Annotation of the results obtained in 2021
We completed development of algorithms and modules serving as a part of the system for automated analysis of metagenomic samples in the format of amplicon sequencing of the 16S rRNA gene that allow one to calculate taxonomic and metabolic characteristics of the samples. The bioinformatic pipeline output taxonomic profiles at different taxonomic levels, containing tables of the relative abundance for detected taxonomic units at the level of species, genus, family, order, class and type of bacteria. Metabolic profiles are provided in the form of tables of cumulative phenotypic indices (CPI) for each studied metabolite. The following pipeline components were developed and implemented: 1) a module for calculating the cumulative occurrence of polysaccharide catabolic enzymes according to the classification from the CAZy database; 2) a module for calculating the resistome potential of a metagenomic sample for known families of drug resistance proteins belonging to 13 functional categories from the Resfams database; 3) a module for identifying taxonomic groups that create the greatest quantitative contribution to the metabolic CPI of a metagenomic sample; the module allows predicting directional changes in the composition of the microbiota by adding certain prebiotics and vitamins; 4) a module for the search for discriminatory metabolic phenotypes between groups of metagenomic samples, determined using various metadata, and allows one to highlight the phenotypes that best explain the differences between groups of samples. The developed software modules were applied for phenotypic profiling of several large datasets of 16S metagenomic samples of the intestinal microbiota. The phenotypic profiling software pipeline was used to analyze 16S metagenomic samples from 10 different in vitro fermentation studies of fecal microbiota. The obtained profiles were analyzed in the context of the metadata available for the samples (concentration and type of prebiotic, SCFA, or other metabolites added to the growth medium), statistically processed, and phenotypes potentially associated with the properties of experimental prebiotics / metabolites were identified. Also, SCFA production phenotypes were calculated for intestinal samples from a metagenomic study of 1-3 years old infants with a genetic predisposition to diabetes mellitus. We completed the development and testing two bioinformatic pipelines for taxonomic and functional analysis of metagenomic samples in the whole genome sequencing (WGS) format. The first pipeline allows obtaining taxonomic and CPI profiles for WGS samples using a reference database of 2856 intestinal genomes for which metabolic reconstructions and phenotypes are available in the form of a Binary Phenotypic Matrix (BPM). The second pipeline makes it possible to assess the representation of individual metabolic genes (signature genes) and biochemical pathways using metabolic reconstructions from the reference database. The pipelines have been used to profile WGS samples from several metagenomic studies allowing to find links between taxa, biochemical pathways and metabolic phenotypes represented in the samples. In particular, we identified metabolic phenotypes and distinct biochemical pathways that are potentially associated with the Crohn's disease. We completed the development and testing of an integrated bioinformatics pipeline for automatic extension of the reference BPM to new bacterial genomes and metagenomic-assembled genomes (MAGs). The pipeline allow one to functionally annotate a target genome or MAG and to assign its metabolic binary phenotypes using three computational methods, namely, phenotype rules, machine learning models and determining the reference phenotype in groups of closely related genomes; with the final assignment of a consensus phenotype. We carried out large-scale testing of this pipeline on a set of >20,000 genomes from three genomic collections of isolates and MAGs, which showed the accuracy of determining the phenotypes of more than 99%. We analysed the obtained extended BPM for human gut genomes, and calculated the metrics of variability of metabolic phenotypes in groups of closely related genomes belonging to the same species. The results allow to find the most variable taxonomic species and metabolic phenotypes in an expanded collection of intestinal genomes and MAGs. With the use of comparative genomic analysis and metabolic reconstruction, we further obtained an extended BPM for ~2,000 reference bacterial genomes representing the human vaginal microbiome. The updated BPM allows one to determine the distribution of metabolic potentials in the vaginal microbiota of healthy women, and to associate them with the main types of vaginal microbiomes previously identified using taxonomic profiles. We performed metabolic reconstruction and functional annotation of the pathways and regulons for carbohydrate utilization pathways in intestinal bacteria belonging to the genus Collinsella. For comparative analysis, 136 Collinsella genomes were selected, which were divided into 18 groups of closely-related strains. As a result, we inferred metabolic phenotypes for catabolism of sugars and oligosaccharides in Collinsella spp., that include both conserved and strain-specific phenotypes. Among the conserved phenotypes in Collinsella, we found the utilization of fructose, mannose, N-acetylgalactosamine, maltose, and sialic acid. New potential carbohydrate transporters and sugar-specific transcription factors were discovered, for which DNA-binding sites were identified and the corresponding regulons were reconstructed.

