sub.k, B.sub.p(k), C.sub.k and D.sub.p(k) as functions of x or y fulfill necessarily the criteria of additivity or homogeneity; i.e. they do not necessarily describe linear systems in terms of the systems theoretical reception of linearity. This reflects that two different recommended nutrient dosages d.sub.v and d.sub.w can have a different individual influence on an inner nutrient component x.sub.j as in the combination of both. As an example: The efficiency of the supplementation of Calcium, largely depends on the inner concentration of Vitamin D3.

(66) The concept of the here described nutrient recommendation function r is easily extendable and will be extended over time to another recommendation function q which derives in the same manner recommendations e to influence the inner nutrient concentration profile x through external factors, as fitness or lifestyle.
e=q.sub.k(x,y,α.sub.e,A.sub.k,β.sub.e,B.sub.p(k),γ.sub.e,C.sub.k,δ.sub.e,D.sub.p(k),ω.sub.k)

(67) Ultimately the concept can be extended such that r and q can be combined with the goal to give combined recommendations in the domains of nutrient supplementations and lifestyle advice to balance the inner nutrient profile:
(d,e)=r.sub.k.Math.q.sub.k

(68) The described approach is tailored towards balancing the inner nutrient concentration profile by deriving recommendations for nutrient supplementation, external factors and the combination of both. Finally, this concept is easily extendable to other objectives in the fields of nutrition, fitness, wellness by defining an appropriate objective function ω.sub.k−x=0 and defining according functions for A.sub.k, B.sub.p(k), C.sub.k and D.sub.p(k) (and others if necessary).

(69) Frequency of Testing:

(70) The frequency of measuring the inner nutrient profile is determined by the recommendation system itself. The effect of a recommendation d on an inner nutrient profile x is statistically predicted. If the confidence bands of the prediction for a given confidence level φ at a future point in time gets too broad, the inner nutrient profile needs to be re-measured. At this time point a new test is pro-actively recommended by the system.

(71) The statistical confidence of the prediction depends mainly on three factors: 1. The on-going degree of personalisation: The less personalised (i.e. the smaller α) the broader are the confidence bands. 2. Time elapsed since the last test: The longer, the less certain the forecast becomes. 3. Changes in the lifestyle, health, wellness, fitness or regional parameters: Changes in intrinsic and extrinsic effects on the physiology of an individual can lead to unpredicted effects on the nutrient demand and consequently profile.
Feedback Loop:

(72) As already mentioned the feedback loop is one essential component of the entire approach: It increases the value for the user over the time he uses the product, through the ongoing personalization of nutrient supplementation by adapting it dosing formula according to the success of its previous dosing formulas. This means that similar to a series of experiments, the delta between actual and target nutrient state is minimized as the dosing formula increasingly better approximates the individual biochemistry of a given user. Its novelty compared to prior art is explicitly in the dynamic nature of this adaption process, i.e. in the fact that the dosing formula changes overtime. A detailed example based in simulated data is given on page 17.

(73) The fundamental assumption is, that every human needs different intake dosages of nutrients daily, but also has different resorption characteristics, making different and tailored recommendations necessary. Determination of the particular resorption characteristics is one challenge, which is tackled by the approach at hand. In the following explanation only the very reduced model
d=r.sub.k.sup.*(x,α,A.sub.k,β,B.sub.p(k),ω.sub.k)
is considered. It is easy to show that the presented principle is also applicable to the generalised mode r.sub.k as well as to q.sub.k and r.sub.k.Math.q.sub.k.

(74) The first time the user uses the product, almost nothing is known of his resorption characteristics. But phenotypic information like gender, age or ethnic affinity allow to group the user k in a particular peer group p(k). Accordingly, a B.sub.p(k) exists which is most suitable for the user. Based on the first measured inner nutrient concentration profile x.sub.0 a recommendation d.sub.1 can be given.

(75) For the first application of the model at t=0 it is α=0.5 and β=0.5 (although not being limited to this start configuration) with A.sub.k=B.sub.p(k). For the next scheduled test at t=1 the outcome of x.sub.1 is predicted and noted as x.sub.1.sup.†. The results of the next test are compared with the prediction and from the deviation x.sub.1.sup.†−x.sub.1 an updated version A′.sub.k is derived, such that the dosing d.sub.1 derived from d.sub.1=r.sub.k.sup.*(x.sub.0, α=1, A′.sub.k, β=0, B.sub.p(k), ω.sub.k) would have led to a prediction x.sub.1.sup.†=x.sub.1. The updated version of A′.sub.k is used for the recommendation at t=2. At the same time α=1−β increases by a portion which depends on the relative difference of x.sub.1.sup.†−x.sub.1. This mechanism is the adaptive learning step, which ensures that the resulting recommendations converge smoothly towards a stable personalization.

