Method and system for probabilistically typing microbial strains

Abstract

The present invention provides a microbiological typing method comprising providing a database of gene profiles of microbial strains; a phylogenetic tree; assignment of gene profiles to positions within the phylogenetic tree; and the frequency of variation of profiles within the tree. Microorganism typing includes measuring the gene profile of a microorganism; determining the probability of variation of each gene profile in the database with respect to the measured gene profile, which is calculated as a function of the frequency of variation; and determining that the microorganism belongs to the taxa in the tree if the probability of variation for at least the taxa exceeds a threshold. According to the present invention, the phylogenetic tree is constructed from the core genome of the microorganism species, and a predefined set of genetic markers is selected from the accessory genome of the microorganism species.

Description

Technical Field

[0001] The present invention relates to the field of genomotyping microorganisms, i.e., characterizing bacterial strains at taxonomic levels lower than the species level based on their genomes. The present invention is particularly applicable in the epidemiological monitoring of such bacterial strains, the monitoring of contamination by such strains, as well as the investigation of the root causes of bacterial contamination and / or infection, regardless of their industrial, environmental, veterinary, clinical origin, or other origin.

Background Art

[0002] In the genomotyping of microorganisms, such as bacteria, all techniques aimed at comparing the genomes of bacterial strains are combined in order to characterize them at levels lower than the species level. Typing makes it possible, in particular, to determine the degree of similarity between two bacterial strains or to classify strains into finer subdivisions than the species level, for example, clone groups, or groups defined by phenotypic characteristics of the bacteria, such as its serotype or its susceptibility to antibiotics, etc.

[0003] Typing techniques can be classified into two categories. The first category is based on the analysis of very small target portions of the bacterial genome, such as "16S rRNA" MLST (multilocus sequence typing) typing, etc. The second category is based on the comparison of genomes obtained by whole-genome sequencing (WGS), such as cg-MLST (core genome MLST) typing, etc. The latter technique has the highest level of accuracy, but whole-genome sequencing has many drawbacks, including time, cost, and the need to perform complex bioinformatics processing. In contrast, when the first technique uses a very limited number of markers, the technique enables the use of a polymerase chain reaction (PCR) amplification platform, which is faster and less costly for the detection of such markers.

[0004] Regardless of the technology used, prior art genomic typing is based on the investigation of conserved parts of the genome and its use to characterize microorganisms. For example, MLST typing is based on housekeeping genes that are highly stable over time. Similarly, cg-MLST typing directly targets all of the core genome (or basic genome: these terms are used interchangeably). This type of approach requires a very large number of markers to achieve high accuracy. As an example, to detect clone groups of the species P. aeruginosa with an accuracy of over 90%, cg-MLST typing requires thousands of markers, which as a result renders the use of conventional PCR platforms meaningless or even impossible.

[0005] In summary, for example, users who wish to perform molecular typing to monitor, for example, microbial contamination, are forced to choose between accuracy and speed. SUMMARY OF THE INVENTION

[0006] The object of the present invention is to propose a molecular typing method using a limited number of highly accurate genetic markers.

[0007] To achieve this object, the present invention provides an in vitro method for identifying microorganisms contained in a sample, a. a computer storage means comprising a database of gene profiles of microbial strains belonging to the microbial species of said microorganism, said profiles consisting of the presence or absence of a predefined set of genetic markers, the phylogenetic tree of the species of said microorganism, the assignment of the gene profiles in the database to positions within the phylogenetic tree, and the frequency of variation of the gene profiles of the microbial strains contained in the database corresponding to predefined levels of the phylogenetic tree and b. measuring the gene profile of the microorganism. c. For at least one level of the predefined set of levels in the phylogenetic tree, The probability of variation of each gene profile in the database belonging to the level for the measured gene profile, calculated as a function of the frequency of variation at the level. Membership of microorganisms included in a sub-sub-level at a level where a first identification performance criterion is met with respect to a sub-sub-level, wherein the first criterion is met if the probability of at least one variation with respect to the sub-sub-level exceeds a first predefined threshold. Determine this using a computer-mediated method. and In the above method Phylogenetic trees are constructed from the core genomes of microscopic species. A predefined set of genetic markers is selected from the accessory genome of a microscopic species. Regarding the method.

[0008] This invention is based on the principle that the frequency of variation in accessory genomes effectively explains the diversity of microbial species when it is structured by a phylogenetic tree built on the core genome. By applying this principle, this invention consists of constructing a highly accurate and flexible genome "sensor" for this information, making it simple, fast, and reliable for the user. This sensor measures the profile of a strain being tested and estimates the probability that the strain belongs to a particular taxa in the tree as a function of the frequency of variation in the profile of that taxa.

[0009] Therefore, the present invention enables taxonomic accuracy in subdivisions of hundreds of species. For example, for Listeria monocytogenes and Salmonella, using 23 and 13 markers respectively, it is possible to obtain an assignment degree of over 80% with a 95% confidence interval in an 18-level tree. This very small number of genetic targets enables the use of amplification platforms, such as PCR platforms, such as the GENE-UP® and FilmArray platforms or isothermal amplification platforms sold by the applicant, or the use of DNA chips.

[0010] According to one embodiment, if the first identification performance criterion at that level is not met, then step c) is performed for the next level located higher in the phylogenetic tree.

[0011] According to one embodiment, the number of markers in the gene profile is less than 50, preferably 16 to 30.

[0012] According to one embodiment, The phylogenetic tree was constructed by applying clustering based on genetic distance. The genetic distance decreases from the roots to the leaves of a tree, so that levels are defined in trees containing 150 to 350 differences within the core genome.

[0013] According to one embodiment, a predefined set of markers is selected with the aim of minimizing the entropy or impurity criterion of the set at a certain level in the phylogenetic tree. In particular, markers for gene profiles are selected with the aim of minimizing the entropy or impurity criterion at a level containing 150 to 350 differences within the core genome.

[0014] According to one embodiment, the first identification performance criterion also includes the difference between the probability of variation for the sub-level and the probability of variation for other sub-levels, and the first identification performance criterion is satisfied if the difference exceeds a second predefined threshold. In particular, the second identification performance criterion also includes the difference between the probability of variation for the group and the probability of variation for other groups, and the second identification performance criterion is satisfied if the difference exceeds a fourth predefined threshold.

