Mitochondrial diseases are a group of inherited disorders characterized by impaired mitochondrial function, leading to diverse clinical manifestations. The identification of reliable biomarkers for mitochondrial diseases is crucial for early diagnosis, accurate prognosis, and the development of targeted therapeutic interventions. Although several diagnostic biomarkers are implemented in clinical practice, they demonstrate relatively low accuracy, and therefore this area of research is still very prospective. In this study, we aimed to discover and explore potential biomarkers of several mitochondrial diseases through the analysis of multi-omics data.
We conducted an in-depth analysis of lipidomics datasets obtained with an untargeted Liquid Chromatography-Mass Spectrometry (LC‐MS) approach on plasma samples from individuals diagnosed with mitochondrial diseases and healthy controls. Our research employed advanced statistical and machine learning techniques to identify the most performant molecules specific to different mitochondrial diseases. The performance of selected compounds was evaluated in a multiclass classification task. Through our condition-specific analysis of identified potential biomarkers of Leber's Hereditary Optic Neuropathy (LHON), we unveiled two distinct clusters of lipids that consistently exhibited dysregulation patterns. Further investigation of candidate molecules using pathway enrichment and correlation analysis provided the ability to propose the chemical class of these lipids. The developed workflow proved to be an efficient strategy for novel biomarkers discovery. In further studies, it can be modified to add new functionality or redirected to use in a wide range of bioinformatics tasks.

Positive-unlabeled (PU) data consists of points with either positive or unknown labels. It is widespread in medical, genetic, and biological settings, creating a high demand for predictive PU models. The performance of such models is usually estimated using validation sets, assumed to be selected completely at random (SCAR) from known positive examples. Then, the estimated performance should be representative of true performance. We show, that in practice SCAR assumption often does not hold.
Using the developed simulation framework, we obtain the distribution of performance estimates under user-specified assumptions about the labeling process, PU data composition, and the true model quality. Applying it to the problem of gene prioritization for complex traits, we illustrate that SCAR assumption is often violated there, causing the inflation of performance estimates, which we call validation bias. We estimate the potential effect of validation bias, as well as other user-specified assumptions, on the performance estimation metric values.

Nuclear receptor (NR)-ligand interactions are essential for the regulation of a wide range of biological processes. However, the factors that influence binding affinity between NRs and ligands are not fully understood. This study investigated the use of machine learning to identify common and unique features that explain binding affinity for three different NRs (VDR, PXR, and CAR). A bile acid-like library was composed and downloaded based on 4 reference molecules. The library was preprocessed using tools provided by Schrodinger Maestro, and docking simulations were performed using GLIDE. The docking results were transferred for analysis using Python programming language , and a Random Forest Regressor model was used to identify the most significant features with respect to the strength of ligand-receptor binding. The results showed that a large number of features were correlated with binding affinity, but that a subset of these features were most important for each NR. The identified features were used to create a set of rules for ligand selection, which could potentially be used to reduce the number of drug candidates in the drug design pipeline.

Sex chromosomes of mammals have a short region of true similarity on the end of chromosomes called pseudoautosomal region (PAR). It has a lot of unique properties but has not been studied for many species of mammals yet. With resent advancements in sequencing technologies hundreds of species with high quality genome assemblies are available. To study PAR region male raw read data for several species of eutherian mammals was aligned to chromosome level assemblies and coverage was calculated to identify PAR region. New algorithm to split chromosome into regions with normal and half coverage based on least squares approach was devised and successfully tested using as benchmark human and mouse genome assemblies with known PAR coordinates. For Grey mouse lemur (Microcebus murinus) andEuropean rabbit (Oryctolagus cuniculus) noPAR candidates were found. The smallest eutherian PAR with length 0.7 Mbis in mouse (Mus musculus) and the biggest was found in Hyrax 11.8Mb (Heterohyrax brucei). Second PAR region was identified only for humans. Across 12 analyzed carnivora species and closely related to them pangolin PAR size was similar (6.6±0.6Mb). Heterozygosity assessment for five species showed that PAR has similar levels of heterozygosity as autosomes. Rough estimation of gene content in identified PAR regions showed that the last PAR gene in most analyzed species is SHROOM2, probably due to its fertility importance. The Mid1 gene in mouse PAR indicates a novel translocation from X to Y chromosome.

The southern black rhino (Diceros bicornis minor) is a subspecies of black rhino inhabiting South Africa that has no natural enemies but humans. Long-term historical demand for their horns led wild populations of this animal to the brink of extinction. The International Rhino Foundation established the Southern Black Rhino Sustainability Program (SBRSP) to manage individuals held in zoos to develop self-sustaining southern black rhino population that allows exhibition, research, and reintroduction of animals to the wild. In this work presented results of evaluation of the genetic status of the SBRSP population. Whole-genome sequencing data of 20 individuals living in the USA were studied using about 3.5 million SNPs for each rhino. Heterozygosity quantified for each individual with mean autosomal heterozygosity 0.00091 heterozygous sites per base.Runs of homozygosity (ROHs) and private alleles detected. Site-frequency statistics collected, population structure studied. Pedigree of population compiled, studbook data confirmed with computation of kinship relatedness coefficient. Determined high impact variants and genes in which they are located. As a result, 20 individuals of SBRSP population are estimated in terms of heterozygosity and inbreeding level. No evidence of recent inbreeding events and bottleneck were detected; no lack of genetic diversity in comparison with wild rhinos identified.

