Method for modelling a gut microbiome

Abstract

The invention relates to a method for modelling a gut microbiome, comprising five steps. First, receiving phenotypic data, medical data, lifestyle data and metagenomic data for a set of individuals. Then, receiving phenotypic data, medical data and lifestyle data for a target individual. Next, constructing a vector of the phenotypic data, medical data and lifestyle data for all these individuals. Additionally, determining individuals, from the set of individuals, with maximum similarity to the target individual. Finally, generating at least one microbiome-scale community model based on the metagenomic data of the individuals with maximum similarity to the target individual. The at least one microbiome-scale community model generated provides a molecular fingerprint of gut microbiome activity by simulating food absorption by the gut microbiome. The molecular fingerprint of the activity of the gut microbiome comprises a metabolic flux vector, the metabolic flux vector comprising a distribution of the metabolic fluxes via the reactions of the microbiome-scale community model.

Description

DESCRIPTION TITLE: Method for modeling an intestinal microbiome TECHNICAL FIELD OF THE INVENTION

[0001] The technical field of the invention is that of metagenomics.

[0002] The present invention relates to a method for modeling a gut microbiome of a person and in particular to a method for modeling a gut microbiome of a person whose metagenomic data is not available. TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0003] Metagenomics is a methodological process that aims to study the microbiome. The microbiome is the set of microorganisms and their genes. Microorganisms can be, for example, bacteria, viruses, fungi, yeasts, or plankton, living in a specific environment such as the intestine, ocean, soil, or air.

[0004] A person's gut microbiome is home to a very large number of microorganisms. This collection of non-pathogenic bacteria, viruses, parasites, and fungi present in the gut constitutes the gut microbiome.

[0005] It is known that a person's gut microbiome varies depending on many factors such as diet, lifestyle, or the use of medications such as antibiotics. However, it should be noted that the variation of the gut microbiome depending on a factor is specific to each person. Thus, for the same diet, the gut microbiome of two people will not evolve in exactly the same way. A diet will be understood here as being able to include a single meal or a multitude of meals over one or more days.

[0006] The role of the gut microbiome is becoming increasingly well-known. Numerous studies highlight the links between the gut microbiome and the functioning and / or overall health of the body. Thus, there is a consensus on the importance of the microbiome for a person's well-being.

[0007] A better understanding of how interactions between the microbiome and its host work should therefore help improve the functioning and overall state of the body. It is therefore necessary to be able to understand how a person's gut microbiome reacts to different diets.

[0008] A better understanding of the gut microbiome could help us understand individual responses to diet, in particular. Molecules processed by the microbiome are absorbed by the host to affect target organs. These molecules have local effects on the gut, are retransformed and / or metabolized by microbes, and are excreted in the feces, thus having a wide range of effects on the individual.

[0009] Gut microbiome modeling allows, in particular, the prediction of a variation in the intestinal microbiome upon absorption of one or more foods within a diet. Indeed, by simulating the absorption of one or more foods by the intestinal microbiome using a gut microbiome model, it is possible to obtain a molecular fingerprint of the activity of the intestinal microbiome. A molecular fingerprint of the activity of the intestinal microbiome, under an individual's diet, is a vector of metabolic fluxes comprising a distribution of metabolic fluxes through the reactions of the community model at the microbiome scale.

[0010] In-silico metabolic modeling of microbes allows the translation of functional annotations of (meta)genomic sequences into mathematical models of cellular metabolism. From these mathematical models, the behaviors of systems can be studied under different conditions, also called constraints. To do this, two elements are essential. The first element is the genome-scale metabolic model (commonly called GSMM) or network (commonly called GSMN) of the cellular system under study, which is reconstructed from the functional annotations of the genes encoded in the genomic sequences.This GSMN is a repository of all biochemical reactions catalyzed by cellular enzymes and transport reactions carried out by different cellular transport systems that describes the drainage and production of cellular metabolites in the organism studied necessary to support cell growth, for example amino acids necessary for protein synthesis, nucleotides necessary for DNA replication and transcription, lipids necessary for the construction of. cell membranes, the cofactors necessary for the activity of different enzymes, etc. The second element is the molecular profile of the environment in which the cellular system lives, in the form of different compounds to which the system has access to develop, for example carbon sources such as glucose, maltose, fructose; nitrogen sources such as ammonia, aerobic or anaerobic conditions defined by the presence / absence of oxygen, etc. These molecular profiles are codified in the form of exchange reactions that introduce these compounds into the extracellular environment of the metabolic network at maximum rates defined by the estimated quantities of these compounds in the environment.

[0011] With the biochemical and transport reactions of GSMM and the exchange reactions defining the environmental inputs, a mathematical model of cellular metabolism is constructed by encoding the metabolic network reactions as a stochiometric matrix S where rows represent chemical compounds (m compounds), columns represent reactions (n ​​reactions), and subscripts represent the stochiometric coefficient of each metabolite in each reaction of the network. Stochiometric coefficients are molar ratios in which substrates are transformed into products in a chemical reaction. The matrix S is the most important structural property of metabolic networks. It describes the architecture and topological properties of the system and remains constant under changing conditions that could alter the kinetic parameters or reaction rates.With this matrix S, it is possible to define differential equations for each chemical compound in the network. The differential equations describe the net change in metabolite concentrations as a function of time, which is equal to the differences between the sum of all reaction fluxes that produce the chemical compound and the sum of all reaction fluxes that consume it.

[0012] ^ = S * v

[0013] This system of differential equations integrates the enzyme kinetics through the time derivative, so that the reaction flux vector v will depend on the concentration of the metabolites and a number of kinetic parameters that are unknown for many reactions in the network in the GSMM. It is in this context where, in order to overcome this lack of knowledge of the kinetic information, the metabolic system is placed in a state of equilibrium, which is the The most fundamental constraint in computational modeling of cellular metabolism. This is justified by the fact that metabolic transitions of internal chemical compounds are faster than cell growth rates and dynamic changes in the organism's environment. Consequently, the metabolic fluxes leading to the formation and degradation of any particular chemical compound must be balanced and their sum must be zero:

[0014] 0 = S u'

[0015] Assuming this steady state, the above system of differential equations is converted into a set of linear equations from which it is possible to calculate the vector of reaction fluxes v that defines the metabolic state of the system under the defined environmental conditions. However, this system of linear equations is generally indeterminate, meaning that there is no unique solution, i.e., no unique reaction flux vector v that satisfies the equilibrium condition, but rather a space of feasible flux distributions, i.e., multiple vectors v, in the N-dimensional space of reaction fluxes that satisfy the equilibrium condition. This space of flux distributions can be more or less wide depending on additional constraints. These additional constraints can be imposed on the behavior of the system in addition to the environmental constraints.These environmental constraints are imposed in the form of maximum and minimum values ​​that different reaction fluxes can reach, derived from experimental data, such as the maximal glucose uptake rates calculated experimentally with Metabolic Flux Analyses (MFA); the zero flux through reactions catalyzed by enzymes not expressed in transcriptomic or proteomic datasets, etc.

[0016] There are several ways to explore the space of feasible flow distributions, the most common of which is linear optimization of an objective function. These linear optimization approaches include: Flux balance analyses, commonly referred to as "Flux Balance Analyses" and known as FBA, which attempt to find the flux distribution across model reactions that maximizes the objective function, Parsimonious flux balance analyses, commonly referred to as "Parsimonious FBA" and known as pFBA, which attempt to find the flux distribution between model reactions that maximizes the objective function while minimizing the overall total flux between model reactions, Flux Variability Analyses, commonly referred to as FVA, attempt to find the range of achievable flux values ​​for the model's reactions that maximizes the objective function.