[0015] According to one embodiment, the database includes pairs of first and second thresholds, and at least one identification performance index for each of the thresholds, the performance index including at least one index relating to the specificity and / or sensitivity of the identification method. In particular, the identification performance index includes the degree of assignment correctness, and / or raw accuracy and / or balanced accuracy. Advantageously, the performance index for the pairs of first and second thresholds is displayed on a screen, where the user selects a numerical value for the displayed identification performance index, and hereby the method associates the first and second thresholds with their numerical values ​​corresponding to the selected numerical values ​​for the performance index.

[0016] According to one embodiment, the database includes, for at least a portion of the gene profiles, the membership of said gene profiles in different microbial communities that are not present in the phylogenetic tree, and the method is The probability of variation in the gene profiles belonging to the group relative to the measured gene profiles, which is calculated as a function of the frequency of variation at a predefined level of the phylogenetic tree, Membership of a microorganism included in one of the groups of microorganisms within the subdivision of the level, where the second identification performance criterion is met for the group, and where the second criterion is met if the probability of at least one variation relating to the group exceeds a third predefined threshold. This includes determining the result via a computer.

[0017] In particular, the second identification performance criterion also includes the difference between the probability of variation for the aforementioned group and the probability of variation for other groups, and the second identification performance criterion is satisfied if the difference exceeds the fourth predefined threshold. In particular, the microbial group is serotype or standard sequence.

[0018] According to one embodiment, the method also includes investigating the root cause in the event that a food product is contaminated.

[0019] According to one embodiment, the gene profile of a microorganism is measured by amplifying the target gene sequence, particularly via polymerization chain reactions, without performing complete sequencing.

[0020] The subject of this invention is a system for typing microorganisms contained in a sample in vitro, a. A database of gene profiles of microbial strains belonging to the microbial species of the aforementioned microorganism, wherein the profile consists of the presence or absence of a predefined set of genetic markers, The phylogenetic tree of the aforementioned microorganism species, the assignment of gene profiles in the database to their positions within the phylogenetic tree, and the frequency of variation in the gene profiles of microbial strains included in the database, corresponding to predefined levels of the phylogenetic tree. Computer storage means including, b. Means for collecting and storing measurements of the gene profiles of microorganisms, c. For at least one level of the predefined set of levels in the phylogenetic tree, The probability of variation of each gene profile in the database belonging to the level for the measured gene profile, calculated as a function of the frequency of variation at the level. Membership of microorganisms included in a sub-sub-level at a level where a first identification performance criterion is met with respect to a sub-sub-level, wherein the first criterion is met if the probability of at least one variation with respect to the sub-sub-level exceeds a first predefined threshold. Computer-mediated means for determining and It is also a system that includes this.

[0021] Such a system is configured to implement the above-described type of method.

[0022] The subject of the present invention is also a computer program product recorded on a computer-readable support, which includes instructions for performing step c of the above-described type of method.

[0023] The subject of the present invention is also a computer program product recorded on a computer-readable support, which includes instructions for constructing a phylogenetic tree as a function of the core genome and as a function of the variability frequency of a set of markers of accessory genomes at the phylogenetic tree level.

[0024] The subject of the present invention is a composition for typing microorganisms contained in a sample in vitro by amplification of a target nucleotide sequence, particularly by polymerization chain reaction, and also a composition comprising a set of oligonucleotides for amplifying at least a set of genetic markers, obtained according to the above method.

[0025] The subject of the present invention is a method for preparing a biological sample containing or tending to contain a bacterial strain for typing the strain by polymerization chain reaction, the method comprising placing the sample in contact with the composition. In particular, the method comprises dissolving the sample obtained by placing it in contact with the composition to release genetic material from the bacteria of the bacterial strain.

[0026] This invention is presented primarily as an example and will be better understood by reading the following description, which is prepared in conjunction with the accompanying drawings. [Brief explanation of the drawing]

[0027] [Figure 1] This figure shows an overview of the method according to the present invention. [Figure 2] This figure shows a flowchart detailing the learning phase of systematic typing according to the present invention. [Figure 3] This figure shows a graph illustrating the reduction in genetic distance for constructing a phylogenetic tree. [Figure 4] This figure shows a schematic diagram of a binary phylogenetic tree. [Figure 5] This figure shows a scheme illustrating the calculation of the frequency of gene profile variation according to the present invention for a taxonomic group in a phylogenetic tree. [Figure 6] This diagram shows a flowchart illustrating the manufacturing of a PCR kit used in typing according to the present invention. [Figure 7] This figure shows a flowchart detailing the phylogenetic type prediction phase according to the present invention. [Figure 8] This diagram shows a flowchart illustrating the calculation of thresholds used in the prediction phase. [Figure 9] This figure shows a flowchart illustrating the learning phase for determining the serotype according to the present invention. [Figure 10] This figure shows a flowchart illustrating the serotype prediction phase according to the present invention. [Figure 11] This diagram shows a flowchart illustrating the calculation of thresholds used in the serotype prediction phase. [Figure 12] This figure shows a graph of the decrease in Shannon entropy observed during the selection period of genetic markers that make up the gene profile for Listeria typing. [Figure 13]This figure shows a series of graphs illustrating the frequency of variations in the gene profile observed in the phylogenetic tree of the genus Listeria. [Figure 14] This figure shows a graph relating the degree of serotype assignment accuracy as a function of equilibrium accuracy in the genus Listeria. [Figure 15] This figure shows a series of graphs relating to the degree of phylogenetic assignment correctness as a function of average accuracy in the genus Listeria. [Figure 16] This figure shows a graph of the decrease in Shannon entropy observed during the selection period of genetic markers that make up the gene profile for Salmonella typing. [Figure 17] This figure shows a series of graphs illustrating the frequency of variations in the gene profile within the phylogenetic tree of the genus Salmonella. [Figure 18] This figure shows a graph relating the degree of serotype assignment accuracy as a function of equilibrium accuracy in the Salmonella genus. [Figure 19] This figure shows a series of graphs relating the degree of phylogenetic assignment accuracy as a function of average accuracy in the genus Salmonella. [Figure 20] This figure illustrates a computer architecture for performing typing according to the present invention. [Figure 21] This figure illustrates a computer architecture for performing typing according to the present invention. [Figure 22] This figure illustrates a computer architecture for performing typing according to the present invention. [Modes for carrying out the invention]