The PIK3/AKT/mTOR signaling pathway plays a key role in cellular regulation, and its disruption is associated with multiple diseases. The deregulated activity of this pathway in humans has been found in many cancer types including breast, brain, lung and renal carcinoma. Therefore, proteins involved in the PIK3/AKT/mTOR pathway can be identified as potential therapeutic targets.The core of the signaling pathway includes proteins PIK3CA, AKT1, PDK1, TSC1/TSC2, Rheb, 4E-BP1, S6K1 and complexes mTORC1, mTORC2. The interactions between these proteins are critical for maintaining processes such as cell growth, aging, protein synthesis, autophagy, and metabolism. Proteins interacting within the pathway throughout the course of evolution tend to co-evolve together to maintain or refine such interactions. However, the structural details and coevolution of the protein-protein interactions (PPIs) within PIK3/AKT/mTOR signaling pathway remain unclear, and so far, only a few models of proteins' interaction complexes have been proposed.
In our study, we combine molecular modeling with co-evolutionary sequence analysis to shed the light on protein interactions in the PIK3/AKT/mTOR pathway focusing on complexes PDK1-AKT1, RAPTOR-S6K1 and RHEB-mTOR. The integration of these approaches may provide fundamental insights into the coevolution of interaction patterns between proteins involved in the PIK3/AKT/mTOR pathway and help in the investigation of new-targeted cancer drugs in the future.

In recent years CAR-T cell therapy has begun to gain popularity. Chimeric T-cell receptors used in it recognize unprocessed tumor antigens independent from antigen presentation of tumor cells. Moreover, the large variability of possible CAR-T receptor constructions allows building a receptor design, specific for al- most any antigen. However, in some cases it is also imperfect, being insufficiently efficient or possessing side cytotoxicity. Thus the dynamics of the CAR-T population is one of the important factors for the monitoring of the patient and the efficiency of the therapy. At that moment, about 6 types of CAR-T receptors are approved by the FDA for lymphoma treatment. In each particular case, the design of the CAR-T receptor depends on many factors, as a result, the exact structure and design of CAR-T receptors varies between various manufacturers and is usually not disclosed. Currently, many patients are undergoing genetic tests to look for biomarkers, and therefore have RNA-Seq data obtained from blood samples or tumors. In such cases, it would be practically helpful to use RNA-Seq data for monitoring of CAR-T population as well.
So in this work, we aim to develop an approach for the detection and estimation of the amount of CAR-T receptor transcripts in RNA-Seq data without knowledge of their exact structure. To do so we use several open datasets, which contain different CAR-T designs used to treat different types of cancers. We also use data from no cancer patients as a negative control for our detection procedure. Our pipeline revealed the presence and relevant dynamics of the CAR-T cell population in the set of samples we used. Such a method does not require additional human costs. It can become a useful instrument in clinical practice in addition to classical methods of monitoring population sizes, such as cytometry analysis.

Benchmarking of the most widely used existing solutions as well as new methods for variant calling from RNA-seq data was performed. The computational pipeline has been developed and its automation has been implemented for easy deployment and usage. Effectiveness of the proposed solution was demonstrated on test data as well as clinical samples: The pipeline showed best results (based on F-score) for all test samples; Causative mutation was prioritized in one clinical sample; Several variants were detected only by RNA-seq analysis for all clinical samples, which demonstrates how the proposed solution can contribute to clinical diagnosis when used in conjunction with targeted DNA-seq.

Gene expression deconvolution is a vital technique in transcriptome analysis, allowing the breakdown of bulk transcriptomes into cell type-specific gene expression profiles and estimation of each cell type's contributions. Linseed 2, an innovative algorithm developed at Maxim Artyomov's lab, builds upon the previous work by Zaytsev et al. to provide a significant advancement in gene expression deconvolution. This thesis focuses on evaluating Linseed 2's quality, comparing it with existing algorithms, and presenting a comprehensive benchmarking approach using quantitative and qualitative methods. We also address the algorithm's weaknesses through data preprocessing, particularly its sensitivity to outliers, by introducing an outlier filtering approach. Our research demonstrates that Linseed 2 produces biologically significant results, outperforming other gene expression deconvolution algorithms in terms of quality and performance, while our preprocessing techniques allow to mitigate the issue of outlier sensitivity.

Multiple myeloma (MM) is a cancer of plasma cells that arises from predisposing germline variants and somatic alterations accumulated during the life. With implementation of NGS methods primary and secondary genomic events leading to clinical manifestation of MM were discovered. However, molecular mechanisms that result in further evolution of tumor after clinical onset of the disease are not completely explored. Therefore, our research aimed to detect and analyze genetic alterations at different time points of MM.Using WES approach, we analyzed samples of bone marrow aspirate collected from a cohort of MM patients at four stages: before treatment, after chemotherapy, after stem cell transplantation and six months after treatment.We detected germline variants associated with MM risk. Also, we identified somatic alterations driving MM development in the study population. Affected biological processes include DNA repair, regulation of epigenetic mechanisms and chromatin organization, transcription, cell cycle, B lymphocytes function, protein metabolism and cell adhesion.In addition, we revealed dysregulation of several oncogenic signaling pathways, most prominent of them wereMAPK and NF-kB pathways.Involvement of several mechanisms at once was observed in all patients, which confirms heterogenicity of MM mutational landscape. Furthermore, we revealed markers of high risk of MM progression, namely,1q gain, which is also associated with resistance to chemotherapeutic drugs.Observation of somatic alterations in dynamics revealed signs of tumor clonal evolution. Also, mutational signatures extracted at different disease stages revealed chemotherapy induced DNA alterations.