(76) Accordingly, A.sub.k is derived from B.sub.p(k) and constantly becomes more precise the more cycle iterations a user undergoes. This is reflected by an increasing weight α and decreasing weight β.

(77) A second benefit is a constant improvement of B.sub.p(k) on the peer group level. The more the content of the digital biobank grows, the more associations of phenotypic peer group information to constantly refined resorption characteristics A.sub.k exists. This allows to diversify the set of B.sub.p(k)'s according to the peer group information and assign from the beginning a better matching and faster improving model to new users from t=0.

(78) Finally, the entire approach allows to not only model and personalise the resorption characteristics (A.sub.k), the influence of external factors (C.sub.k) and subsequently refine the according peer group data (B.sub.p(k) and D.sub.p(k)). Over time, and with a sufficient degree of personalization it even allows to match the user's subjective input, reflecting immediately his fitness, wellness, against his inner nutrient profile and accordingly perform an adaptive learning process on the optimal inner nutrient profile ω.sub.k with the goal to even personalize this component of the model.

(79) An Example:

(80) The described feedback mechanism is now rendered as a specific example. For this purpose the following simplifications are made without any loss of generality: External influences C.sub.k or D.sub.p(k) are not considered. Recommendations e for external factors are not considered. Consequentially, combined recommendations (d, e) ∀e≠0 are not considered. The functional relation of the recommendation profile d on the inner nutrient status x is approximated by linearization. This means in particular that the change of the inner nutrient status, does not depend on the status itself. A depletion of inner nutrient levels is not explicitly taken into account. Cross interactions of nutrients are not considered, i.e. every recommended nutrient effects exactly one inner nutrient.

(81) The following example is given for two inner nutrients: The 25-OH-Vitamin D.sub.3 (Calcifediol) level and the Omega-3 fatty acid level.

(82) Those nutrients can be affected by the direct supplementation of Calcifediol and Eicosapentanoic acid (EPA)/Docosahexanoic acid (DHA): For Calcifediol a blood (serum) level between (50-125) nmol/L is considered to be normal, while a level in the range of (70-100) nmol/L is considered to be optimal. Typical daily doses of supplemented Calcifediol are in the range between 60 mmol and 300 mmol (and can even vary largely from this range). For the Omega-3 status a blood level of (4-12) % EPA+DHA of the total fatty acid content is considered as normal, while (8-11) % are considered to be optimal. These nutrients are typically supplemented as fish oil capsules, whereas the typical ratio of DHA:EPA is 2:3. Usual recommendations for supplementation are in the range of (1-3) g EPA+DHA.

(83) We assume now, that for an individual k∈p(k) at t=0 the following inner nutrient profile was assessed:

(84) $x_0\left(\begin{array}{c}\mathrm{Calcifediol}\\ \mathrm{Omega}\text{-}3\end{array}\right)=\left(\begin{array}{c}30\mathrm{nM}\\ 3\%\end{array}\right)$

(85) Based on this, the user is receiving a recommendation profile:

(86) $d\left(\begin{array}{c}\mathrm{Calcifediol}\left[\mathrm{mmol}\right]\\ \mathrm{EPA}\left[g\right]\\ \mathrm{DHA}\left[g\right]\end{array}\right).$

(87) Furthermore, we assume for this particular user the following absorption characteristics:

(88) $A_k=B_{p\left(k\right)}=\left(\begin{array}{ccc}5& 0& 0\\ 0& 0.05& 0.05\end{array}\right).Math.\left(\begin{array}{cc}\mathrm{nM}\text{/}\mathrm{mmol}& 0\\ 0& \%\text{/}g\end{array}\right).$

(89) Accordingly, the inner nutrient profile is expected to reach the state x.sub.t+1 after a single instance of compliance with the recommendation profile:
x.sub.t+1=x.sub.t+(α.Math.A.sub.k+β.Math.B.sub.p(k)).Math.d
x.sub.t+1=x.sub.t+R.sub.1.Math.d

(90) For the first cycle of adaption the recommendation profile d is to be found which would lift after 20 iterations the inner nutrient profile to

(91) $x_{1,20}^{†}=\left(\begin{array}{c}60\mathrm{nM}\\ 5\%\end{array}\right).$
In its general form the recommendation function would be reduced to

(92) $r_k\text{:}{d}_1=\frac{1}{n}.Math.R_1^{-r}\left(x_{t+n}-x_t\right)$
Were R.sub.1.sup.−r denotes the right inverse of R.sub.1. We find (for α=β=0.5) under the constraint that typically the ratio of EPA:DHA in dietary supplements is 3:2, that

(93) $R_1^{-r}=A_k=B_{p\left(k\right)}=\left(\begin{array}{cc}0.2\mathrm{mmol}\text{/}\mathrm{nM}& 0\\ 0& 12g\text{/}\%\\ 0& 8g\text{/}\%\end{array}\right)$
and finally

(94) $d_1=\left(\begin{array}{c}0.3\mathrm{mmol}\\ 1.2g\\ 0.8g\end{array}\right).$

(95) The simulation of x.sub.t over 20 iterations (days) of the recommended intake is depicted in FIG. 4. The uncertainty of the predicted x is assumed to be

(96) $\pm \frac{\beta }{100}=\pm 0.5\%$
per iteration.