[0028] A.Definition In this document, the following definitions apply: - Biological sample: Represents any sample that contains or tends to contain one or more microorganisms; - Core genome of a microbial species (or "basic genome" represented as "Gb"): This refers to the portion of the genome (genes or groups of genes) shared by all microbial strains belonging to one or more microbial species under consideration. In light of the purposes of the present invention, the core genome can evolve if the set of strains from which the core genome is generated evolves. Preferably, the core genome is constructed from at least 100 different microbial strains, preferably at least 1000 strains. It should be noted that the genomic portions of microbial strains belonging to the core genome may differ from one another. For example, a particular gene may belong to the core genome and have several different alleles. In the following text of this specification, the term "core genome," when applied to microbial strains, refers to the portion of the strain's genome belonging to the core genome of a species. - Accessory genome of a microbial species (represented as "Ga"): A portion of the genome (genes or groups of genes) of a set of bacterial strains that does not belong to the core genome. Such a portion of the genome is therefore not present in the genome of at least one of the bacterial strains included in the set of bacterial strains under consideration. In light of the purpose of the present invention, the accessory genome may evolve if the set of strains from which the accessory genome is generated evolves. Preferably, the accessory genome is constructed from at least 100 different microbial strains, preferably at least 1000 strains. - Genetic markers: These represent genes or nucleotide sequences located at specific locations within the genome (also called "locuses"), and may correspond to a single nucleotide sequence, such as two alleles differing by only one nucleotide, or single nucleotide polymorphisms (or SNPs in the case of "single nucleotide polymorphisms") in the context of mutation. - Microorganisms: This refers to all microscopic organisms, especially bacteria, yeasts, and fungi. - Phylogenetic tree of microbial species: This refers to a tree that represents the evolutionary relationships in a set of microbial strains, determined as a function of the similarities and differences in the core genomes of the set of microbial strains. In light of the object of this invention, the phylogenetic tree can be enriched as the set of strains from which it is generated evolves. Preferably, the phylogenetic tree is constructed from at least 100 different microbial strains, preferably at least 1000 strains. By convention, the root of the phylogenetic tree is represented as the first level of the tree, and the leaves of the tree are represented as the last levels. Another level "higher" corresponds to a level closer to the root.

[0029] B. Embodiments of the present invention A probabilistic typing method according to the present invention, applicable to monitoring contamination by certain bacterial strains present in tissues, particularly in manufacturing facilities for food intended for human consumption, and for detecting their serotypes, is disclosed below in this specification.

[0030] Referring to Figure 1, the typing according to the present invention includes the following main phases: - Phase 10: Establish a database consisting of the complete genomes of strains belonging to bacterial species; - Phase 20: Known as “learning” or “training,” this involves generating a phylogenetic tree as a function of genomes in a database, selecting appropriate gene profiles, and the frequency of variation in said gene profiles along the tree; - "Prediction" Phase 30: Perform typing of unknown bacterial strains as a function of the tree and the frequencies generated in Phase 20; and - A phase optionally complemented by “epidemiological” Phase 40: During this period, corrective and / or preventive measures are taken within the organization as a function of the results of typing unknown strains.

[0031] Preferably, the complete genome database reflects the genetic diversity of the species as accurately as possible. Advantageously, if the initial version is incomplete, the database is continuously supplemented by adding complete genomes with the aim of targeting this diversity. In this case, the typing according to the present invention is performed when the database contains at least 100 genomes, preferably at least 1000 genomes, and more preferably several thousand genomes of the species.

[0032] Referring to Figures 2, 3, and 4, training phase 20 begins by generating a core genome in 202 and an accessory genome in 204 from genomes in database 200. In the first variant, the genomes include annotations that classify genes as belonging to one or the other of these genomes. In the second variant, they are generated ad hoc from the database, with genes having allele frequencies above a threshold, e.g., 95%, belonging to the core genome and others belonging to the accessory genome. In the third variant, the core and accessory genomes are generated from the database by comparing them to a reference genome of the species stored in the database, with genes having allele frequencies greater than a threshold, e.g., 95%, belonging to the core genome and others belonging to the accessory genome. In the fourth variant, a database has already been established for the species, and similar versions of its core and accessory genomes are also established.

[0033] The typing according to the present invention continues to generate a phylogenetic tree as a function of the core genome.206 Several techniques are possible for constructing such a tree and are described, for example, in the documents “Phylogenetics Algorithms and Application” (Hu et al., Ambient Communications and Computer Systems, 2019), “A Review: Phylogeny Construction Methods” (Shaktawat et al., International Journal of Emerging Science and Engineering, 2019), or “Essential Bioinformatics” (Chapter 11 by Jin Xiong, Cambridge University, 2011).

[0034] In a modified version of the present invention, the tree is constructed from cg-MLST (core genome polylocus sequencing typing) coding of the core genome. This core genome is converted to a cg-MLST profile using, for example, BioNumerics software (bioMerieux, Marcy l'Etoile, France), SeqSphere+ (Ridom, Munster, Germany), BIGSdb (accessible at https: / / github.com / kjolley / BIGSdb, and described in the document "Whole genome sequencing options for bacterial strain typing and epidemiologic analysis based on single nucleotide polymorphism versus gene-by-gene-based approaches" by Schurch, A. et al., Clin. Microbiol. Infect., 2018), or chewBBACA (accessible at https: / / github.com / B-UMMI / chewBBACA, and described in the document "chewBBACA: a complete suite for gene-by-gene scheme creation and strain identification." by Silva M et al., Microb. Genom., 2018). One advantage of the cg-MLST profile is that it is a fixed-length array consisting of hundreds of components that are aligned with each other, allowing for the direct use of tree construction methods.