CD8 T cells play a crucial role in adaptive immunity. Specifically, they are responsible for detection and elimination of different pathogens including cancer, viruses and bacteria. This is possible because T cells can use their TCR receptors to bind and recognize antigens or peptides within the major histocompatibility complex (MHC). This process is called pMHC-TCR interaction and it's the main event that launches T cell activation. However, the mechanism is still poorly understood. In particular, there is a phenomenon of hump-like kinetics of the number of pMHC-TCR complexes which has been observed in previous studies including ours. However, nobody was able to explain or model this manifestation. Here, we developed a computational model with different sets of biologically relevant parameters which was able to recapitulate hump-like behavior of pMHC-TCR interactions. We have shown that the CD8 coreceptor is the crucial complex for describing this type of non-linear kinetics. Moreover, we found that pMHC and CD8 as well as pMHC-TCR and CD8 interaction had the greatest impact on non-linearity of the pMHC-TCR binding kinetics.

In this study, we investigated the relationship between common and rare genetic variants and the risk of Parkinson's disease (PD) in a Russian cohort (544 unrelated PD patients and 433 controls). The specific objectives were to replicate genome-wide association study (GWAS) results, identify novel associations and analyze rare variants previously implicated in PD genes. Through an additive model of logistic regression, we analyzed the association between common genetic variants and PD risk, but no statistically significant associations were found in our cohort. Additionally, we explored the impact of rare variants within 50 genes located in known PD risk loci. By utilizing the ClinVar and CADD databases, we identified 10 rare pathogenic/likely pathogenic variants among PD patients. Notably, we discovered a novel mutation (p.E46K) in the SNCA gene, which had not been previously reported in the Russian population. This mutation was found in a patient with early-onset PD and a family history of the disease. Furthermore, we identified pathogenic variants in GBA (p.D409H) and LRRK2 (p.Gly2019Ser and p.Arg1441Cys) genes. The LRRK2 p.Gly2019Ser mutation was the most common in our cohort (2.4% of patients), corroborating findings from other studies. Additionally, we observed a compound carrier of two rare pathogenic heterozygous variants (p.Arg421Trp/p.Gly99Ser) in the MCCC1 gene, which had not previously been reported as associated with PD. Employing the SKAT-O method for statistical analysis of rare variants, we found nominal associations with PD in 15 genes, including for those identified pathogenic variants (GBA, LRRK2, MCCC1, NPC1, and SNCA). However, after adjusting for multiple testing, none of the rare variants showed a statistically significant association with PD. Therefore, further research is necessary to validate and replicate the findings from this study.

The Eriophyoidea is a superfamily of plant-eating mites (Acariformes), widespread on most continents of the world [...].
This work presents an attempt to study the phylogenetics and metagenomics of poorly studied and new species of Eriophyoid mites, based on whole genome sequencing metagenomic data. The metagenomic analysis was performed on these samples aimed to identify symbionts common to such mites, as well as to statistically analyze the influence of the presence of symbiotic bacteria on the ability of the samples to induce gall formation. This, in turn, allowed for the detection of a slight correlation between the presence of Wolbachia bacteria and the induction of gall formation by mites, which corresponds to the hypothesis of other studies [3]. Phylogenetic analysis was performed by identifying orthologous sequences of arachnids in these assemblies. As a result, phylogenetic trees were obtained using different models, indicating the phylogenetic proximity of the studied Eriophyoid mite specimens to organisms from the Nematalycidae family.

Nowadays, there are plenty of health outcomes, more especially diseases, for which the probability of risk of such incidents can be calculated using several machine learning and deep learning techniques. There were already several attempts to develop such models for different health outcomes. However, the model design depends on the dataset and a certain health outcome, only a small number of known features is used and resulted model can predict a certain incident only within a small time window (5-10 years). Also, not all researches include information on how to include polygenic risk score into a clinical dataset, to estimate embedded risk.
The current work tries to present an effective design of a model, that can pre- dict in a long time window (more than 10 years) a certain health outcome on selected best features from the whole dataset. Survivability was chosen as an example, but the same design is applicable to predict other health outcomes. The resulting design was tested on DeepSurv and Random Forest Regressor models.
Finally, different techniques for the stratification of individuals by depression incidents were proposed. Polygenic risk score for depression was included in a clinical dataset with depression incidents and embedded risks were calculated. The results can indicate that there is no significant correlation between polygenic risk score for depression and survivability.

In this study, we aimed to assess the genetic diversity of the Yakutian wolverines. For this purpose we resequenced the genomes of eight modern individuals from Yakutia. For comparison, we also included previously published data for a single Scandinavian individual. Heterozygosity analysis revealed similar and relatively low diversity of all analyzed Yakutian wolverines with 0.45 hetSNPs per kbp in average. We observed a relatively uniform distribution of heterozygous sites along pseudochromosomes with some increase in density at the ends. Scandinavian individual showed approximately twice lower heterozygosity (0.23 hetSNPs/kbp) but also with uniform distribution in the genome. Coalescent-based (PSMC) reconstruction of demographic history elucidated similar trajectories of effective population size for Scandinavian and Yakutian individuals during the past 300 thousand years. It allows us to suggest extensive and continuous gene flow between two geographically distant populations in the recent and distant past and/or origin from the same glacial refuge. However, our comparative analysis was limited by the availability of only one Scandinavian sample. Resequencing of additional individuals will improve the quality of all analyses and make it possible to model different demographic scenarios by applying AFS-based (allele frequency spectrum) methods.