[0017] The mentioned mathematical methods (FBA, pFBA, FVA) explore this space of feasible flux distributions using mathematical optimization of linear programming, which allows finding the best flux distribution v that optimizes a given objective function, which is defined in the form of a biomass equation. This biomass equation is usually written as an additional reaction in the model that reflects the cell's needs (amino acids, nucleotides, lipids, cofactors, possibly macromolecules such as DNA, RNA or proteins in some models) to make one gram of cell dry weight, and is usually obtained by reviewing the relevant scientific literature regarding experimental measurements of the biomass constituents of the studied organisms or by adapting the biomass equation of related organisms if experimental information from the studied organism is scarce.In summary, linear programming optimization on a metabolic system, with biomass production maximization as the objective function, yields the maximum growth yield achievable by the system under a given set of environmental constraints. The environmental constraints are defined by the nutrient inputs as well as the reaction flux distribution (v) that drives the cell growth phenotype, from which it is possible to extract the nutrient consumption or cellular metabolite secretion rates that define the metabolic state of the system.

[0018] This approach relies on the strong assumption that cellular systems have optimized their growth performance under a subset of possible environmental conditions during their evolution, so that maximization of biomass can be considered a guiding principle of metabolic functioning. This hypothesis has been experimentally validated in organisms such as Escherichia coli or Helicobacter pylori, for which experimentally measured growth rates have been shown to correlate with biomass production rates in model simulations with FBA and cell biomass as the objective function.

[0019] Currently, methods for modeling a person's gut microbiome require the collection of metagenomic data from that person, for example by collecting their feces.

[0020] This collection of metagenomic data is expensive and complex. Thus, it greatly limits the possibility of using gut microbiome modeling for a large number of people for whom metagenomic data are not available.

[0021] Therefore, there is a need to provide a method for modeling a gut microbiome of a person whose metagenomic data is not available. SUMMARY OF THE INVENTION

[0022] The invention provides a solution to the problems mentioned above, by enabling the modeling of a person's intestinal microbiome from phenotypic data, medical data and lifestyle data of the person.

[0023] One aspect of the invention relates to a computer-implemented method of modeling a gut microbiome comprising the steps of: Receiving, for each individual of a set of individuals, phenotypic data, medical data, lifestyle data and metagenomic data, Receiving, for a target individual, said target individual being an individual not included in the set of individuals, phenotypic data, medical data, and lifestyle data, Construction, for each individual in the set of individuals and for the target individual, of a vector of phenotypic data, medical data, and lifestyle data, Determination, among the set of individuals, of at least one individual maximizing a similarity with the target individual, the similarity being evaluated from the vectors constructed for each individual of the set of individuals and for the target individual, and Generating at least one microbiome-scale community model based on metagenomic data of individuals maximizing similarity to the target individual, the generated microbiome-scale community model providing a molecular fingerprint of gut microbiome activity by simulating food uptake by the gut microbiome, the molecular fingerprint of gut microbiome activity comprising a vector of metabolic fluxes, said vector of metabolic fluxes comprising a distribution of metabolic fluxes across reactions of the microbiome-scale community model.

[0024] The invention allows the generation of at least one community model at the scale of the personalized microbiome of a target individual, without having access to the metagenomic data of the target individual. Thus, the at least one community model at the scale of the microbiome is generated based on the metagenomic data of people similar to a target individual. This is particularly interesting since the inventors have noticed, during experimental studies, that there is a positive correlation between the similarity determined according to the invention between two individuals from their phenotypic, medical and lifestyle data and the similarity of their metagenomic characteristics. In other words, the people considered similar according to the invention from the phenotypic, medical and lifestyle data also have similar metagenomic characteristics.

[0025] In addition to the characteristics which have just been mentioned in the preceding paragraph, the method for modeling an intestinal microbiome according to one aspect of the invention may have one or more complementary characteristics among the following, considered individually or in all technically possible combinations: the generation of at least one community model at the microbiome scale includes the steps of: Combining the metagenomic data of individuals maximizing a similarity to the target individual by calculating an average plurality of metagenomic taxon abundances from the pluralities of metagenomic taxon abundances of individuals maximizing a similarity to the target individual, Obtaining at least one genome-scale metabolic model for each metagenomic taxon of the average plurality of metagenomic taxon abundances and at least one additional exchange reaction, each genome-scale metabolic model comprising exchange reactions defining environmental conditions, an individual biomass of the metagenomic taxon simulating cell growth, and biochemical reactions and transport reactions encoded by protein-coding genes of the metagenomic taxon genome, Receiving at least one additional exchange reaction, and Constructing a microbiome community model comprising the previously obtained genome-scale metabolic models linked by at least a portion of the exchange reactions comprised in the genome-scale metabolic models, the genome-scale metabolic models exchanging molecules with a common luminal compartment shared by the genome-scale metabolic models of the community model, the community model comprising the additional exchange reactions defining the admission of molecules from the diet into the common luminal compartment and the release of the molecules from the common luminal compartment into a fecal compartment, the method according to the invention further comprises the steps: Obtaining a first molecular fingerprint of the gut microbiome activity by simulating, using the constructed microbiome-scale community model, a monitoring of a first diet by the gut microbiome of the target individual, the first diet being determined from the target individual's lifestyle data, obtaining the first molecular fingerprint comprising the use of a linear optimization method with community biomass maximization as the objective function and with the diet dataset as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the metabolic flux vector comprising a distribution of the metabolic fluxes across the reactions of the microbiome-scale community model, and Obtaining at least one second molecular fingerprint of the gut microbiome activity by simulating, using the constructed microbiome-scale community model, a monitoring of at least one second diet by the gut microbiome of the target individual, the at least one second diet being different from the first diet, obtaining the second molecular fingerprint comprising the use of a linear optimization method with maximization of the community biomass as the objective function and with the diet dataset as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the metabolic flux vector comprising a metabolic flux distribution through the reactions of the community model at the microbiome scale, Identification, from the phenotypic and / or medical data of the target individual and the first molecular fingerprint of the activity of the intestinal microbiome, of a molecular fingerprint of the activity of the intestinal microbiome among the at least one second molecular fingerprint making it possible to modify phenotypic and / or medical characteristics of the target individual, and Determination of dietary recommendations for the target individual to obtain the molecular fingerprint of the identified intestinal microbiome activity. The generation of at least one community model at the microbiome scale includes the steps of, for each individual of the individuals maximizing a similarity with the target individual: Obtaining at least one genome-scale metabolic model for each metagenomic taxon of the plurality of metagenomic taxon abundances and at least one additional exchange reaction, each genome-scale metabolic model comprising exchange reactions defining environmental conditions, an individual biomass of the metagenomic taxon simulating cell growth, and biochemical reactions and transport reactions encoded by protein-coding genes of the genome of the metagenomic taxon, Receiving at least one additional exchange reaction, and Construction of a microbiome community model comprising the previously obtained genome-scale metabolic models linked by at least part of the exchange reactions comprised in the genome-scale metabolic models, the genome-scale metabolic models genome exchanging molecules with a common luminal compartment shared by the genome-wide metabolic patterns of the community model, the community model including the additional exchange reactions defining the admission of molecules from the diet into the common luminal compartment and the release of the molecules from the common luminal compartment into a fecal compartment, and And the method according to the invention further comprises the steps of: for each individual of the individuals maximizing a similarity with the target individual, obtaining a molecular fingerprint of the activity of the intestinal microbiome by simulating, using the microbiome-scale community model constructed for said each individual, a monitoring of a diet, determined from the lifestyle data of the target individual, obtaining the molecular fingerprint comprising the use of a linear optimization method with the maximization of the community biomass as the objective function and with the diet data set as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the vector of metabolic fluxes comprising a distribution of the metabolic fluxes across the reactions of the microbiome-scale community model, and Construction of a composite molecular fingerprint by performing a weighted average of the molecular fingerprints of intestinal microbiome activity obtained for each individual of the individuals maximizing similarity with the target individual, and Determining dietary recommendations for the target individual based on the constructed composite molecular fingerprint. Generating at least one microbiome-scale community model further comprising the steps of, for each individual, maximizing similarity to the target individual: Obtaining at least one genome-scale metabolic model for each metagenomic taxon of the plurality of metagenomic taxon abundances and at least one additional exchange reaction, each genome-scale metabolic model comprising exchange reactions defining environmental conditions, an individual biomass of the metagenomic taxon simulating cell growth, and biochemical reactions and transport reactions encoded by protein-coding genes of the genome of the metagenomic taxon, Receiving at least one additional exchange reaction, Construction of a microbiome community model comprising the previously obtained genome-scale metabolic models linked by at least a portion of the exchange reactions comprised in the genome-scale metabolic models, the genome-scale metabolic models exchanging molecules with a common luminal compartment shared by the genome-scale metabolic models of the community model, the community model comprising the additional exchange reactions defining the admission of molecules from the diet into the common luminal compartment and the release of the molecules from the common luminal compartment into a fecal compartment, The method further comprises the steps of: for each individual of individuals maximizing a similarity with the target individual: • obtaining a molecular fingerprint of the activity of the intestinal microbiome by simulating, using the microbiome-scale community model constructed (53b) for said each individual, a monitoring of a diet determined from the lifestyle data of said each individual, obtaining the molecular fingerprint comprising the use of a linear optimization method with the maximization of the community biomass as the objective function and with the diet data set as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the vector of metabolic fluxes comprising a distribution of the metabolic fluxes across the reactions of the microbiome-scale community model, and • Determination of dietary recommendations allowing modification of phenotypic and / or medical characteristics of said individual from the molecular fingerprint obtained (60c) for each individual of the individuals maximizing a similarity with the target individual, and • Combination of the determined dietary recommendations to obtain dietary recommendations for the target individual. The determination, among the set of individuals, of one or more individuals maximizing a similarity with the target individual consists of a determination of one or more individuals, among the set of individuals, minimizing a Gower distance calculated between the vector constructed for the target individual and the vector constructed for each individual in the set of individuals. Phenotypic data includes at least one of: sex, age, body weight, height, body mass index (BMI), body mass index category, and waist circumference and one or more menopausal status data. Lifestyle data includes at least one of: one or more data on dietary, physical and social behavior, one or more data on smoking status, one or more data on physical activity practices, one or more data on diet and one or more socio-demographic data and one or more geographical data. The medical data includes at least one of: one or more data on a sleep apnea status, one or more data on a diabetes status and a diabetes treatment, one or more data on a dyslipidemia status, one or more data on a dyslipidemia treatment, one or more data on a hypertension status, one or more data on a hypertension treatment, one or more data on a hypertriglyceridemia status, one or more data on a hypertriglyceridemia treatment, one or more data on a hypercholesterolemia status, one or more data on a hypercholesterolemia treatment, one or more data on an arthritis status, one or more data on an arthritis treatment, one or more data on a varicose vein disease status, one or more data on a varicose vein disease treatment, one or more data on the perception of emotional and physical well-being,one or more data on the perception of emotions and data on the perception of sleep quality, one or more data on a type of stool on a Bristol scale., Metagenomic data result from metagenomic sequencing of a fecal sample and include a plurality of metagenomic taxa abundances.