(97) Accordingly a total uncertainty of 20% is reached after 20 days. At this time point a reassessment of x and a comparison with x.sub.1,20.sup.†should be made.

(98) We assume now, that the reassessed data for x.sub.1 (i.e. at the end of cycle 1) was found to be

(99) $x_1=\left(\begin{array}{c}48\mathrm{nM}\\ 4\%\end{array}\right).$
The training coefficient α is then calculated by

(100) 0${\alpha }_1=1-0.5.Math.\sqrt{{\left(\frac{x_1^{†}-x_1}{x_1^{†}-x_0}\right)}_{\mathrm{Calcifediol}}^2+{\left(\frac{x_1^{†}-x_1}{x_1^{†}-x_0}\right)}_{\mathrm{Omega}\text{-}3}^2}$
which gives in this case α=0.68 and accordingly β=0.32. Furthermore, the actual resorption characteristic of the first cycle was

(101) $A_k=\left(\begin{array}{ccc}3& 0& 0\\ 0& 0.025& 0.025\end{array}\right).Math.\left(\begin{array}{cc}\mathrm{nM}\text{/}\mathrm{mmol}& 0\\ 0& \%\text{/}g\end{array}\right)$
as this one solves the equation x.sub.1=x.sub.0+20.Math.A.sub.k.Math.d.sub.1. Accordingly, the resorption characteristics for a second cycle is

(102) $R_2=\left(\alpha .Math.A_k+\beta .Math.B_{p\left(k\right)}\right)=\left(\begin{array}{ccc}3.64& 0& 0\\ 0& 0.033& 0.033\end{array}\right).Math.\left(\begin{array}{cc}\mathrm{nM}\text{/}\mathrm{mmol}& 0\\ 0& \%\text{/}g\end{array}\right).$

(103) Additionally, the uncertainty of the next iteration reduces to

(104) $\pm \frac{\beta }{100}=\pm 0.32\%$
which gives in total 31 iterations (days) until the total uncertainty of the prediction reaches 20%. This whole procedure was simulated over three cycles. All calculated parameters are found in table 1. Note that the actual reassessments x.sub.i are random, yet realistic guesses.

(105) TABLE-US-00001 TABLE 1 Depicted are the main parameters of the simulation over 3 cycles. Note that according to α the number of iterations in cycle 3 should be 100. As it would, however, not be recommended to reassess the actual value of x_i after a longer period than a month, the number of iterations was set to 31. Cycle x.sub.i Iterations d xi.sup.† α A.sub.k 1 $\hspace{1em}\left(\begin{array}{c}30\mathrm{nM}\\ 3\%\end{array}\right)$ 20 $\hspace{1em}\left(\begin{array}{c}0.3\mathrm{mmol}\\ 1.2g\\ 0.8g\end{array}\right)$ $\hspace{1em}\left(\begin{array}{c}60\mathrm{nM}\\ 5\%\end{array}\right)$ .50 $\hspace{1em}\left(\begin{array}{ccc}5.00& 0& 0\\ 0& 0.050& 0.050\end{array}\right)$ 2 $\hspace{1em}\left(\begin{array}{c}48\mathrm{nM}\\ 4\%\end{array}\right)$ 31 $\hspace{1em}\left(\begin{array}{c}0.2\mathrm{mmol}\\ 2.34g\\ 1.56g\end{array}\right)$ 0$\hspace{1em}\left(\begin{array}{c}70\mathrm{nM}\\ 8\%\end{array}\right)$ .68 $\hspace{1em}\left(\begin{array}{ccc}3.64& 0& 0\\ 0& 0.033& 0.033\end{array}\right)$ 3 $\hspace{1em}\left(\begin{array}{c}66\mathrm{nM}\\ 7.6\%\end{array}\right)$ 31 $\hspace{1em}\left(\begin{array}{c}0.14\mathrm{mmol}\\ 1.45g\\ 0.97g\end{array}\right)$ $\hspace{1em}\left(\begin{array}{c}80\mathrm{nM}\\ 10\%\end{array}\right)$ .90 $\hspace{1em}\left(\begin{array}{ccc}3.21& 0& 0\\ 0& 0.032& 0.032\end{array}\right)$

(106) A complete overview of the whole simulation is found in FIG. 5. From the data shown it becomes obvious, that the strategy of adaptive adjustment of the resorption characteristics A.sub.k based on the tracked feedback of the reassessments ensures a targeted repletion of deficient inner nutrient profiles.