Non-invasive prenatal testing (NIPT) based on cell-free DNA has emerged as a powerful tool for the detection of single gene disorders during pregnancy. This study aimed to address the need for comprehensive NIPT options for single gene disorders by developing a custom pipeline that exclusively utilizes cell-free DNA from maternal blood samples. The study evaluated various methods reported in the literature and selected two promising approaches, namely Bayes factor and Maximum Likelihood Estimation (MLE), for further investigation.
Comparative analysis revealed that MLE demonstrated comparable accuracy, sensitivity, and specificity to the Bayes factor method, without relying on the estimation of cutoff values or prior genotype probabilities. Consequently, MLE was selected as the preferred method for implementation. A prediction model was developed using Python, tailored to our project's specific requirements, enabling the construction of a pipeline that processes raw DNA sequencing reads as input and generates estimated fetal genotypes along with their respective probabilities as output.

Due to the increased affordability of next-generation sequencing, it generates vast amounts of data that clinical laboratories utilize to elucidate the complex etiology of genetic diseases. To systematically interpret this wealth of data, the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) have developed standards for sequencing variant interpretation. With the availability of public databases, these standards can be automated. This work presents a tool for automated ACMG/AMP classification, which utilizes Ensembl Variant Effect Predictor (VEP) annotated files (VEP-Classification Annotation Tool, VEP-CAT). To enhance the accuracy of pathogenic and benign variant classification, VEP-CAT implements the calculation of posterior probability of pathogenicity based on ACMG criteria. Benchmarking demonstrates that VEP-CAT outperforms existing tools, such as InterVar and TAPES, in determining pathogenicity class in terms of quality metrics such as ROC-AUC and Average Precision (AP).

Domestication is permanent genetic modification of a bred lineage that leads to, among other things, a heritable predisposition toward human association. Unlike other domesticated animals, domestication of cats was not a result of artificial selection for agricultural purposes, but rather a result of natural selection. Cats, that were behaviorally more predisposed to live among humans, were eventually domesticated. Thus, by comparing the genomes of wildcats, who were predecessors of domestic cats, with the genomes of domestic cats, we can determine the genes behind the domestic phenotype.

We performed a variant calling analysis and counted statistics, calculated heterozygosity levels and compared them among samples. Heterozygosity level of clone (Kurt) appeared to be the same as heterozygosity level of clone donor (Kupor) and lower, but comparable with other individuals. Heterozygosity per chromosomal scaffold was counted and appeared to be the same in clone and clone donor. The highest heterozygosity appeared to be in 12th, 21st and 26th chromosomes in each individual. We also counted private alleles per each individual. The clone and clone donor have almost the same level of private alleles, lower than other individuals. That can be the reason of previous filtration as far as donor's cells were cultured for a long time, and due to cloning process. We also estimated SNP density and made plots. Interestingly it appeared that each individual has a zone of high SNP density on 20th chromosome. Runs of homozygosity (ROH) were estimated using PLINK command. Kurt and Kupor both have high numbers and lengths of ROHs, which we consider as a reason of high inbeeding level in a pedegree line of this individuals.

Long-read metagenomic sequencing has recently been used to recover complete bacterial genomes from various complex metagenomic communities. Metagenome assembly algorithms however are still facing challenges in deconvolution of closely-related species and strains. De novo assemblies of highly heterogeneous bacterial species typically result in tangled assembly graphs, where some sequences could be strain-specific, while others represent species-level consensus. Such partially-collapsed representation of bacterial strains does not take full advantage of the ability of long reads to phase small variants. In this work we present an algorithm called MetaPhase that extends metagenomic phasing approaches to assembly graphs. Our algorithm operates on graph paths rather than single contigs, and iteratively simplifies assembly graphs with newly reconstructed strain contigs. We benchmark our algorithm using mock communities and show that it produces accurate and complete strain-level reconstructions and substantially improves over the initial partially-collapsed assemblies.

Presentation_Kazantseva (slides)

In this work we focused on conservation genomics of the Baikal seal (Pusa sibirica). This pinniped species isolated in Lake Baikal hides a great mystery of its origin. Even today the scientific community has plenty of assumptions and disputes about how and when seals appeared in the lake now separated from the ocean by thousands of kilometres. In the first part of this work we assessed the quality of chromosome-length assembly of the baikal seal, generated by our team early. It demonstrated a bit lower but comparable quality to available assemblies of other pinnipeds. Second part of the project was focused on annotation of analysed genome assembly. We detected coordinates of the pseudoautosomal region (PAR) on the X chromosome and performed prediction of tRNA, rRNA and protein-coding genes. As a final stage of annotation we established a connection between chromosome-length assemblies and karyotypes of baikal seal (Pusa sibirica) and five other pinniped species. Final part of work included heterozygosity and demographic history analysis for baikal seal, grey seal and spotted seal. All three pinniped species showed low heterozygosity levels. It could be explained by inbreeding in a closed ecosystem (Pusa sibirica), decrease in population size (Halichoerus grypus) and vulnerable status (Phoca largha). Reconstruction of demographic history showed the presence of a severe bottleneck happened ~300 ka for spotted and grey seals. Surprisingly, for baikal seal bottleneck is shifted by ~150 ka to present time. That could be explained by different environmental conditions that species faced after speciation. For Pusa sibirica demographic history supports the hypothesis that seals were isolated in Baikal for at least for 80-150 ka depending on mutation rate.