[0026] A second aspect of the invention relates to a computer program product comprising instructions which, when the program is executed by a computer, cause the latter to implement the method according to the invention.

[0027] A third aspect of the invention relates to a computer-readable recording medium comprising instructions which, when executed by a computer, cause the latter to implement the method according to the invention.

[0028] A fourth aspect of the invention relates to a system comprising the means adapted to execute the method according to the invention.

[0029] The invention and its various applications will be better understood by reading the following description and examining the accompanying figures. BRIEF DESCRIPTION OF THE FIGURES

[0030] The figures are presented for information purposes only and in no way limit the invention. Figure 1 shows a schematic representation of a method for modeling an intestinal microbiome according to the invention. Figure 2 shows a schematic representation of a community model obtained by a method of modeling an intestinal microbiome according to the invention. Figure 3 shows an example of a list of data that can be used in the method according to the invention. Figures 4A, 4B, and 4C show examples of Gower distances calculated between a target individual and individuals maximizing similarity to the corresponding target individual. Figure 5 shows an example of Gower distance calculated between different individuals in the set of individuals. Figure 6 shows an example of the distribution of the median Gower distance between all pairs of subjects in a set of individuals compatible with the invention. Figure 7 shows examples of Z-scores calculated for different meals. Figure 8 shows characteristics of an example of a set of individuals compatible with the method according to the invention. DETAILED DESCRIPTION

[0031] Figure 1 shows a schematic representation of a method for modeling a person's intestinal microbiome according to the invention. "Modeling" means providing a molecular fingerprint of the activity of the community under a given diet or more generally under a given lifestyle, such as the practice of one or more physical activities for example.

[0032] The method according to the invention makes it possible to obtain a molecular fingerprint of the intestinal microbiome of an individual, in order to have an understanding of the functioning of the intestinal microbiome of said individual and to be able to modify its behavior, for example by modifying the diet as a starter.

[0033] The method according to the invention also allows the modeling of the intestinal microbiome of a target person without requiring metagenomic data from this person. Indeed, the method uses the metagenomic data of people similar to the target person to generate at least one community model at the microbiome scale.

[0034] The method can reduce symptoms of discomfort, such as bloating, during the digestion of a meal. The method can also improve the metabolic health of an individual whose gut microbiome is negatively affected, for example, by taking medications such as antibiotics. The method can also improve an individual's metabolic health for purely aesthetic purposes. For example, to combat hair loss, the method can aim to increase the abundance of biotin.

[0035] The method according to the invention is "computer-implemented". By "computer-implemented" is meant that the steps, or at least some of the steps, are executed by at least one computer or processor or any other similar system. Thus, steps are carried out by the computer, possibly fully automatically, or semi-automatically. In examples, the triggering of at least some of the steps of these methods may be carried out by user-computer interaction. The level of user-computer interaction required may depend on the level of automation intended and balanced against the need to implement user wishes. In examples, this level may be user-defined and / or may be predefined.

[0036] A typical example of computer implementation of a method is to execute the method with a system adapted for this purpose. A system 200 may be configured to implement all the steps of the method and / or its various embodiments according to the invention. To do this, the system 200 comprises a memory and a computing unit, the memory being configured to store instructions which, when executed by the computing unit, cause the computing unit to implement the steps of the method according to the invention and / or the various embodiments of the method according to the invention. The system 200 further comprises at least one network interface for communicating with remote entities, i.e. for sending and receiving data to said entities, via at least one network. The entities may be servers or computers storing data.The data may include phenotypic data, medical data, lifestyle data, or metagenomic data. The data may be stored on the same server or on different servers / databases.

[0037] The method for modeling an intestinal microbiome according to the invention comprises a first step of receiving 10, for each individual of a set of individuals, phenotypic data, medical data, lifestyle data and metagenomic data. The reception 10 of the data can be carried out by the system 200. The method for modeling an intestinal microbiome according to the invention can also be used with only part of this data. For example, the phenotypic data and / or the medical data and / or the lifestyle data can be optional.

[0038] The method for modeling an intestinal microbiome according to the invention comprises a second step of receiving 20, for a target individual, phenotypic data, medical data and lifestyle data. The reception 20 of the data can be carried out by the system 200. The terms "target individual" designate an individual not included in the set of individuals whose data was received in step 10. The method for modeling an intestinal microbiome according to the invention can also be used with only a part of this data. For example, the phenotypic data and / or the medical data and / or the data lifestyle changes could be optional. The target individual is an individual for whom metagenomic data are not accessible and for whom gut microbiome modeling is necessary.

[0039] In a variant of the invention, the phenotypic data, for each individual of the set of individuals and for the target individual, comprise at least one data item from among: a sex, an age, a body weight, a height, a body mass index (BMI), a body mass index category, a waist circumference and one or more data items on a menopause status.