(107) Furthermore it is shown that this adaptive adjustment strategy is strictly personalized: Even if two individuals from the same peer-group would have started the whole process with the same nutrient profile x.sub.0 and accordingly with the same dosage recommendation d.sub.1 they would have been treated with different recommendations after the first cycle if they would have differed in their x.sub.1.

(108) Notion of Power Shots:

(109) To make recommendations as convenient as possible for the user, power shots can be used. In cases where nutrients are not supplemented but are solely taken up by natural food, or a mixture of natural food and supplements, the recommended daily doses can be too abstract to understand. Hence, a power shot is a particular amount of one nutrient, or a combination of nutrients. Quantities of foods containing these nutrients can then be expressed in power shots.

(110) The concept is similar to the German notion of a Broteinheiten, defining an amount of carbohydrates in natural food.

(111) This concept can even be expanded, to Power shots as equivalents to particular lifestyle actions, e.g. a walk in the sun is worth x power shots for Vitamin D.sub.3 negative power shots, where a certain action, intake or event can have a negative effect on a particular nutrient profile value.

Mobile App

(112) Functionality:

(113) One integral part of this entire approach is a software—preferably a mobile application—which supports the entire approach and guides the user. For the remainder of the document the software is abbreviated with app.

(114) The software has mainly five functions (see also FIG. 3): 1. As part of the digital round trip, measurement values are returned to the user from the lab through the digital platform to the app. It represents the information in different perspectives, i.e. Plain measurement values Aggregated measurement values for different outcome categories (i.e. a dashboard view, for e.g. physical fitness) History of former measurements, for particular markers or fitness categories. 2. The app displays the recommendations derived from the personalised modelling based on the measurement. It also lets the user track to which extent he complied with the recommendations. 3. It also enables the user to input and track other relevant daily data from the domains, of fitness, lifestyle and nutrition. Ideally the app is also integrated with other third party services. 4. The app keeps track of any data shared with or generated by the digital platform. The user is at any time in full control of his data and is able to revoke access to and from third party services and can also delete data generated by using the system. 5. Through the app the complete lifecycle of kits, samples and supplements can be tracked: Ordering the kit, registering a sample, checking a samples status, ordering supplements, ordering third party products, managing the subscription, being informed of recommendations, including the recommendation to redo a test.

(115) From this description it follows, that the app is an unobtrusive digital companion to the user. This is important, as it closes the feedback loop and drives the confidence levels for increasing personalization.

Personalised Nutrients

(116) Set of Nutrients

(117) To ensure beneficial impact for the users, one needs to make it as easy as possible for them to take action. One concrete product to facilitate action and drive compliance on nutrients is to provide users with dietary supplements that are personalized according to their nutrient needs as determined by our recommendation system.

(118) The nutrients which can be supplemented include, but are not limited to, amino acids, carotenoids, fatty acids, quantity elements, trace elements and vitamins.

(119) Personalization Process:

(120) The nutrients are delivered to the user in concentrations that are personalized according to his nutrient needs as determined by our recommendation system. This can be both in the form of fully personalized nutrient concentrations, i.e. for n=1 (“custom-tailored”), or an appropriate combination of a given set of nutrient building blocks, i.e. for n>1 (“made-to-measure”). This can include as well the use of dietary supplements that are being produced and sold by other parties. a combination of the latter two.

(121) The personalization process can be done both manually as well as automatically via compounding machines. Intermediary stages are possible as well.

(122) Possible Galenic Formulations:

(123) To ensure beneficial impact for the users, one needs to make it as easy as possible for them to take action. This includes the convenience of taking the personalized supplements, which is to a good amount driven by the galenic formulation(s), used. These can include, but are not limited to, liquid, powder and/or compressed (tablets, including effervescent forms thereof, suppositories) versions.

(124) Delivering Personalised Supplements:

(125) The personalized supplements can be delivered conveniently to the users. This can be done via direct home delivery or via third parties. The packaging of the personalized supplements will be convenient and intuitive to facilitate compliant intake. For this, the daily doses will be packed into daily packs ready for convenient intake.

(126) TABLE-US-00002 LIST OF REFERENCE SIGNS 1 user/subject 2 solid substrate for blood sample 3 blood sample 4 barcode on 2 5 envelope 6 high throughput laboratory analysis and data management and nutritional information calculation facility 7 individualized nutritional supplement formulation 8 smart phone or other communication device 9 high throughput laboratory analysis facility 10 database and calculation facility 11 facility for manufacturing individualized nutritional supplement formulations 12 sample kit 13 digital platform