Presentation_Yakupova (slides)

Medulloblastoma is one of the most prevalent and highly malignant childhood brain cancers. Finding the most effective treatment strategy for this type of cancer is particularly challenging due to its remarkable molecular heterogeneity. [...] In this work, we show how single-cell RNA sequencing data analysis can be applied to uncover the origins of human medulloblastoma tumors through mapping to healthy developing human brain cells. We applied existing methods for correlation-based comparison of gene expression between different cell types as well as developed a new approach for joint analysis of cerebellar and tumor single-cell RNA sequencing data based on non-negative matrix factorization. The results demonstrated that medulloblastomas from the SHH, group 3 and group 4 are transcriptionally similar to the cells from granule neuron/unipolar brush cell cerebellar lineage, while identifying specific cells-of-origin for each group requires additional investigation.

Presentation_Petrova (slides)

Assessment of the strength of binding between a drug and its target (drug-target affinity) is an important aspect of the drug discovery process. Obtaining this data experimentally is both time-consuming and costly. Therefore, computational methods for predicting binding strength are being widely used and developed. However, the experimental data used to train these prediction models are highly inconsistent and unevenly represented for different drug-target pairs. This leads to biased algorithms that do not generalize well to new data. As a result, several approaches have been developed to address this problem. One of them is the use of multi-task learning. However, the extent of the application of this technique in this field is currently heavily undeveloped. The approach proposed in this work hopes to fill this gap. This work explores the application of methods that have not been thoroughly explored before: the use of multi-task learning with auxiliary tasks, the use of positional embedding for small molecules, the use of loss functions with the ability to fill in missing values, and the use of residual connections. The proposed method was found to achieve an accuracy of over 90%.

Presentation_Vinogradova (slides)

Analysis of the data on expressional changes associated with astrocyte-neuron interaction during synapse formation obtained using hiPSC technology highlights the importance of TGFβ signaling in neuronal response to astrocytic activation and reveals the connections between astrocyte-promoted synaptogenesis, schizophrenia and sleep deprivation. Signaling by PDGF appears to be a common significantly enriched pathway in these contexts, at the same time being closely connected to TGFβ signaling. Ligand-receptor interactions potentially involved astrocyte-neuron crosstalk are prioritized. Risk gene ranking for schizophrenia using the pipeline developed in the project emphasizes involvement of dopaminergic synapse, retrograde endocannabinoid signaling, MAPK signaling and nicotine addiction molecular mechanisms in disease pathogenesis, while also complying with involvement of genes related to astrocyte-neuron crosstalk in it.

Presentation_Pinakhina (slides)

Despite all efforts, cancer is still a major problem worldwide. Liver cancer is one of the most common cancers worldwide and is thought to be number 3 by projected cancer deaths in the USA by 2030. Standard of care for liver cancer currently includes immune checkpoint blockade application, with a clear need of biomarker guidance for treatment. Previous works established subtypes of liver cancer, but none of them efficiently described microenvironment. In this research work we collected a meta-cohort from 11 publicly available datasets of hepatocellular carcinoma expression data with 1432 total number of samples after quality control. Using previously described knowledge-based functional gene expression signatures and unsupervised clustering approach, we identified 5 clinically and biologically meaningful clusters with different properties, activated processes and survival based on microenvironment and tumor features. We validated survival findings on the holdout data with 103 samples. These clusters could be potentially used to predict prognosis and preferable treatment options for single samples of tumor tissue, but further validation is needed.

Presentation_Danilov (slides)

In recent years, the idea of lymphatic endothelial cells being an important part of the lymphatic system that is responsible for more than just transporting lymph to the lymph nodes, began to be widely spread. A number of studies have demonstrated that lymphatic endothelial cells are a highly heterogeneous cell population that can carry out a vast range of different functions depending on their anatomical location. However, despite the growing interest in this cell type, at the moment they are still poorly studied, especially in terms of application of single-cell RNA sequencing. To fill this gap, we created a compendium of lymphatic endothelial cells by extracting them from publicly available single-cell RNA-seq datasets from studies dedicated to other cell types. Almost all tissues require lymph drainage and thus contain some small amounts of lymphatic endothelial cells that we can use in our analysis. Using the existing markers for lymphatic endothelial cells we extracted them from publicly available datasets. We integrated obtained lymphatic endothelial cells to remove effects introduced by different datasets and further analyzed them in order to gain insight into tissue variations, canonical transcription factors, and other key genes. We found new possible markers for lymphatic endothelial cell population and their subpopulations and extended the list of known markers for lymphatic endothelial cells. Usage of the updated list of markers as gene signature makes the discovery of more unique lymphatic endothelial cell subsets possible, leading to a more comprehensive analysis.

Presentation_Lupova (slides)

The GAM-clustering algorithm is a method based on the joint clustering in network and correlation spaces producing metabolic modules. Previously the GAM- clustering method was adapted to the bulk RNA-Seq data. The application to the single-cell data required user supervised data preprocessing. Here we present automated pipeline for the preprocessing of the single-cell RNA sequencing data prior to the GAM-clustering including two rounds of clustering with ChooseR and kmeans. We compare results obtained using automated pipeline and simple kmeans clustering approach revealing advantages of the implemented automated pipeline. Metabolic networks built on KEGG and Rhea databases as well as metabolite and atom based networks were used in the analysis. GAM-clustering results obtained using different metabolic networks appeared to be similar with certain differences depending on the network type used in the analysis. Benchmarking of the method was performed with TiCoNE and Compass. Both methods were able to reveal biochemical properties also found by the GAM-clustering. However, GAM- clustering was able to identify unique metabolic pathways that were not found by other methods.