[0040] In a second variant of the invention, compatible with the previous variant, the medical data, for each individual of the set of individuals and for the target individual, comprise at least one data item from among: one or more data items on a sleep apnea status, one or more data items on a diabetes status and a diabetes treatment, one or more data items on a dyslipidemia status, one or more data items on a dyslipidemia treatment, one or more data items on a hypertension status, one or more data items on a hypertension treatment, one or more data items on a hypertriglyceridemia status, one or more data items on a hypertriglyceridemia treatment, one or more data items on a hypercholesterolemia status, one or more data items on a hypercholesterolemia treatment, one or more data items on an arthritis status, one or more data items on an arthritis treatment, one or more data items on a varicose vein disease status,one or more data on treatment of varicose vein disease, one or more data on perception of emotional and physical well-being, one or more data on perception of emotions and data on perception of sleep quality, one or more data on a type of stool on a Bristol scale.,

[0041] In a variant of the invention, compatible with the preceding variants, the lifestyle data, for each individual of the set of individuals and for the target individual, comprise at least one data item from among: one or more data items on dietary behavior, on physical behavior or on social behavior, one or more data items on smoking status, one or more data items on one or more physical activity practices, one or more data items on diet, one or more socio-demographic data items and one or more geographical data items. The socio-demographic data items are, for example, housing, civil and family situation, the language(s) spoken, the level of education. of an individual. Geographic data may include, for example, an individual's place of birth, place of residence or place of work, or historical geographic data such as the different places of residence or places visited over the past few weeks, months or years.

[0042] In one embodiment of the invention, consistent with the preceding embodiments, the diet data comprises the abundance of at least one molecule included in at least a portion of the individual's diet. The abundance of molecules in the diet data set corresponds to the absolute nutrient composition per person per day. The diet data can be obtained from various dietary data collection methods, such as food frequency questionnaires or 24-hour dietary recalls that collect individuals' dietary habits. The foods in these dietary records are translated into quantitative macromolecule and micromolecule profiles. The quantitative macromolecule and micromolecule profile of an individual's diet is referred to as "diet data."Generating a quantitative profile from a list of foods can be done based on the results of laboratory experiments and / or on knowledge from the scientific literature. Translating foods into a quantitative profile can also be done using integrated food databases such as the United States Department of Agriculture's FoodData Central (https: / / fdc.nal.usda.gov) which contains the molecular breakdown of these different foods, in order to be able to define personalized molecular profiles of an individual from their dietary data.

[0043] In a variant of the invention, compatible with the preceding variants, the metagenomic data result from a metagenomic sequencing of a fecal sample and comprise a plurality of abundances of metagenomic taxa. A metagenomic sequencing of a fecal sample comprises an extraction of nucleotides, for example DNA (deoxyribonucleic acid) and / or RNA (ribonucleic acid) from a fecal sample, a library construction and a sequencing in order to obtain, after a taxonomic assignment, abundance tables and therefore the identification and quantification of the species forming the intestinal microbiome. A metagenomic dataset can be obtained from a fecal sample using a sequencing platform such as Nanopore or a high-throughput sequencing platform such as Illumina®'s NovaSeq® products. A metagenomic dataset is, for example, one or more abundance tables comprising the abundance of at least one taxon present in the fecal samples. A taxon is understood to mean a taxonomic unit, such as, for example, the bacterium Akkermansia muciniphila present in a relative abundance of 1.5% (15,000 metagenomic reads classified as A. muciniphila / 1,000,000 total metagenomic reads in the fecal sample). The abundance of the taxa may be a relative abundance, for example, as a percentage of the microbial composition. In a first embodiment, compatible with the previous variants, the set of individuals whose data are received 10 have varied characteristics.“Various characteristics” means data having values ​​presenting a great diversity in different domains. The determination, in step 40, of several individuals maximizing a similarity with the target individual will make it possible to find among this set of individuals, the individuals necessary for the generation of a personalized community model at the scale of the intestinal microbiome of the target individual. This mode of implementation allows the maximization of the number of individuals present in the set of individuals, since no selection criteria are applied to integrate the set of individuals. In addition, this set of individuals, due to its great diversity, makes it possible to increase the probability that one or more individuals in the set of individuals is similar to the target individual.

[0044] In a second implementation mode, the individuals in the set of individuals may constitute a population having one or more similar characteristics. “Similar characteristics” means data having substantially similar values ​​in at least one domain. For example, if the target individual has a specific medical characteristic, such as diabetes, the set of individuals could be made up of individuals having the same diabetes. In this example, the method could be specifically dedicated to people suffering from diabetes. In this second implementation mode, the individuals in the set of individuals having one or more similar characteristics may come, after filtering, from the individuals in the set of individuals having varied characteristics. Thus, for example, among the thousands of individuals having varied characteristics, there It is possible to filter all women suffering from diabetes, to obtain a set of individuals with two similar characteristics: sex and diabetic status.

[0045] The method for modeling an intestinal microbiome according to the invention comprises a third step of construction 30, for each individual of the set of individuals and for the target individual, of a vector of the phenotypic data, the medical data, and the lifestyle data. The term "vector" designates a container of elements ordered and accessible by indices, the size of which is dynamic. Thus, an index is assigned to each piece of data of the phenotypic, medical, and lifestyle data. In addition, the data can be modified in order to optimize their storage and the determination 40 of individuals maximizing a similarity with the target individual.

[0046] In a variant of the invention, compatible with the previous variants, the vector constructed in step 30 comprises at least the data of: Sex, Age, Obesity categorization, Weight, Body Mass Index, Waist measurement, Diabetes status, Diabetes treatment status, Cognitive restriction, Emotional eating score, Level of professional activity, Diet quality (alternative healthy diet index), and Dietary diversity (dietary diversity score).

[0047] In a variant of the invention, compatible with the previous variants, the vector constructed in step 30 comprises at least the data provided in Figure 3. In addition, additional variables can be used to complement: the eating behavior variable: a cognitive restraint variable, an emotional eating score, an uncontrolled eating score, and a general health variable; quality of life and physical, mental, and emotional limitations: a physical functioning variable, a role limitations due to physical health variable, a role limitations due to emotional problems variable, a social functioning variable, a pain variable, an emotional well-being variable, and an energy fatigue variable; physical activity: work activity, sports activity, and leisure activity; diet quality and diversity: dietary diversity with a dietary diversity score and diet quality or an alternative healthy eating index; and sleep: sleep quality.

[0048] The method for modeling an intestinal microbiome according to the invention comprises a fourth step 40 of determining individuals maximizing a similarity with the target individual among the set of individuals. The similarity is evaluated from the vectors constructed in step 30 for each individual of the set of individuals and for the target individual. More concretely, the constructed vectors will have a similar architecture, that is to say that the data of the same type will have the same index in the different constructed vectors. For example, the data “weight of a person” will have the same index in the different constructed vectors. Then, the similarity between each data item of the constructed vectors is calculated. Finally, an average similarity is calculated from the different similarities calculated previously. This average similarity is the similarity between the two constructed vectors 30.During this step, a predefined number of similar individuals to be determined can be predetermined. Thus, this step will, for example, determine the 1, 3, 5 or 10 individuals maximizing the similarity with the target individual among the set of individuals. Alternatively, a minimum similarity threshold can be predetermined. Thus, this step will, for example, determine the. individuals having a similarity greater than the minimum similarity threshold with the target individual among the set of individuals. These two alternatives can be combined during this step to have a minimum or maximum number of individuals to determine and / or a minimum similarity threshold to respect.

[0049] In a variant of the invention, compatible with the preceding variants, the determination 40, among the set of individuals, of one or more individuals maximizing a similarity with the target individual consists of a determination of one or more individuals, among the set of individuals, minimizing a Gower distance calculated between the vector constructed for the target individual and the vector constructed for each individual of the set of individuals.

[0050] In a variant of the invention, compatible with the preceding variants, the determination 40, among the set of individuals, of one or more individuals maximizing a similarity with the target individual comprises the identification of a predetermined number of individuals, for example ten individuals, among the set of individuals, maximizing a similarity with the target individual. To carry out this identification, the Gower distance can be used on the basis of a data set as presented previously. The Gower distance is an index that uses numerical and categorical data to measure the degree of dissimilarity between two samples. The Gower distance varies between a value equal to zero, when two samples are identical, and a value equal to one, when two samples are completely dissimilar. In this variant of the invention, the following formula can be used:

[0052] With :

[0053] Sj, the partial similarity between data x t and Xj.