[0035] This tree is favorably generated by applying a series of groupings using single-linkage clustering, and is therefore based on the genetic distance between core genomes, calculated, for example, using normalized Hamming distance. Next, the subdivision of taxa within the tree consists of integrating core genomes that differ from one another by a distance smaller than a predefined threshold into the same subgroup. Software such as HierCC (accessible at https: / / github.com / zheminzhou / pHierCC and described in the document "HierCC: a multi-level clustering scheme for population assignments based on core genome MLST" by Zhou et al., Bioinformatics, 2021) and Pathogenwatch from the Center for Genomic Pathogen Surveillance (accessible at https: / / pathogen.watch / ) can be used. Once the tree is constructed, it is stored in database 208.

[0036] In a particularly preferred variant of the present invention, the number of levels and clustering distances within the tree are selected with the aim of generating at least one level within the tree such that the core genome diversity at that level corresponds to or approximates the clonal diversity observed in the species, thereby enabling the taxa at that level to distinguish different clonal complexes from bacterial strains corresponding to complete genomes in the database. For example, in the genera Salmonella and Listeria, the levels have 150 to 350 differences, preferably 200 to 300 differences, in the core genome of the strains. In particular, the clustering distance decreases linearly, for example, from the roots toward the leaves.

[0037] Learning Phase 20 continues with the selection of the type of gene profile within the accessory genome, i.e., the selection of a set of genetic markers (210) derived from the accessory genome of the species whose presence or absence will be measured in the strains to be tested during Prediction Phase 30. Profile selection consists of selecting one of the levels within the tree, preferably a level representing clonal diversity, and then selecting a set of markers having the lowest entropy or impurity at that level. The total number of markers selected to construct the gene profile is N. M This is favorably selected as a function of the specific PCR platform used to perform the typing. For example, in the case of the applicant's GENE-UP platform, this number N M This is equivalent to 23 markers for the genus Salmonella and 16 markers for Listeria monocytogenes.

[0038] In practice, the set of markers is selected using a greedy algorithm with the aim of minimizing an entropy criterion (e.g., Renyi entropy, Hartley entropy, Shannon entropy, collisional entropy, or minimum entropy) or a Gini impurity criterion.

[0039] For illustrative purposes, and referring to the phylogenetic tree in Figure 4, Q N = 20 classification groups (T1, T2, ..., T 20 The selection of a set of markers to minimize the Shannon entropy criterion of level N, consisting of the entropy criterion: This involves selecting the first marker M1 derived from the accessory genome that minimizes TIFF2026519548000002.tif14170, where, The file is TIFF2026519548000003.tif7170, where NG is the total number of core genomes used to construct the phylogenetic tree, and n i The marker is the i-th taxonomic group T. iis the number of occurrences within the accessory genome in. When marker M1 is selected, this procedure is repeated to select a second marker M2 within the accessory genome removed from marker M1. Thus, the j-th marker M j is a marker represented as TIFF2026519548000004.tif14170, where TIFF2026519548000005.tif7170, where NG is the total number of core genomes used to construct the phylogenetic tree, and n i is the number of times marker M j is present within the accessory genome in the i-th taxon T i .

[0040] As a variant form, stepwise marker selection is performed by minimizing the conditional entropy at level N according to the following equation: TIFF2026519548000006.tif15170, where - TIFF2026519548000007.tif7170, where NG is the total number of core genomes used to construct the phylogenetic tree, and n i is the number of times the marker is present within the accessory genome in the i-th taxon T i . - TIFF2026519548000008.tif8170, where n i is the l-th value of the marker set {M1, M2,..., M j-1 , M j} (the set may contain 2 j values, among which) is the number of times it is present in the i-th taxon T i , and N i is the number of core genomes contained in taxon T i .

[0041] The inventors have observed that the typing method according to the present invention requires 50 markers or less, preferably fewer than 30 markers, and more specifically 16 to 30 markers, to achieve high typing performance.

[0042] Learning phase 20 continues with the extraction (212) of accessory genomes from the gene profiles formed from selected markers (stored in database 214). Thus, the gene profiles of accessory genomes are The vector P in TIFF2026519548000009.tif5170 is such that the j-th component P(j) of vector P is the marker M j Corresponding to the value, this value is marker M j The value is equal to 0 if it does not exist, and equal to 1 if it does exist.

[0043] Once the profile database 214 is formed, the frequency of profile variations along the tree is calculated in 216. Referring to Figure 5, the subdivisions (T1, T2, ..., T Q For each taxonomic group T of the tree, which is divided into subgroups (T1, T2, ..., T Q ) and a subset of the profiles associated with each. Set of gene profiles associated with taxa T, categorized in TIFF2026519548000010.tif8170 {P1,P2,...,P T} exists. Gene profile marker M j For each sub-taxa (T1, T2, ..., T), the frequency of variation P(Mj|T) of this marker for each taxonomic group T is divided by Q. Q It is calculated as being equal to the number of ), in which case marker M j This includes both the numerical value 0 and the numerical value 1, i.e., the relationship: The frequencies are taken based on TIFF2026519548000011.tif13170, where card(E) represents the cardinality function of set E. Therefore, in the frequency calculation, for each subdivided taxa T, The frequency vector P(.|T) of TIFF2026519548000012.tif5170 is obtained, where P(.|T)(j) = P(M j |T) is calculated, and the frequency vector is stored in database 218.

[0044] The learning phase 20 concludes with the calculation of a grid of hyperparameters (220) to be used in the prediction phase 30 (this calculation will be described in detail once the prediction phase 30 is described), and the parameters are stored in the database 222.

[0045] In summary, in learning phase 20, the gene profiles extracted from each complete genome and its accessory genomes are placed within a tree, resulting in a phylogenetic tree in which each taxa within the tree, excluding the leaves, is associated with the frequency of variation of each genetic marker in the profile.