Presentation_Chikina (slides)

The clinical significance of somatic mutations, its expected low allelic frequency and load of high throughput sequencing data with errors determine the demand of approach that allow accurate automatic artifact variant filtration. This project aimed to develop machine learning based approach for artifact variant filtration from high throughput sequencing data. The achieved goals: the pipeline for automatic variant calling and quality metrics collection was developed; load of high throughput sequencing data with false variants called by standard variant calling tools was evaluated; set of true positive and true negative variants for machine learning was prepared; Random Forest algorithm was applied for variant classification based on quality metrics from standard variant callers. The accuracy of variant classification with developed algorithm achieved AUC 0.9996, True Positive Rate 0.9978, True Negative Rate 0.9333 – for SNV, and 0.9987, 0.9867, 0.9380, respectively - for INDELs.

Presentation_Yakushina (slides)

Adaptive immune system is aimed to control the interaction with pathogens through antigen-antibody binding. Sulfated tyrosine (sY) is suggested to play a crucial role in the binding processes, considering it is found on the binding sites of some co-receptors, like CCR5. CCR5 is a protein used by HIV-1 to infect the human cell, while the antibodies that mimic CCR5 can successfully inhibit the virus. Despite the great therapeutic potential of antibodies with sulfated Ys, we still know how they are generated. The aim of this research is to investigate if sulfated tyrosines can help antibodies to mimic such receptors; find their common features. The target motif of coupled aspartic and/or glutamic acids neighboring sulfated tyrosine ([E/D]+sY) that can help antibodies mimic receptor CCR5 was found in about 8% of the reads containing sulfated tyrosine. These reads were aligned to the D genes, and the possible algorithms of recombinations that lead to the target motif formation were revealed. The common feature of reads with aligned D genes was distinguished – the highest counts have the genes IGHD4-17, IGHD3-22, IGHD3-16.

Atherosclerotic cardiovascular disease is an inflammatory disease of the arteries. During atherosclerosis progression, the special structure with complex cellular composition called atherosclerotic plaque is formed. The differentiation relationships still remain unclear within the intima myeloid cell population associated with atherosclerotic plaque: the differentiation process which leads to foam macrophages formation in the plaque still remains unknown. New computational approaches for trajectory analysis designed for single-cell RNA sequencing data provide an opportunity to reconstruct trajectories for cells of interest. The important feature of trajectory inference is the possibility to identify genes and regulons which are statistically significant associated with the identified lineages. The project is focused on secondary analysis of public single-cell RNA sequencing studies dedicated to atherosclerosis. The results covered multiple topics including processing, integration and annotation of scRNA-seq datasets and trajectory inference of myeloid cells which were identified in all prepared scRNA-seq atherosclerosis datasets.

Presentation_Firulyova M. (slides)

Skin cancers, such as cutaneous melanoma, harbor the highest mutation burden among all malignancies. While the vast majority of these changes are consistent with UV-induced mutations, the biological effects of UV carcinogenesis have yet to be fully elucidated. We performed in silico analysis of the TCGA cutaneous melanoma cohort to find genes that selectively accumulate mutations in high UV-burden tumors. Subsequently our findings were replicated in in vivo tumors derived from human melanocytes with controlled UV-exposure to confirm UV-induced nature of identified mutations. TCGA melanoma tumors were separated into 3 groups by their UV-signature burden, analysis of per gene mutational burden was adjusted relevant clinical features, overall tumor mutational burden and gene length. In vivo tumors were generated by UV-irradiation of human melanocytes and further injection into mice. Somatic variant calling was performed on lab-generated tumors using original melanoma cell line and tumor resulting from non-irradiated cells as comparison.

Presentation will be available after publication of results

It is a well-known fact that the immune system is critical in cancer development and progression. The immune surveillance theory suggests that the immune system permanently controls the cells and tissues of the body and is responsible for recognition and killing cancer cells. However, this immune surveillance causing selection of cancer cells that are poorly immunogenic or have extensive mechanisms, allowing escape from immune detection. As a result, the malignant cells arise with the capability to slip away from immune destruction, proliferate and manifest clinically as cancer. T-cells are important component of cell-mediated immunity against cancer and they are controlled by the number of costimulatory and inhibitory signals that serve as checkpoints. Checkpoint regulators guarantee that T-cell responses maintain self-tolerance, effectively protect the organism from pathogens and malignancies. Immune checkpoint inhibitors represent a new class of immunotherapy and have demonstrated a rapid increase of overall survival rate in patients with different types of advanced cancer. [...] The aim of this work is to reveal transcription factors that can serve as predictors for treatment with immune checkpoint inhibitors in different types of cancer.

Presentation_Terekhova M. (slides)

Orthonectida Giard, 1877 is a small phylum of poorly known marine invertebrates. [...] Orthonectids dramatic loss of complexity and a unique life cycle is the only case among the large Annelida group, but the origin of orthonectids parasitism is still unknown. The main adaptation to parasitism, the plasmodium, remains underexplored and many questions related to the biology of the orthonectids parasitic stage have not yet been resolved. Discovering genes expressed explicitly in the plasmodium is an essential step towards revealing the mechanisms behind the development and functioning of orthonectids' parasitic stage. It will help to explore orthonectids adaptations to a parasitic lifestyle. Here, we present the identification and annotation of orthonectids' plasmodium-specific hypothetical proteins, intending to understand the processes behind the plasmodium functioning and the orthonectids adaptations to parasitism.