[0054] Furthermore, it is possible to note that the similarity between two samples can be calculated differently if the data are numerical or categorical. Several known formulas are compatible with the method according to the invention.

[0055] Figures 4A, 4B, and 4C show examples of Gower distances calculated between a target individual and individuals maximizing similarity with the corresponding target individual. Figure 4A illustrates for example that a first target individual has a Gower distance equal to 0.0982100 with individual 470081 101 . Figure 4B illustrates for example that a second target individual has a Gower distance equal to 0.0540370 with individual 910072261 . Figure 4C illustrates for example that a third target individual has a Gower distance equal to 0.0824474 with individual 5504691 1 1 . The proximity of the target individuals to the individuals maximizing a similarity is highlighted both by the low values ​​of the Gower distance and by the comparison with the distance matrix in Figure 5, which is an extract of the complete matrix summarized in Figure 6. Figure 5 illustrates the Gower distance between different individuals in the set of individuals identified by an identifier. For example, the Gower distance between individual 10001 1101 and individual 100036071 is equal to 0.2152812.Figure 6 illustrates a distribution of the median Gower distance between all pairs of subjects in a set of individuals that can be used in the method according to the invention. Figure 6 therefore shows that the average distance between the samples is preferably relatively small and therefore that the homogeneity of the set of individuals also allows sufficient diversity to find close matches with the target individual. In a variant, compatible with the previous variants, the set of individuals comprises for example between 500 and 1000 individuals, for example 750 individuals.The method for modeling an intestinal microbiome according to the invention comprises a fifth step of generating 50 one or more microbiome-scale community models based on the metagenomic data of the individuals maximizing a similarity with the target individual, the generated microbiome-scale community model making it possible to obtain a molecular fingerprint of the activity of the intestinal microbiome by simulating an absorption of food by the intestinal microbiome, the molecular fingerprint of the activity of the intestinal microbiome comprising a vector of metabolic flows, said vector of metabolic flows comprising a distribution of metabolic flows through reactions of the microbiome-scale community model.

[0056] In one variant of the invention, consistent with previous variants, generation 50 of the microbiome-scale community model comprises four sub-steps.

[0057] A first sub-step 51a consists of combining metagenomic data of individuals maximizing a similarity with the target individual in calculating an average plurality of metagenomic taxon abundances from the pluralities of metagenomic taxon abundances of the individuals maximizing a similarity to the target individual.

[0058] In a variant of the invention, compatible with the preceding variants, the average plurality of abundances of metagenomic taxa is a weighted average plurality of abundances of metagenomic taxa of the individuals according to the value of the similarity of these individuals. Thus, the plurality of abundances of metagenomic taxa of an individual considered to be closer to another individual will have a greater weight than the plurality of abundances of metagenomic taxa of the other individual in the calculation of the average plurality of abundances of metagenomic taxa.

[0059] A second sub-step 52a of the generation 50 of at least one microbiome-scale community model consists of obtaining at least one genome-scale metabolic model. These genome-scale metabolic models can be requested by the system 200 from a database or a server, after receiving metagenomic data in step 10 and combining them in step 51. The system 200 receives in response the genome-scale metabolic models (GSMMs), one for each taxon of the metagenomic dataset. The genome-scale metabolic models can be obtained using a repository called AGORA (http: / / dx.doi.org / 10.1038 / nbt.3703) which contains GSMMs for more than 800 microorganisms of the gut microbiome, reconstructions that have received varying degrees of manual processing in terms of reaction content and biomass equation definition which substantially improves its predictive ability compared to automatically generated GSMMs.

[0060] A genome-scale metabolic model of a metagenomic taxon includes genes, reactions, and metabolites. A genome-scale metabolic model allows for computational simulation of the taxon's cell growth. It defines environmental conditions, exchange reactions, and an individual biomass of the metagenomic taxon. Exchange reactions are pseudo-reactions that make different metabolites available for biochemical and transport reactions encoded by protein-coding genes in the taxon's genome. For example, an exchange reaction for L-Glucose (EX_Glc_L: =>L_Glc[e]) makes the metabolite L-glucose (L_Glc) available in the environment extracellular compartment of GSMM ([e]), from which GSMM-specific transport reactions can transport the metabolite to the cytoplasmic / intracellular compartment (e.g., by an ATP-dependent transport system represented by the reaction L_Glc_ABCtrp: L_Glc[e] + ATP[c] => L_Glc[c] + ADP[c] + Pi[c]), and intracellular reactions can transform the metabolite for a given metabolic purpose (e.g., in the case of L-glucose, the metabolite is channeled by central metabolism reactions that allow the generation of energy (ATP) and precursors for different biomass constituents).Environmental conditions are the nutrients available to simulate cell growth (e.g. glucose in the example above) and the individual taxon biomass is a specific reaction that reflects the cell's needs (amino acids, nucleotides, lipids, cofactors, possibly macromolecules like DNA, RNA or proteins in some models) to make one gram of cell dry weight (cell growth). As an example, the biomass equation for the bacterium A. muciniphila, commonly found in the human gut microbiome, is shown below:.

[0061] 0.0030965 nadp[c] + 35.5403 h2o[c] + 0.21909 glu_L[c] + 0.0030965 ACP[c] + 0.0030965 coa[c] + 0.0030965 thmpp[c] + 40.1 102 atp[c] + 0.0030965 nad[c] + 0.13541 gtp[c] + 0.0030965 amet[c] + 0.21909 gln_L[c] + 0.016021 dttp[c] + 0.084104 ctp[c] + 0.016021 datp[c] + 0.016021 dctp[c] + 0.016021 dgtp[c] + 0.0030965 adocbl[c] + 0.20083 asp_L[c] + 0.42793 ala_L[c] + 0.50987 gly[c] + 0.0030965 10fthf[c] + 0.0030965 thf[c] + 0.079264 his_L[c] + 0.37539 leu_L[c] + 0.21 107 thr_L[c] + 0.076146 cys_L[c] + 0.0030965 mqn8[c] + 0.2467 arg_L[c] + 0.20083 asn_L[c] + 0.1278 met_L[c] + 0.0030965 ca2[c] + 0.010648 pg180[c] + 0.010648 clpn180[c] + 0.010648 pgai17[c] + 0.010648 clpnai17[c] + 0.010648 pgi17[c] + 0.010648 clpnil 7[c] + 0.0030965 cobalt2[c] + 0.090832 utp[c] + 0.17946 ser_L[c] + 0.0030965 q8[c] + 0.0030965 cu2[c] + 0.28544 lys_L[c] + 0.0030965 fe2[c] + 0.0030965 fe3[c] + 0.0030965 fad[c] + 0.0030965 2dmmq8[c] + 0.0030965 gthrd[c] + 0.15452 phe_L[c] + 0.18435 pro_L[c] + 0.12068 tyr_L[c] + 0.0030965 pheme[c] + 0.02501 1 colipa[c] + 0.0030965 k[c] + 0.2418 ile_L[c] + 0.02501 1 udcpdp[c] + 0.0030965 5mthf[c] + 0.0030965 mg2[c] + 0.0030965 mn2[c] + 0.0030965 ptrc[c] + 0.0030965 pydx5p[c] + 0.025011 PGP[c] + 0.010648 pe180[c] + 0.010648 peai17[c] + 0.010648 pei17[c] + 0.0030965 ribflv[c] + 0.0030965 so4[c] + 0.0030965 sheme[c] + 0.0030965 spmd[c] +. Tl 0.047202 trp_L[c] + 0.35223 val_L[c] + 0.0030965 zn2[c] + dnarep[c] + proteinsynth[c] + 0.0030965 cl[c] + rnatrans[c] -> 40 h[c] + 39.9969 pi[c] + 0.4846 ppi[c] + 40 adp[c] + 0.0030965 apoACP[c] + 0.0030965 cbi[c] + biomass[c] + 0.0030965 dmbzid[c] + 0.025011 PGPm1 [c]

[0062] A third sub-step 53a of the generation 50 of the microbiome-scale community model is to receive at least one additional exchange reaction, which will be used when building 54 the microbiome-scale community model.