[0046] Referring to Figure 6, the present invention relates to the design and manufacture of a composition or “kit” for carrying out amplification for the purpose of detecting a selected marker, - Selection of amplification primers for amplifying selected genetic markers (500): - Manufacturing of the primers, and, if necessary, manufacturing of a detection probe intended to detect the presence of the marker in at least one of the marker variants, preferably in the maximum number of such variants (502); - Selected primers, detection probe (if necessary), dNTPs (deoxyribonucleotide triphosphates) to provide energy, nucleotide bases required for amplicon generation, amplification enzymes, and appropriate concentrations of salts, e.g., Mg, to enable the enzyme to function properly. 2+ Preparation of a buffer solution containing ions or NaCl, etc. (504). Dyes may also be added to this solution. - Preparation of lysis buffers containing beads, such as magnetic beads or ceramic beads, and optionally dyes. Including design and manufacturing 50.

[0047] The manufactured kit is then used in prediction phase 30. Referring to Figure 7, this phase begins with the collection of a biological sample 300 obtained from a tissue, for example, a food product manufacturing facility. In the subsequent step 302, the sample is prepared with the aim of detecting a predetermined pathogen using methods known in the prior art, such as culture on a plate, VIDAS® type immunological detection, or molecular biology using, for example, a GENE-UP® platform. After the detection of the pathogen, a confirmation step may be performed.

[0048] Simultaneously, if the colony is identified as belonging to a species, the prediction phase 30 continues to 302, which uses the isolated colony or enriched broth for characterization by PCR amplification. In particular, the sample undergoes a series of preliminary steps, such as lysis to release DNA from the bacteria constituting the sample, optionally using beads, such as magnetic beads or ceramic beads, or DNA extraction from the lysate. Once the bacterial strain's DNA is released, it is added in 304 to a PCR kit designed to target at least one genetic marker selected in the learning step 20. Finally, PCR is performed in 306, with the aim of measuring the strain's genetic profile, for example, using the applicant's GENE-UP® platform.

[0049] Measured profile (P mes The bacterial strains (represented as ) are then analyzed in 308 to place them within a phylogenetic tree. Starting from tree level N, for example, the level just above the leaf level, analysis 308 includes the following steps: a. Classification group at level N (T i,N For each (represented as the i-th taxa at the Nth level): a1. Taxon T i,N The stored profile (P) in database 214 that was assigned to obs (represented as) for each, the measured profile P mesProfile P obs The probability P is the variation. mut (P mes ,P obs ) is related by: Calculate according to TIFF2026519548000013.tif15170. Therefore, the measured profile P mes is profile P obs If it is identical to [another value], this probability is equal to 1 (no change), and profile P mes Profile P obs If it deviates from the norm, it decreases rapidly, and the frequency P mut (M j |T i,N Note that the decrease is even greater when the value is low. a2. Taxon T i,N The calculated probability P mut (P mes ,P obs ) maximum Select TIFF2026519548000014.tif5170. In b.310, the measured profile P mes of, As shown in TIFF2026519548000015.tif17170 (wherein S1 and S2 are predetermined positive thresholds excluding zero), a classification group T of level N. j,N Assigning to one of them; c. If the taxonomic group at level N does not meet the criteria of process b, move to a higher level N-1 in the tree and perform processes a and b for this new level.

[0050] Profile P mes Once assigned to a level in the phylogenetic tree and a taxonomic group at that level, prediction phase 30 delivers a report to the user, for example, in the form of a screen display and / or a file stored on a computer and / or a message to the user (e.g., email). If, upon completion of step b, two taxonomic groups have the same result, the report includes both taxonomic groups or is listed as undecidable typing.

[0051] The position of a newly tested bacterial strain in the phylogenetic tree allows for the typing of the new strain relative to the bacterial strain corresponding to its complete genome in database 200. In addition to this type of typing, which references strains observed to date and incorporated into various databases, the typing according to the present invention allows for the relative typing of two bacterial strains when those two strains are assigned to the same taxonomic group. Furthermore, the present invention allows for the definition of a lineage for one strain relative to a strain in the database or to another tested strain. Specifically, the relative positions of two strains in the tree allow for the characterization of their relative evolution.

[0052] The phylogenetic tree is traced from the leaves to the roots, and unless the assignment relationships are verified, the phylogenetic tree is traced upwards. In a variant form, the phylogenetic tree is traced from the roots to the leaves, and unless the assignment relationships are verified, the phylogenetic tree is traced downwards.

[0053] The threshold values ​​S1 and S2 are hyperparameters that are preferably determined by cross-validation in step 220 of the learning phase 30. Referring to Figure 8, a grid of values ​​for thresholds S1 and S2 is first determined, and for each pair of thresholds in the grid, the degree of assignment correctness and the average accuracy are determined as follows: - The complete genome database 200 is divided into a training database 800 and a test database 802 according to a cross-validation strategy 804, for example, a strategy called "Tenfold cross-validation"; - Perform training phase 20 on training database 802; - Prediction phase 30 is performed for each gene profile derived from testbase 802, and the mean accuracy ACC and assignment correctness TA are calculated in 806 for the gene profiles derived from database 802 within the phylogenetic tree; - At 808, the maximum values ​​for average accuracy and assignment correctness are selected across the folds of the cross-validation strategy. Thus, these two values ​​correspond to the average accuracy and assignment correctness of the phylogenetic prediction phase 30 for thresholds S1 and S2, respectively. The pair of thresholds with the best accuracy and degree is selected for phylogenetic typing according to the present invention. In one variant, the user can select a desired accuracy and / or assignment degree, and phylogenetic typing selects a pair of thresholds (whose accuracy and degree are closest to those selected by the user).

[0054] In the preferred method, the number of levels in the phylogenetic tree and the distance decay law of clustering (e.g., affine) are also hyperparameters, and therefore the method described in relation to Figure 8 is complemented by the selection of grids for these two parameters, as well as the combination of grid paths and grid paths for thresholds S1 and S2.

[0055] Herein, a preferred variant of the invention is described, comprising predicting the membership of bacterial strains within a specific group of bacterial species as a function of the measured genetic profiles of the bacterial strains. In the example shown, this group is serotype, but this variant can be applied to any explicit subdivision at some level of bacterial species. It should be noted that such subdivisions of species are defined independently of, but rather do not correspond to, the first level of the phylogenetic tree.