 

Publications

---

Annotation of the results obtained in 2019
In 2019, the following unique scientific results were obtained. A metabolic reconstruction and functional annotation was carried out using comparative genomics, which significantly expanded the database of bacterial metabolism, representing the microbiota of the human gastrointestinal tract (GIT). For the reporting year, 2662 bacterial genomes were analyzed, among them a search was made for enzymes of the metabolic pathways of catabolism and utilization of monosaccharides, oligo- and polysaccharides with prebiotic properties. As a result of this analysis, an extended collection of metabolic phenotypes was obtained for the utilization of the following carbohydrates (modulator phenotypes): lactulose, raffinose, allose, arabinan and arabinoligosaccharides (AOS), xylan and xylooligosaccharides (XOS), pectin, fructan and fructooligosaccharides (FOS) also oligosaccharides of breast milk (HMO). New metabolic pathways and regulators were reconstructed for the utilization of metabolites of the processing of food proteins by intestinal bacteria. A new family of fructosolysine-specific transcription factors (FrlR) has been found to be involved in the utilization of fructosolisin in various taxonomic groups of bacteria. Using comparative genomics, DNA binding sites for FrlR regulators were identified. Various pathways for the degradation of amino acids and urea were reconstructed. Functional annotations of the corresponding catabolic enzymes were obtained in the reference collection of intestinal bacteria. The corresponding metabolic phenotypes were assigned and used to assess the metabolic potential of the metagenomic communities of intestinal bacteria for degradation of the following amino acids: tryptophan, histidine, lysine, isoleucine, leucine, valine, methionine, threonine, proline. We carried out the genomic study of the global regulators AraQ and MalR1, controlling the utilization and metabolism of sugars in 70 genomes of bifidobacteria belonging to the dominant intestinal microbiota of children of the first year of life. A genomic set of carbohydrate-active catabolic enzymes encoded in the reference collection of the GIT collection of bacterial genomes was obtained. It included a total of 109,000 carbohydrate-active catabolic enzymes, which were classified into 229 families of glycosyl hydrolases and 57 families of polysaccharide lyases. We completed the comparative genomic analysis and metabolic reconstruction of the utilization of short chain fatty acids (butyrate, lactate, ethanolamine and propanediol) by GIT bacteria. We continue the development of the Phenotype Profiler software package for the automated analysis of metagenomic samples of the human intestine with the aim of predicting their total metabolic phenotypes. The following software modules were developed and implemented in 2019: 1) a module for multitaxonomic profiling that provides detailed assignment of taxonomies for metagenomic samples of amplicon sequencing format; 2) a module for renormalization of taxonomic representations by the number of copies of the 16S rRNA gene; 3) a module for calculating taxonomic and phenotypic alpha diversity based on taxonomic and phenotypic profiles of bacterial communities. An extensive collection of sequencing data and metadata was composed for available sets of metagenomic samples of GIT microbiota from relevant published studies of interest that was used for subsequent taxonomic and phenotypic profiling. For the selected studies, the primary sequencing data were downloaded from public databases and nucleotide archives. The Phenobase database has been created to systematize and standardize the description of metagenomic data from various studies. The collected metadata allows one to break down samples according to various classifications, for example, geography, sex and physiological characteristics of microbiota donors, groups healthy people or patients with a specific medical diagnosis, as well as various lifestyles, dietary preferences, the use of antibiotics, vitamins, etc. Using the developed software for metabolic profiling, we performed bioinformatics analysis of 7 metagenomic studies selected from Phenobase, as well as for two studies on the cultivation of fecal microbiota in vitro. The obtained taxonomic and phenotypic profiles were statistically processed, associations between phenotypes and diseases and / or concentrations of metabolites from metadata were identified using statistical approaches and machine learning methods. The obtained phenotypic profiles allowed us to classify the processed samples according to metadata and also to detect phenotypes with the largest contribution to the obtained classifications. The first prototype of the Phenotype Predictor algorithm was developed for the metabolic annotation of new bacterial genomes using the genomic encyclopedia of metabolic pathways, rules, and phenotypes collected as part of this project. The Phenotype Predictor algorithm will allow to quickly expand the phenotype database by newly sequenced genomes. The algorithm was tested on the current reference genome collection, and then applied to two extensive collections of intestinal microorganisms that appeared in 2019. In total, more than 2250 new bacterial genomes were analyzed. The obtained metabolic phenotypes for biosynthetic pathways for vitamins and amino acids in the newly analyzed genomes were used to assess the intra- and interspecies variability of binary phenotypes within the extended collection of GIT bacterial genomes (totally about 5000 genomes).

 