[0063] A fourth substep 54a of generation 50 of the microbiome-scale community model is to build a microbiome-scale microbiome community model. The microbiome-scale community model is created by linking individual GSMMs through their exchange reactions. Individual GSMMs exchange molecules between their individual cytoplasmic environment and a luminal compartment shared by all GSMMs in the community. This allows for the exchange of molecules between GSMMs and ultimately for the linking of GSMMs. Additional exchange reactions allow for the simulation of the release of molecules into an external compartment such as the fecal compartment.These additional reactions are linked to the individual GSMM and may be received during step 54a or during a previous step, or may constitute a step in itself. In another embodiment, this step may be performed from metagenomic taxon abundances and genome-wide metabolic models gathered at higher levels of the taxonomic hierarchy than species-level abundances, e.g., genus, family, order, class, phylum. In a first embodiment, the metagenomic gene abundances are derived from non-redundant gene catalogs of the human gut microbiome and the genome-wide metabolic models are derived from functional annotations of said non-redundant gene catalogs.In another embodiment, building a microbiome community model includes metagenomic taxa specific to the target individual, product of de-novo genome assembly with the metagenomic dataset generated from the corresponding fecal sample and reconstructing the GSMM from the functional annotation of the genome.

[0064] With the microbiome-scale community model, it is possible to obtain a molecular fingerprint of the community activity. The term "community" refers to the set of taxa in the combined metagenomic data. The metabolic flux vector of the metabolic fingerprint is calculated based on the microbiome-scale community model created as described above. The calculation of the metabolic flux vector includes the calculation of admissible flux distributions that satisfy the steady-state constraint and the use of a linear optimization method with maximization of the community biomass as the objective function and with the diet data set as the external constraint. This linear optimization method can be, as described in the prior art, FBA, pFBA, FVA or any other suitable linear optimization method.Community biomass is the sum of the individual biomasses of taxa included in the received GSMMs from taxa, weighted by the abundances of metagenomic taxa. In other words, if a metagenomic taxon is present in greater abundance than another metagenomic taxon, its individual biomass will contribute more to the community biomass. The diet dataset is used as an external constraint, i.e., it defines the input metabolites of the community model. The resulting metabolic flux vector Vec includes a distribution of metabolic fluxes across community reactions that maximizes community biomass, which constitutes the molecular spring of community activity under the defined dietary constraints.Below is shown an example of a metabolic flux vector (Flux column) produced from pFBA simulations on a community model of an individual comprising 402479 reactions from 252 metagenomic taxa identified in the simulated fecal sample under a dietary profile derived from the individual's Food Frequency Questionnaire (FFQ). For simplicity, only 44 reactions are shown, corresponding to the dietary exchange reactions that define the consumption of food metabolites by the community (Diet_EX_ reactions), the communityBiomass reaction corresponding to the objective function that is maximized by the pFBA step, a subset of GSMM reactions of two microbial taxa in the community (Acidaminococcus fermetans and Dorea formicigenerans) with their corresponding biomass reaction (biomass400 and biomass137 in Table 1) and one. subset of exchange reactions from the lumen to the fecal compartment (EX_ reactions in Table 1).

[0065]

[0165] The invention therefore makes it possible to specifically trace the source of chemical compounds from the diet to the fecal compartment through the reaction fluxes in the community model, as illustrated in Figure 2. As shown in Figure 2, it is possible, using the invention, to trace TMA (trimethylamine) and TMAO (trimethylamine oxide) from the fecal compartment to the dietary source (dietary choline (CHOL) being the source of TMA / TMAO in the fecal compartment). For example, with a European diet, two bacteria (S. wadsworthensis_3_1_45B and E. lenta_DSM_2243) can be shown to use luminal TMAO as an electron acceptor in respiration, releasing TMA and decreasing TMAO in the fecal compartment.

[0166] In this same variant, the method according to the invention further comprises three additional steps.

[0167] A first additional step 60a consists of obtaining a first molecular fingerprint of the activity of the intestinal microbiome. This first molecular fingerprint is obtained by simulating, using the generated microbiome-scale community model 50, the monitoring of a first diet of the intestinal microbiome of the target individual, the first diet being determined from the data on the lifestyle of the target individual. Monitoring a diet is defined as the absorption of food as part of this diet. Thus, the absorption of food as part of a first diet is simulated in this step 60a.

[0168] A second additional step 70a consists of obtaining molecular fingerprints of the activity of the intestinal microbiome by simulating, using the same model of the microbiome-scale community model as that used in step 60, the monitoring of different diets of the intestinal microbiome of the target individual. These diets are different from the diet used in step 60. These diets can be determined from data from a food data catalog. In addition, several simulations are performed at this step 70. Indeed, the food intake for each of the different diets is simulated at this step 70. Thus, the molecular fingerprint of the activity of the intestinal microbiome will be different for each diet among the different diets simulated at this step.

[0169] A third additional step 80a consists of identifying, among the molecular fingerprints of the activity of the intestinal microbiome obtained in step 70a, one or more molecular fingerprints of the activity of the intestinal microbiome having an interest for the target individual. The interest of a molecular fingerprint of the activity of the intestinal microbiome is determined from the phenotypic and / or medical data of the individual. In other words, the identification of one or more molecular fingerprints of the activity of the intestinal microbiome is carried out according to its beneficial effect on the functioning and the overall state of the target individual. For example, if a target individual has recently lost weight, then a molecular fingerprint promoting the stabilization of the new weight will be considered to be of interest for the target person.The “interest” of a molecular fingerprint may be predetermined, determined automatically based on the individual's phenotypic and / or medical data, or determined manually by an operator. The interest of a molecular fingerprint of gut microbiome activity may also be determined by comparing the molecular fingerprints of gut microbiome activity obtained in step 70a with the molecular fingerprints of the set of individuals for which data were received in step 10, for example, using a z-score calculated for certain targeted metabolites. The targeted metabolites may be determined based on a predetermined minimum or maximum richness for all the metabolites present in the molecular fingerprint. Thus, the targeted metabolites are metabolites whose abundance is to be modified.Targeted metabolites can also be determined based on a predetermined maximum difference between the richness of a metabolite present in the first molecular fingerprint of gut microbiome activity and the at least one second molecular fingerprint.

[0170] In a variant of the invention, compatible with the previous variant, the method according to the previous variant further comprises the determination 90a of dietary recommendations enabling the molecular fingerprint of the activity of the identified intestinal microbiome to be obtained 80a. The terms "enabling to obtain" mean in this application that the dietary recommendations enable the differences between the molecular fingerprint of the target individual and the identified molecular fingerprint 80a to be reduced.

[0171] In one embodiment of the invention, generation 50 of at least one microbiome-wide community model is performed by constructing a microbiome community model for each individual of the individuals maximizing similarity to the target individual.

[0172] From these microbiome community models constructed for each individual of the individuals maximizing a similarity with the target individual, a molecular fingerprint of the community activity at the microbiome scale in the context of the diet is calculated for each of these models. Then, two variants are possible. In the first variant, for each individual of the individuals maximizing a similarity with the target individual, the community model at the microbiome scale constructed is used to obtain, in step 60b, a molecular fingerprint of the activity of the intestinal microbiome by simulating a monitoring of a diet determined from the data on the lifestyle of the target individual. Then, from the molecular fingerprints obtained in step 60b, a composite molecular fingerprint is constructed 70b by performing, for example, a weighted average of the vectors of the metabolic fluxes molecular fingerprints obtained in step 60b.Finally, dietary recommendations based on the constructed composite molecular fingerprint 70b can be determined in step 80b for the target individual. The determination of dietary recommendations can be carried out using scientific literature in order to identify from the constructed composite molecular fingerprint 70b: one or more metabolites whose abundance must be modified, and the dietary recommendations making it possible to obtain these modifications.