[0056] Referring to Figure 9, the learning phase 20 described above in relation to Figure 2 is complemented by serotype learning 20A, which includes the following: a. Collect serotypes from all or part of the bacterial strains corresponding to the complete genome. The serotypes will be stored in database 224; b. NS (ST1, ST2, ...ST) NS) to distribute the gene profiles observed in the database 214 of serotyped strains (226). Gene profiles associated with a serotype will be referred to as “serotyped” below in this specification; c. Calculate the frequency of variation in the gene profile for which serotyping has been determined (228). In particular, marker M j Each time, serotype ST i The frequency of that variation P(M) j |ST i ) is serotype ST i The number of times this marker takes the value 1 is equal to the total number of gene profiles assigned to it; d. For the gene profile in which the serotype has been determined among different serotypes, the frequency of variation P(M) j |ST i ) should be stored in database 230.

[0057] If the elements of learning phase 20 described in relation to Figure 2 remain unchanged, in particular meaning that the type of gene profile used is determined as a function of the phylogenetic tree rather than as a function of serotype subdivisions, then prediction phase 30 can take two forms. In the first form, the already described phase 30 is complemented by serotype prediction of the novel strain being tested, in parallel with steps 308-312 in Figure 2. In the second form, only the prediction is performed, replacing steps 308-312.

[0058] Referring to Figure 10, the prediction of the serotype of a new bacterial strain 30A is performed according to steps 300-306, which involve measuring the genetic profile of the bacterial strain of the species contained in the biological sample, and includes the following steps: a. Each serotype (ST i Regarding (expressed as): a1. Group ST of database 228 i For each gene profile in which the serotype was determined (P ST (represented as), measured profile P mes Profile P STThe probability P is the variation. mut (P mes ,P ST ) is related by: Calculate according to TIFF2026519548000016.tif15170; a2. Serotype ST i The calculated probability P mut (P mes ,P obs ) maximum Select TIFF2026519548000017.tif5170; b. Measured profile P mes of, Serotype ST, as in TIFF2026519548000018.tif19170 (wherein S3 and S4 are pre-determined positive thresholds excluding zero). j Assigning to one of them; c. Prepare and send a report in 318 regarding the serotypes detected for bacterial strains that have already undergone serotyping.

[0059] In the variant form of serotype determination 30A, the second condition If TIFF2026519548000019.tif7170 is not met, the serotype determination is, with the highest probability In addition to the first two or three serotypes associated with TIFF2026519548000020.tif5170, the aforementioned probabilities are also returned.

[0060] The numerical values ​​of thresholds S3 and S4 are hyperparameters, preferably determined by cross-validation in step 220 of the learning phase 30. Referring to Figure 11, a grid of numerical values ​​for thresholds S3 and S4 is first determined, and for each pair of thresholds in the grid, the degree of correctness and average accuracy of the assignment for serotype typing are determined as follows: - The whole genome database 200 and serotype 224 are divided into training database 1100 and test database 1102 according to a cross-validation strategy 1104, for example, a strategy called "Tenfold cross-validation"; - Perform training phases 20-201 on training database 1102; - Prediction phase 30A is performed for each gene profile derived from test database 1102, and the mean accuracy ACC and assignment correctness TA are calculated in 1106 for the gene profiles derived from database 802 within the serotype; - At 1108, the maximum values ​​for average accuracy and assignment correctness are selected across the folds of the cross-validation strategy. Thus, these two values ​​correspond to the average accuracy and assignment correctness of serotype prediction phase 30A for thresholds S3 and S4, respectively. The pair of thresholds having the best accuracy and degree is selected for serotype typing according to the present invention. In one variant, the user can select a desired accuracy and / or assignment degree, and the serotype determination according to the present invention selects a pair of thresholds (whose accuracy and degree are closest to those selected by the user).

[0061] In the variant described above, the determination of the serotype of the strain being tested involves the calculations described in relation to Figures 9 and 10. In another variant, for each taxa in the phylogenetic tree, the probability of each serotype in that taxa is calculated as a function of the serotype database 224. For example, taxa T i The probability of a given serotype within ST jThis probability is equal to the number of genomes grouped within this taxa that possess this serotype divided by the total number of genomes in the taxa. This probability is stored in a reference table or dictionary, and the prediction of the serotype of the strain to be tested consists of first determining the taxa of the phylogenetic tree to which the strain belongs, in the manner described in steps 308-312. Once the taxa is identified, the determination of the strain's serotype consists of querying the probabilities in the reference table associated with the taxa and providing the user with at least the serotype with the highest probability.

[0062] In the modified forms of typing and serotyping described above, once the profile is measured, the numerical value P mut , Calculation and measured profile P for TIFF2026519548000021.tif5170 mes The assignment is performed. In a variant form, known profiles are directly associated with the taxa (one or more) of the binary phylogenetic tree, for example, using a reference table. Specifically, the taxa are known because the assignment calculation has been performed in advance. In particular, profile P obs It is assigned to its associated classification group by the calculations described above. Therefore, the profile of the new stock is Profile P obs If it is identical, the reference table is called for this profile and the associated taxa are output directly. Similarly, a new profile P mes If a value is measured and it differs from all profiles in the reference table, its assignment is calculated as previously described, and this new profile is then stored in the reference table along with the identified taxa(s). This avoids the calculation of assignments that have already been performed by simply calling the reference table. As a variation, instead of a reference table, each known profile is stored in a memory space dedicated to the taxa(s) of the phylogenetic tree. In other words, taxa in the tree are assigned an "address" that is equivalent to a known and assigned profile. [Examples]