Presentation_Skalon E. (slides)

The guar plant (Cyamopsis tetragonoloba, (L.) Taub.) is a short day annual herbaceous flowering plant native to India and Pakistan, industrially important for serving as the main source of guar gum, widely used in the oil and gas industry. The exploitation of this plant in countries of northern latitudes is limited due to longer day lengths during the vegatitive season compared to its native habitat. The ability to efficiently identify early flowering guar varieties would greatly accelerate breeding efforts. Genomic resources of guar are limited, despite its economic significance. This study presents a new highly contiguous guar genome assembly and 10736 variant sites derived from RADseq data of a cohort of 192 guar plants of different varieties. A pilot genome wide association study was performed that found a number of SNP markers in proximity to previously established genes that regulated flowering time that could be used as markers in marker assisted breeding of this economically important crop.

Presentation_Beatovich A. (slides)

Human traits and diseases are the results of individual or combinatorial factors of genetics and environment. Over the last decade, genome-wide association studies (GWAS) have discovered a substantial number of associated variants for many complex traits. The success of GWAS in finding and replicating thousands of associations for thousands of phenotypes has demonstrated the usefulness of previous approaches and ushered in a new era of human genetics. However, even within European-centered GWAS data, there are local subpopulations significantly under-represented in these studies. For example, Russians, being one of the largest ethnic groups among the Europeans, remained significantly under-represented in GWASs for years. The aim of the present work was to test whether UK biobank GWAS results could be successfully applied for estimation of the polygenic risk scores in samples of Russian-descent.

Presentation_Rezapova V. (slides)

Lipids are an important class of biomolecules that are involved in many vital cellular processes. Due to their hydrophobic nature, lipids are the major constituents of biological membranes and are thus the physical basis of all living organisms because they provide the ability to separate living entities from their natural surroundings. Another task that lipids fulfil is the storage of surplus energy for later consumption. Finally, lipids are also involved in extra- and intracellular signaling processes, where they transduce signals and amplify regulatory cascades. Since lipids play a crucial role in many biological processes, any imbalance in their homeostasis can lead to serious conditions in living organisms, such as chronic inflammation, cardiovascular diseases, diabetes, and neurodegenerative diseases. Therefore, the importance of lipid influence in biomedical research should not be underestimated. [...] For now, lipidomics data cannot be easily integrated into current pipelines and it remains unclear of the particular lipid roles in metabolism, their exact function and impact in various biological processes. The aim of this study was to extend the applicability of the metabolic network analysis to lipidomics data. To do this, it was needed to build comprehensive metabolic and lipid-specific graphs, then update the currently existing pipeline for network analysis to suit lipid-specific analysis and test the pipeline on real datasets.

Presentation_Emelianova M. (slides)

Guar (Cyamopsis tetragonoloba (L.) Taub.) is an annual legume crop native to India and Pakistan. Seeds of the plant serve as a source of galactomannan polysaccharide (guar gum) used in the food industry as a stabilizer (E412) and as a gelling agent in oil and gas fracturing fluids. There were several attempts to introduce this crop to countries of more northern latitudes. However, guar is a plant of a short photoperiod, therefore, its introduction to Russia is complicated by a long day length during the growing season. Breeding of the new guar varieties insensitive to photoperiod is slowed down due to the lack of information on functional molecular markers, which, in turn, requires information on guar genome. In this work presented an attempt to use integrative transcriptome-metabolome integration approach to understand the genetic determination of flowering time variation among guar plants with different in their photoperiod sensitivity. This study was performed on nine early and six delayed flowering guar plants with the goal of finding a connection between biomarkers and differentially expressed transcripts. Metabolome-transcriptome integration was done by two different approaches: WGCNA and Shiny GAM.

Presentation_Grigorieva E. (slides)

Next generation sequencing has greatly increased the amount of data available for both research and clinical uses. However, in order to utilize this data there is a need for both accurate tools and standardized analytic pipelines, as well as resources such as variant databases that capture variation across all populations. There are several large databases that are worldwide in scope, however they often under-represent certain populations, leading to initiatives that focus on these underserved groups. One such group is the various ethnic populations within Russia. We developed a pipeline to conduct variant calling on WES data from 1739 individuals from the Russian Federation. Lastly we conducted an over-representation analysis to assess the frequency of disease causing alleles in the Russian population compared with a reference population.

Presentation_Futey M. (slides)

MiRNAs were suggested to be the key players during flax stem development. Numerous studies have shown that indeed, this type of small RNAs demostrate a great impact on the regulation of both intrusive elongation and cell wall thickening which are the most important stages of flax development. Samples of phloem fibers at the late stages of development from three poorly studied Flax varieties (fiber, linseed and wild), were used in our work. Novel miRNAs and their targets were computationally predicted for all samples. Differentially expressed miRNAs specific for fiber cultivar were identified and their mRNA targets among differentially expressed genes were predicted.

Presentation_Dun A. (slides)

Neurons receive through thousands of synapses distributed throughout dendrites. These synapses are transformed into electrical signals, which undergo specific integrations while travelling across the dendrite down to soma. Scientists have been studying dendritic computation properties for decades, but the exact mechanisms remain to be unravelled. One of the most arguable aspects of this question is the input location. Does it provide any significant influence on the output when considering in vivo network scale? To address this puzzle, we studied pathways in the superior colliculus (SC) responsible for evoking innate defensive behavior from visual stimuli. Wide-field (WF) neurons play the leading role in these pathways, receiving their inputs directly from retinal ganglion cells (RGC) 1 and inhibitory interneurons (Gad2). We used in-vivo recordings from RGC, Gad2 and WF neurons in mouse brain to build linear-nonlinear models of a WF neuron. Simple linear model fittings showed the correlation of RGC subtypes and the layer of WF neurons, to which they contribute. Implementation of the same model for inhibitory Gad2 recordings demonstrated strong inhibition in deeper, proximal to soma layers of WF neurons. When applying the activation function model to the data, no notable improvements have been observed, thus concluding that dendrites process local signals linearly. To fully tackle the problem of input locations, we still need to probe the built models on a different set of stimuli and comparison with other known models.