[0173] Alternatively, in order to determine dietary recommendations, it is possible to identify one or more individuals of interest from the set of individuals whose data have been received. The identification of these individuals of interest can be based on a criterion of richness of their intestinal microbiome, for example individuals having a microbiome richness greater than a predefined gut microbiome richness value and / or greater than the gut microbiome richness of the target individual are considered individuals of interest. The identification of these individuals of interest can also be based on an absence of metabolic syndrome or lower blood sugar, waist circumference and cholesterol levels than the target individual. Finally, dietary recommendations for the target individual can be determined to obtain the molecular fingerprint of the identified gut microbiome activity of one or more individuals of interest.

[0174] In the second variant, for each individual of the individuals maximizing a similarity with the target individual, the constructed microbiome-scale community model is used to obtain, at step 60c, a molecular fingerprint of the activity of the intestinal microbiome by simulating a monitoring of a diet determined from the data on the lifestyle of said each individual. Then, from these calculated molecular fingerprints, dietary recommendations are determined at a step 70c from the molecular fingerprint obtained at a step 60c for each individual of the individuals maximizing a similarity with the target individual. These dietary recommendations can be determined in order to improve the metabolic richness of the intestinal microbiome of each individual or in order to obtain a molecular fingerprint identified as making it possible to modify phenotypic and / or medical characteristics of said each individual.Finally, in order to determine dietary recommendations for the target individual, the dietary recommendations determined at step 70c are combined with step 80c.

[0175] In one embodiment of the invention, consistent with the preceding embodiments, step 60c may comprise generating a flux profile, specific to the target individual, of several metabolites associated with metabolic health. For example, this may relate to the following 14 metabolites: Acetate, "Acetate" in English, Butyrate L-isoleucine Indole L-leucine Propionate Trimethylamine, "Trimethylamine" in English, Tryptophan, "Tryptophan" in English, L-valine Trimethylamine-oxide, "Trimethylamine-oxide" in English, P-cresol, “P-cresol” in English, Lithocholic acid, "Lithocholic acid" in English, Deoxycholic acid, "Deoxycholic acid" in English, Indole 3-acetate, "indole-3-acetate" in English,

[0176] and / or these 10 metabolites: Folate, Octadecanoate (n-C18:0), Riboflavin "Riboflavin" in English, Uracil "Uracil" in English, Methanethiol, "Methanethiol" in English, 5 Aminopentanoate, Valeric acid or pentanoic acid, "Valeric Acid" in English, Isoprene, "Isoprene" in English, P-cresol, “P-cresol” in English, and Deoxythymidine monophosphate (C10H13N2O8P).

[0177] The fluxes of these metabolites can be calculated using a methodology developed within the framework of the invention. This methodology uses in particular the microbiota compositions derived from sequencing as well as the nutritional data of 77 meals (available in England, France, Spain, and Belgium) for all individuals. For example, these meals are low-calorie meals taken as part of a weight loss program. The meals are varied and contain in particular pasta, salads, soups, rice, meats, seafood and / or fish. For example, 5% of the meals can be Vegan: 20% of meals can be vegetarian, 20% can include fish, and the remaining meals can include meat. Additionally, approximately 50% of meals can be gluten-free.

[0178] More specifically, the methodology for calculating the flow of these metabolites developed within the framework of the invention may comprise a first step of calculating the combined Z-score for each meal and for each individual maximizing the similarity with the target individual determined in step 40. The following formula may for example be used:

[0179]

[0180] With: xi, the value of the flux of metabolite i for the set of individuals maximizing the similarity with the target individual, n x , the average concentration of metabolite i for all individuals maximizing similarity with the target individual, and o x , the standard deviation value of the metabolite concentration i for the set of individuals maximizing the similarity with the target individual.

[0181] Then, the weighted average of the Z-scores, based on Gower distances, can be calculated for all indicators to obtain a score specific to the target individual for each meal. The Z-scores can then be ranked from the highest, corresponding to the meal with the largest positive effect on the relevant metabolites, to the lowest, corresponding to the meal with the largest negative effect on all 14 relevant metabolites. The results of these steps are summarized in Figure 7, which presents a subset of the first 10 meals recommended for a target individual. Figure 7 also shows the results for individuals maximizing similarity to the target individual, called “proxies” in Figure 7.Thus, the second column titled "Z Score of the Microbiome of the target individual" indicates the values ​​calculated using the metabolic fluxes derived from the sequencing of the target individual, while the fourth column titled "Z Score of the Microbiome of the proxies" indicates the values ​​calculated using the metabolic fluxes derived from the sequencing of the individuals maximizing the similarity with the target individual, called "proxies".

[0182] In a variant of the invention, compatible with the preceding variants, the method according to the preceding variant comprises the following of dietary recommendations by the target individual. In other words, the target individual follows the determined diet 90. Under the effect of the new diet followed by the target individual, the intestinal microbiome of the individual will be progressively modified and will thus make it possible to have a positive influence on the phenotypic and / or medical characteristics of the target individual.

[0183] Unless otherwise specified, the same element appearing in different figures has a single reference. Results

[0184] The effectiveness of the method according to the invention for generating personalized nutritional recommendations at the microbiome level in the absence of a stool sample has been demonstrated. This demonstration was made for example on a cohort of 639 individuals with microbiome samples, using a catalog of 77 meals and using k-fold cross-validation to compare the recommendations generated by the method according to the invention for a target individual and the recommendations for the same individual resulting from metabolic modeling of the microbiome.

[0185] The set of individuals includes 639 individuals with the characteristics listed in the table below: W186] In addition, Figure 8 details the statistical characteristics of this example set of individuals for different numerical variables.

[0187] In this demonstration, meals were scored and ranked based on their ability to optimize a defined list of microbiome-derived, health-related metabolites, such as the 14 metabolites listed above.

[0188] Thus, it was found that the method according to the invention provides a ranking of recommended meals, for example 10 or 50 meals, with an accuracy of 76.4% when 10 meals are recommended and 99.7% when 50 meals are recommended, compared to the recommendations resulting from the metabolic modeling of the microbiome. Furthermore, for the upper quarter of meals, i.e. 19 meals, recommended for a target individual, 75% of the meals are present in a similar manner in the recommendations resulting from the method according to the invention and in the recommendations resulting from the metabolic modeling of the microbiome. This similarity is even equal to 85% and 92% for the recommendations relating to half or three-quarters of the meals, i.e. 38 and 58 meals respectively.

Claims

CLAIMS

1. A computer-implemented method of modeling an intestinal microbiome comprising the steps of: - Reception (10), for each individual of a set of individuals, of phenotypic data, medical data, lifestyle data and metagenomic data, - Reception (20), for a target individual, said target individual being an individual not included in the set of individuals, of phenotypic data, medical data, and lifestyle data, - Construction (30), for each individual of the set of individuals and for the target individual, of a vector of phenotypic data, medical data, and lifestyle data, - Determination (40), among the set of individuals, of at least one individual maximizing a similarity with the target individual, the similarity being evaluated from the constructed vectors (30) for each individual of the set of individuals and for the target individual, and - Generation (50) of at least one microbiome-scale community model based on metagenomic data of individuals maximizing similarity to the target individual, the generated microbiome-scale community model making it possible to obtain a molecular fingerprint of the activity of the intestinal microbiome by simulating an absorption of food by the intestinal microbiome, the molecular fingerprint of the activity of the intestinal microbiome comprising a vector of metabolic fluxes, said vector of metabolic fluxes comprising a distribution of metabolic fluxes across reactions of the microbiome-scale community model.