[0063] D. Examples D.1. Listeria monocytogenes 37,000 Listeria strains were collected, fully sequenced, and serotyped. The number of levels in the tree is equal to 23, which is the total number of targets on the applicant's GENE-UP® PCR platform. A distance decay rule is selected for clustering so that diversity similar to the clonal diversity of 10–15 levels is obtained in a phylogenetic tree containing 24 levels. Figure 12 shows that the Shannon entropy of the gene profile decreases as markers are selected from accessory genomes at level 13 of the phylogenetic tree, and the entropy of the final profile is only a few percent away from the free entropy at that level. Figure 13 shows the frequency of profile variation at different levels of the tree, where the curves in the graph correspond to the taxa at that level, and the thick curves correspond to the frequency of the displayed markers. As can be seen from Figure 14, which shows serotype typing for 12 known serotypes of the genus Listeria, when the degree of serotype assignment accuracy is 100%, the mean equilibrium accuracy is equal to 90%, demonstrating the effectiveness of this typing according to the present invention. Note that it is possible to select a higher accuracy for satisfactory results even with a lower degree of assignment accuracy. In particular, at a degree of 80%, the accuracy is approximately 95% in the 95% confidence interval, and is fully encompassed in the last decile. Figure 15 shows the mean accuracy of the level as a function of the degree of phylogenetic assignment accuracy. Again, the excellent performance of the present invention is noted, with these two values ​​exceeding 95% and 70% respectively for all levels above level 15, and close to 100% for levels above level 9.

[0064] D.2. Salmonella genus 39,000 Salmonella strains were collected, fully sequenced, and serotypes were determined. The number of levels in the tree is equal to 23, which is the total number of targets on the applicant's GENE-UP® PCR platform, and a distance decay rule is selected for clustering so that diversity similar to the clonal diversity of 10-15 levels is obtained in a phylogenetic tree containing 24 levels. Figure 16 shows that the Shannon entropy of the gene profile decreases as markers are selected from accessory genomes at level 13, and the entropy of the final profile is virtually zero. Figure 17 shows the frequency of profile variation at different levels of the tree, where the curves in the graph correspond to the taxa at that level, and the thick curves correspond to the average frequency at that level. As can be seen from Figure 18, which shows serotype determination for more than 2,600 known serotypes of Salmonella, the mean equilibrium accuracy is over 90% when the degree of serotype assignment accuracy is 100%, demonstrating the effectiveness of this typing according to the present invention. It should be noted that even with a lower degree of assignment, it is possible to select a higher accuracy for satisfactory results. In particular, at a degree of 80%, the accuracy is nearly 100% in the 95% confidence interval and is fully encompassed in the last decile. Figure 19 shows the average accuracy of the level as a function of the degree of correctness of the systematic assignment. Again, the excellent performance of the present invention is noted, with these two values ​​exceeding 95% and 80%, respectively, for all levels above level 13.

[0065] E. Preventive and corrective measures The solution's ability to easily and rapidly identify strains allows for consideration of its use in investigating root causes when contamination of final food products is identified during routine check periods. This includes, in particular, investigating strains within the relevant factory environment by collecting surface samples using, for example, swabs, sponges, or cloths. Strains can also be investigated in the starting materials used in the manufacture of the relevant final product to link the contamination to one or more suppliers and / or carriers, and / or storage locations. The solution may also enable the identification of contamination during the laboratory detection phase by laboratory strains or their DNA that may have distorted release tests that identified the presence of the species in question. In this case, the solution allows for the development of plans to retest the final product and avoid its disposal. Finally, the solution can be used during health crises to rapidly characterize the causative strain(s) and link them to strains identified in the food, cosmetic, veterinary drug, or other product in question. Its usefulness therefore covers, for example, internal needs in food product manufacturing facilities, investigation needs during health crises, or on-site audits by regulatory authorities. Thus, the use of this solution makes it possible to streamline and implement appropriate preventive and / or corrective measures to reduce or even eliminate the risk of recurrence of contamination and, consequently, infection.

[0066] F. Implementation of the present invention on computer hardware The learning phases 20, 20A, and prediction phases 30-30A are performed by a computer, i.e., by hardware circuitry including computer memory and one or more microprocessors or processors (whether organized in the form of computing nodes) necessary to execute computer instructions stored in memory (cache, RAM, ROM, etc.) to perform the phases, with the exception of the steps of preparing the sample and measuring the genetic profile of the strain to be tested. Several architectures are possible, for example, as shown in Figures 20-22. In a first architectural variant (Figure 20), a first organization 2000, e.g., the applicant, hosts or controls one or more computing servers 2002 associated with one or more databases 2002 for storing genomes, phylogenetic trees, phylogenetic frequencies, observed genetic profiles, serotypes, and serotype frequencies, and the learning phases 20, 20A are performed by the organization 2000 on its server(s) 2002. A second organization 2006, for example, a microbiology laboratory, hosts a PCR platform 2008 connected to or integrated into computer 2010 and performs all steps up to part of prediction phase 30, namely the measurement of the genetic profile of the bacterial strain being tested. The measured profile is then sent to the first organization 2000 via a remote connection network to perform the remainder of prediction phase 30, 30A on server 2002. The generated report is then sent to the second organization 2006 via network 2012, where the second organization 2006 decides whether to take epidemiological action based on the report. The second architectural variant (Figure 21) differs from the first variant in that a copy of database 2004 and a copy of the software for performing prediction phase 30, 30A are downloaded to the second organization, and the second organization performs all of prediction phase 30, 30A using computer 2010 or a computing server (not shown).In the third variant (Figure 22), all learning phases 20, 20A and prediction phases 30, 30A are performed by a single organization 2006.

[0067] G. Extension of the Instructions of the Embodiments One embodiment of the present invention has described the use of PCR to measure the genetic profile of a bacterial strain. In a variant form, this profile is measured using a DNA chip that targets the marker in a manner known in itself.

[0068] Complete genome databases, at least in their initial versions, have described embodiments that are independent of the bacterial ecology of the tissues performing the typing for their own needs. In a variant form, this database consists substantially of bacterial strains that originate solely in tissues. However, it is also possible to provide a minimal initial version of the database to an tissue and then supplement it with strains derived from that tissue.

[0069] Embodiments applicable to bacterial species have been described. The present invention is also applicable to yeasts and fungi.

[0070] We have described methods for serotyping. Other subtypes of a species can also be characterized, for example, based on standard sequences, for resistance and / or susceptibility to microbial factors or groups.

[0071] In addition to the degree of assignment accuracy, equilibrium accuracy or non-equilibrium accuracy have also been described as criteria for selecting prediction hyperparameters. Other criteria, such as sensitivity and specificity, or confidence intervals different from 95%, are also possible.