Publications

---

Annotation of the results obtained in 2020
We have continued development of software modules for the automated analysis of metagenomic samples to calculate their taxonomic and metabolic/phenotypic characteristics (metrics). In particular, we developed and implemented the following software modules: 1) 16S amplicons-mapping module which uses the reference collection DNA sequences; 2) Alpha-diversity and beta-diversity calculating module; 3) a module for visualization of the obtained taxonomic and metabolic metrics for metagenomic samples; 4) a module which provides export of obtained taxonomic profiles during the WGS sequencing data analysis. The development of an improved pipeline, which were obtained from a number of experimental studies, has been successfully applied for the 16S intestinal microbiota samples from previously published scientific papers with available 16S sequencing data category, and those which were collected during collaboration with leading experimental groups, allowing to classify each metagenomic sample according to its phenotype profile. We developed an algorithm pipeline for metabolic annotation of novel bacterial genomes and metagenomically-assembled genomes using the database of experimentally elucidated metabolic pathways along with rules and phenotypes. This pipeline allows fast and accurate expansion of the genomic database of metabolic phenotypes. It consists of two parts: 1) search for the genes/proteins orthologs from the current encyclopedia of metabolic pathways using the similarity search-based algorithm that is capable to determine the functions of multidomain proteins robustly and reject overpredictions; 2) a range of binary phenotype assignment algorithms for each gene occurrence pattern within single metabolic subsystem (using machine learning models and formal rules of phenotypes determined on a training set of 2600 reference genomes and their phenotypes using decision trees). The developed software pipeline was used to analyze metabolic phenotypes in two recent datasets of genomes: (i) a set of contained more than 2300 bacterial genomes of the intestinal tract representatives, covering a wide phylogenetic space of bacterial species, and (ii) a set of 300 new metagenomically-assembled genomes obtained from of human intestinal samples. Two other software pipelines that were developed are capable for functional metabolic analysis of metagenomic samples in whole-genome sequencing (WGS) format and use information on the metabolic pathways and genes obtained from the comparative genomic reconstruction for more than 2600 reference genomes of human intestinal microbiome. The first pipeline is based on the taxonomic profiling of WGS samples. The second pipeline includes de novo annotation methods to identify functionally important genes in a sample and seed-extend-based mapping methods to assess sample coverage. Both developed pipelines were used for analysis and functional classification of gut metagenomic samples in WGS format that the published HMP2 study on inflammatory bowel disease. Also, we applied the first pipeline to an extensive paired 16S/WGS dataset of fecal samples of 1-3 years old children. The metabolic profiling results for the analyzed metagenomes in the WGS and 16S formats have shown a high correlation rate. In another task, we have expanded metabolic reconstructions and functional annotations of new metabolic pathways and phenotypes for 2662 representatives of the human intestinal microbiota. Genomic reconstructions were obtained for the following new phenotypes: 1) the microbial transformation of the conjugated bile acids; 2) polysaccharide degradation using an expanded collection of carbohydrate-active catabolic enzymes, including the degradation of cellodextrin, cellulose, starch, galactan/arabinogalactan, arabinoxylan, beta-mannan, rhamnogalacturonan, mucin, N- and O-glycans, as well as degradation of human milk oligosaccharides and sialic acid-containing polysaccharides. In order to increase metagenomic samples coverage, the previously described metabolic pathways were also analyzed in 200 novel bacterial genomes to ensure the reference collection expansion. Bioinformatic analysis was performed using the developed technology and software pipelines for metabolic profiling of a large number of metagenomic datasets in 16S format, that were collected from the previously published studies of intestinal microbiota from patients suffering from various chronic diseases. For selected studies, we downloaded the raw sequencing data from public databases and nucleotide archives and used the the developed phenotypic profiling software pipelines to analyze their metabolic phenotype potentials. The collected metadata made it possible to split the samples into groups of sick and healthy people and conduct a comparative analysis and classification. In particular, the metagenomes of cohorts of people suffering from type 1 diabetes mellitus, colorectal cancer, Parkinson's disease and inflammatory bowel disease were analyzed. In each dataset, we identified metabolic phenotypes that are potentially associated with the development of the disease. We also carried classification of samples using the machine learning approaches and basing on the predicted metabolic phenotype indices. The most important phenotypic predictors were subject to a t-test to evaluate significance of mean values differences in community phenotype indices between the sick and healthy cohorts. The classifiers were comparable in quality to conventional taxonomy-based classifiers but provided new findings giving insights into possible mechanisms of pathogenesis. Feature-wise partial dependence plots analysis of contribution to the classification result revealed a diversity of patterns. These observations suggest a constructive basis for defining functional homeostasis of the healthy human gut microbiome. The developed features are promising interpretable candidate biomarkers for assessing microbiome contribution to disease risk for the purposes of personalized medicine and clinical trials. Last, in order to explore the potential for productions of short-chain fatty acids and lactate in the gut microbiome we performed the following analyses. For each fermentation product and each genome, binary phenotypes were assigned, encoding the ability or inability of a given organism to produce the corresponding metabolite. The derived phenotypes were used for metabolic profiling of metagenomic samples from the intestinal microbiota. In particular, Cumulative Phenotypic Indexes, Phenotypic Alpha-Diversity and Phenotypic Beta-Diversity were calculated for each analyzed samples from several large metagenomic datasets of gut microbiota, including healthy populations from two broad western cohorts, one hunter-gatherer cohort from Tanzania, and several cohorts of people with chronic diseases. As result, we identified a wide range of metabolic potentials for the short-chain fatty acids production by the intestinal microbiota in healthy populations. To validate our model, we processed multiple published metagenomics datasets of the intestinal microbiota from studies that analyze the effect of diets on the intestinal microbiota (in vivo), and also from studies that analyze the effect of various prebiotics on in vitro cultured fecal microbiota. The obtained propionate and butyrate synthesis metabolic potentials reveal positive correlation with experimental data on the measured concentrations of these metabolites in these intestinal samples.

 

Publications

---