2. A method according to claim 1 wherein: - the generation (50) of at least one microbiome-scale community model comprises the steps of: - Combination (51 a) of metagenomic data of individuals maximizing similarity with the target individual in calculating an average plurality of metagenomic taxon abundances from the pluralities of metagenomic taxon abundances of the individuals maximizing a similarity to the target individual, - Obtaining (52a) at least one genome-wide metabolic model for each metagenomic taxon of the average plurality of metagenomic taxon abundances and at least one additional exchange reaction, each genome-wide metabolic model comprising exchange reactions defining environmental conditions, an individual biomass of the metagenomic taxon simulating cell growth, and biochemical reactions and transport reactions encoded by genes encoding proteins of the genome of the metagenomic taxon, - Reception (53a) of at least one additional exchange reaction, and - Construction (54a) of a microbiome community model comprising the previously obtained genome-scale metabolic models linked by at least a portion of the exchange reactions comprised in the genome-scale metabolic models, the genome-scale metabolic models exchanging molecules with a common luminal compartment shared by the genome-scale metabolic models of the community model, the community model comprising the additional exchange reactions defining the admission of molecules from the diet into the common luminal compartment and the release of the molecules from the common luminal compartment into a fecal compartment - The method (100) further comprises the steps: - Obtaining (60a) a first molecular fingerprint of the activity of the intestinal microbiome by simulating, using the microbiome-scale community model constructed (54a), tracking a first diet by the gut microbiome of the target individual, the first diet being determined from the lifestyle data of the target individual, obtaining (60a) the first molecular fingerprint comprising using a linear optimization method with maximization of community biomass as the objective function and with the diet dataset as the external constraint, the community biomass being the sum of individual biomasses weighted by the abundances of the metagenomic taxa, the metabolic flux vector comprising a distribution of metabolic fluxes across the reactions of the microbiome-scale community model, and - Obtaining (70a) at least one second molecular fingerprint of the activity of the intestinal microbiome by simulating, using the model of the constructed microbiome-scale community model (54a), a monitoring of at least one second diet by the intestinal microbiome of the target individual, the at least one second diet being different from the first diet, obtaining the second molecular fingerprint comprising the use of a linear optimization method with the maximization of the community biomass as the objective function and with the diet dataset as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the vector of metabolic fluxes comprising a distribution of the metabolic fluxes across the reactions of the microbiome-scale community model, - Identification (80a), from the phenotypic and / or medical data of the target individual and the first molecular fingerprint of the activity of the intestinal microbiome, of a molecular fingerprint of the activity of the intestinal microbiome among at least one second molecular fingerprint allowing modification of phenotypic and / or medical characteristics of the target individual, and - Determination (90a) of dietary recommendations for the target individual to obtain the molecular fingerprint of the activity of the identified intestinal microbiome (80a).

3. A method according to claim 1 wherein: - the generation (50) of at least one community model at the microbiome scale comprises the steps of, for each individual of the individuals maximizing a similarity with the target individual: - Obtaining (51 b) at least one genome-wide metabolic model for each metagenomic taxon of the plurality of metagenomic taxon abundances and at least one additional exchange reaction, each genome-wide metabolic model comprising exchange reactions defining environmental conditions, an individual biomass of the metagenomic taxon simulating cell growth, and biochemical reactions and transport reactions encoded by genes encoding proteins of the genome of the metagenomic taxon, - Reception (52b) of at least one additional exchange reaction, - Construction (53b) of a microbiome community model comprising the previously obtained genome-scale metabolic models linked by at least a portion of the exchange reactions comprised in the genome-scale metabolic models, the genome-scale metabolic models exchanging molecules with a common luminal compartment shared by the genome-scale metabolic models of the community model, the community model comprising the exchange reactions additional defining the admission of molecules from the diet into the common luminal compartment and the release of molecules from the common luminal compartment into a fecal compartment, - The method (100) further comprises the steps of: - for each individual of the individuals maximizing a similarity with the target individual, obtaining (60b) a molecular fingerprint of the activity of the intestinal microbiome by simulating, using the microbiome-scale community model constructed (53b) for said each individual, a monitoring of a diet determined from the lifestyle data of the target individual, obtaining the molecular fingerprint comprising the use of a linear optimization method with the maximization of the community biomass as the objective function and with the diet data set as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the vector of metabolic fluxes comprising a distribution of the metabolic fluxes across the reactions of the microbiome-scale community model, and - Construction (70b) of a composite molecular fingerprint by performing a weighted average of the molecular fingerprints of the activity of the intestinal microbiome obtained (60b) for each individual of the individuals maximizing a similarity with the target individual, and - Determination (80b) of dietary recommendations for the target individual based on the constructed composite molecular fingerprint (70b).

4. A method according to claim 1 wherein: - the generation (50) of at least one community model at the microbiome scale further comprising the steps of, for each individual of the individuals maximizing a similarity with the target individual: - Obtaining (51 c) at least one genome-wide metabolic model for each metagenomic taxon of the plurality of metagenomic taxon abundances and at least one additional exchange reaction, each genome-wide metabolic model comprising exchange reactions defining environmental conditions, an individual biomass of the metagenomic taxon simulating cell growth, and biochemical reactions and transport reactions encoded by genes encoding proteins of the genome of the metagenomic taxon, - Reception (52c) of at least one additional exchange reaction, - Construction (53c) of a microbiome community model comprising the previously obtained genome-scale metabolic models linked by at least a portion of the exchange reactions comprised in the genome-scale metabolic models, the genome-scale metabolic models exchanging molecules with a common luminal compartment shared by the genome-scale metabolic models of the community model, the community model comprising the additional exchange reactions defining the admission of molecules from the diet into the common luminal compartment and the release of the molecules from the common luminal compartment into a fecal compartment, - The method (100) further comprises the steps of: for each individual of the individuals maximizing a similarity with the target individual: - obtaining (60c) a molecular fingerprint of the activity of the intestinal microbiome by simulating, using the microbiome-scale community model constructed (53b) for said each individual, a monitoring of a diet determined from the lifestyle data of said each individual, obtaining the molecular fingerprint comprising the use of a linear optimization method with the maximization of the community biomass as the objective function and with the diet data set as the external constraint, the community biomass being the sum of the individual biomasses weighted by the abundances of the metagenomic taxa, the vector of metabolic fluxes comprising a distribution of the metabolic fluxes across the reactions of the microbiome-scale community model, and - Determination (70c) of dietary recommendations allowing modification of phenotypic and / or medical characteristics of said individual from the molecular fingerprint obtained (60c) for each individual of the individuals maximizing a similarity with the target individual, and - Combination of the determined dietary recommendations (70c) in order to obtain dietary recommendations for the target individual.

5. Method according to any one of the preceding claims, characterized in that the determination (40), among the set of individuals, of one or more individuals maximizing a similarity with the target individual consists of a determination of one or more individuals, among the set of individuals, minimizing a Gower distance calculated between the vector constructed for the target individual and the vector constructed for each individual of the set of individuals.

6. Method according to any one of the preceding claims, characterized in that: - phenotypic data includes at least one of: sex, age, body weight, height, body mass index (BMI), body mass index category, and waist circumference and one or more data on menopause status, - lifestyle data includes at least one of: one or more data on dietary, physical and social behavior, one or more data on smoking status, one or more data on physical activity practices, one or more data on diet and one or more socio-demographic data and one or more geographical data, - the medical data includes at least one of: one or more data on a sleep apnea status, one or more data on a diabetes status and a diabetes treatment, one or more data on a dyslipidemia status, one or more data on a dyslipidemia treatment, one or more data on a hypertension status, one or more data on a hypertension treatment, one or more data on a hypertriglyceridemia status, one or more data on a hypertriglyceridemia treatment, one or more data on a hypercholesterolemia status, one or more data on a hypercholesterolemia treatment, one or more data on an arthritis status, one or more data on an arthritis treatment, one or more data on a varicose vein disease status, one or more data on a varicose vein disease treatment, one or more data on the perception of emotional and physical well-being,one or more data on the perception of emotions and data on the perception of sleep quality, one or more data on a type of stool on a Bristol scale, and, - metagenomic data resulting from metagenomic sequencing of a fecal sample and comprising a plurality of abundances of metagenomic taxa.

7. A computer program product comprising instructions which, when the program is executed by a computer, cause the latter to implement the method according to any one of claims 1 to 7.

8. A computer-readable recording medium comprising instructions which, when executed by a computer, cause the computer to carry out the method of any one of claims 1 to 7.

9. System (200) comprising means adapted to carry out the method according to any one of claims 1 to 7.

Images