Claims

1. An in vitro method for typing microorganisms contained in a sample, a. To provide a computer storage means, wherein the computer storage means is A database of gene profiles of microbial strains belonging to the microbial species of the aforementioned microorganism, wherein the profile consists of the presence or absence of a predefined set of genetic markers, The phylogenetic tree of the microorganism species, the assignment of the gene profiles in the database to their positions within the phylogenetic tree, and the frequency of variation in the gene profiles of the microbial strains included in the database, corresponding to predefined levels of the phylogenetic tree. To provide computer storage means including, b. Measuring the gene profile of the microorganism, c. For at least one level of the predefined set of levels of the phylogenetic tree, The probability of variation of each gene profile in the database belonging to the level for the measured gene profile, the probability of variation calculated as a function of the frequency of the variation at the level, Membership of the microorganisms included in the sub-sub To determine this via computer Includes, In the above method, The aforementioned phylogenetic tree is constructed from the core genomes of the aforementioned microorganisms, The aforementioned predefined set of genetic markers is selected from the accessory genome of the microorganism. method.

2. The method according to claim 1, wherein if the first identification performance criterion at the level is not met, then step c) is performed for the level located directly above in the phylogenetic tree.

3. The method according to claim 1 or 2, wherein the number of markers in the gene profile is less than 50, preferably 16 to 30.

4. The aforementioned phylogenetic tree is constructed by applying clustering based on genetic distance. The genetic distance decreases from the roots to the leaves of the tree, such that a level is defined in the tree containing 150 to 350 differences within the core genome. The method according to any one of claims 1 to 3.

5. The method according to any one of claims 1 to 4, wherein the set of predefined markers is selected with the aim of minimizing the entropy or impurity criterion of the set at a certain level of the phylogenetic tree.

6. The method according to claim 4 or 5, wherein the markers of the gene profile are selected with the aim of minimizing the entropy or impurity criterion at a level that includes 150 to 350 differences within the core genome.

7. The method according to any one of claims 1 to 6, wherein the first identification performance criterion also includes the difference between the probability of variation for the sub-sub-level and the probability of variation for other sub-sub

8. The method according to claim 7, wherein the second identification performance criterion also includes the difference between the probability of variation for a group and the probability of variation for other groups, and the second identification performance criterion is satisfied if the difference exceeds a fourth predefined threshold.

9. The method according to any one of claims 1 to 8, wherein the database includes a pair of first thresholds and second thresholds, and at least one identification performance index for each of the thresholds, the performance index including at least one index related to the specificity and / or sensitivity of the identification method.

10. The method according to claim 9, wherein the identification performance index includes the degree of assignment correctness, and / or raw accuracy and / or balanced accuracy.

11. The method according to claim 9 or 10, wherein the performance indicators for the pair of first and second thresholds are displayed on a screen, the user selects a numerical value for the displayed identified performance indicator, and the method sets the first and second thresholds to the numerical values ​​corresponding to the selected performance indicator.

12. The database includes, for at least a portion of the gene profiles, membership of the gene profiles in different microbial communities that are not present in the phylogenetic tree, and the method is The probability of variation in the gene profiles belonging to the group relative to the measured gene profiles, wherein the probability of variation is calculated as a function of the frequency of the variation at a predefined level of the phylogenetic tree, Membership of a microorganism that is included in one of the groups of microorganisms within the subdivision of the level, where the second identification performance criterion is met for the group, and where the second criterion is met if the probability of at least one variation relating to the group exceeds a third predefined threshold. The method according to any one of claims 1 to 11, comprising determining by computer.

13. The method according to claim 12, wherein the second identification performance criterion also includes the difference between the probability of variation for the group and the probability of variation for other groups, and the second identification performance criterion is satisfied if the difference exceeds a fourth predefined threshold.

14. The method according to claim 12 or 13, wherein the microbial group is a serotype.

15. The method according to any one of claims 1 to 14, comprising an investigation into the root cause of contamination of a food product during its period of contamination.

16. The method according to any one of claims 1 to 15, wherein the measurement of the gene profile of the microorganism is carried out by amplifying the target gene sequence, particularly via a polymerization chain reaction, without performing complete sequencing.

17. A system for typing microorganisms contained in a sample in vitro. d. Computer storage means, A database of gene profiles of microbial strains belonging to the microbial species of the aforementioned microorganism, wherein the profile consists of the presence or absence of a predefined set of genetic markers, The phylogenetic tree of the microorganism species, the assignment of the gene profiles in the database to their positions within the phylogenetic tree, and the frequency of variation in the gene profiles of the microbial strains included in the database, corresponding to predefined levels of the phylogenetic tree. Computer storage means including, e. Means for collecting and storing measurements of the gene profile of the microorganism, f. For at least one level of the predefined set of levels of the phylogenetic tree, The probability of variation of each gene profile in the database belonging to the level, with respect to the measured gene profile, the probability of variation calculated as a function of the frequency of variation at the level. Membership of the microorganisms included in the sub-sub Computer-mediated means for determining and A system that includes this.

18. The system according to claim 16, configured to carry out the method described in any one of claims 2 to 16.

19. A computer program product recorded on a computer-readable support, comprising instructions for performing step c of the method according to any one of claims 1 to 16.

20. A computer program product stored on a computer-readable support, which includes instructions for constructing the phylogenetic tree as a function of the core genome and as a function of the frequency of variation in the set of markers of the accessory genome at the phylogenetic tree level.

21. A composition for typing microorganisms contained in a sample in vitro by amplification of a target nucleotide sequence, particularly by polymerization chain reaction, comprising a set of oligonucleotides for amplifying at least a set of genetic markers, obtained according to claim 5 or 6.

22. A method for preparing a biological sample containing or likely to contain a bacterial strain for the purpose of typing the strain by polymerization chain reaction, the method comprising placing the sample in contact with the composition according to claim 21.

23. The method according to claim 22, comprising dissolving the sample resulting from placing it in the aforementioned contact state, thereby releasing genetic material from the bacteria of the bacterial strain.

Images