Presentation_Baymacheva N. (slides)

RNA sequencing (RNA-seq) is a powerful tool to study gene regulation and functioning on a transcriptional level that has been successfully used in application to various scientific questions in a plethora of organisms. Sequencing of RNA 5'-ends is an important approach for studying gene regulation with a particular focus on transcription initiation level. Many program packages for analyzing 5'-end sequencing are at the disposal of researchers. The majority of them, however, fail to address special issues arising in the context of transcription initiation and regulation processes characteristic of different domains of life thus making necessary the development of a specialized approach that would take into account these differences. The aim of this study was to develop a bacterial 5`-end RNA sequencing data analysis method with prospects of application of this method to Cappable-seq — specialized 5'-end sequencing method used for analysis of bacterial transcription start sites. In this work, we analyzed a dataset of matched samples sequenced with RNA-seq and Cappable-seq and implemented several functionalities on top of the existing ecosystem for 5'-end data analysis. Our implemented utilities allow assaying gene-expression on operon level as well as subtracting non-enriched libraries used in Cappable-seq.

Presentation_Etezova L. (slides)

The nuclear lamina is a polymer located on the inner surface of a nuclear membrane. Lamina supports the structure of the nucleus, participates in the organization of chromatin, regulation of gene expression and the processes of cell division. The major components of nuclear lamina – proteins lamin A and C – are encoded by a single gene called LMNA. Mutations in the LMNA cause diseases that are united into the laminopathy group. These disorders include cardiomyopathy, neuromuscular diseases, myo- and lipodystrophy, and metabolic syndrome. Laminopathies caused by missense mutations p.G232E and p.R482L in LMNA affect skeletal muscle tissue. To date, treatment of laminopathy is symptomatic and there are no effective medications against disease. Despite the big number of fundamental scientific researches of LMNA mutations, the exact molecular mechanisms of disorder development and muscle specificity remain unknown. In this work, we investigate gene expression and molecular pathways of muscle tissue that was altered by mutations G232E and R482L in lamin A/C gene using cell model of myoblasts C2C12 and transcriptome analysis.

Presentation_O. Ivanova (slides)

Chromothripsis is a mutational phenomenon representing a unique type of tremendous complex structural variation: initially described in cancerous genomes, as well as in other disorders, chromothripsis presents massive genomic alterations during a single cellular event characterized by the simultaneous shattering of chromosomes followed by random reassembly of the DNA fragments and subsequent ligation of broken segments' ends, ultimately resulting in newly formed, mosaic derivative chromosomes. The identification of such unforeseeable catastrophic instance has deeply modified the comprehension of the genesis and the etiology of complex genomic rearrangements and has provided new insights on cellular and molecular mechanisms for genomic instability and the role of genome maintenance pathways. Several nonexclusive mechanistic models have been proposed to explain the cause and high complexity of chromothripsis event but the molecular mechanism of such cellular catastrophe remains unclear and poorly understood, especially from the point of its prediction. The aim of present work is dedicated to analyze chromothripsis from the light of spatial genome organization and to answer such questions: do the chromothripsis rearrangements breakpoints appeared in cancer have the spatial predisposition at the genome organization of normal tissue; how the spatial location of chromothripsis breakpoints can be compared with other structural variations (SV) of non-chromothripsis origin; does the whole chromothripsis cluster has more spatial proximity within this region compared to other genome loci without chromothriptic events.

Presentation_N. Petukhova (slides)

The vast majority of human phenotypes, including diseases, are complex traits. The involvement of multiple genes and biological pathways in such phenotypes, among other factors, results in a relatively small contribution of each associated genetic marker. Genotyping array technology provides an affordable tool to find the genetic nature of the disease. Nevertheless, major complication in understanding disease biology only from GWAS often arises from inability to directly identify a complete set of causal genes. <...> We developed a novel Positive-Unlabeled (PU) learning based gene prioritization method - Gene Prioritizer (GPrior), intended for prioritizing disease-relevant genes given a matrix of gene-level features and sets of reliably causal genes. It is an ensemble of five PU bagging classifiers that finds the optimal combination of the predictions among individual PU algorithms. We tested our approach on both simulated and experimental data and estimated gene priorities for several traits (Schizophrenia, Education attainment, IBD and coronary-artery disease). GPrior delivers significantly better prediction qualities compared to individual PU-learning algorithms, conventional ML approaches, and other gene-prioritization tools used in the field. GPrior is yet not another fine-mapping approach rather it is a gene-level prioritization tool using hidden patterns of functional relatedness among the disease-relevant genes. At the same time GPrior is complementary to any fine-mapping approach and GWAS results post-processing. Altogether, GPrior fills an important and currently underdeveloped niche of methods for GWAS data post-processing, significantly improving the ability to pinpoint disease genes compared to existing solutions.

Presentation_N. Kolosov (slides)

GeneQuery is a novel geneset-based phenotype search engine that can be applied across all publicly available microarray experiments independent of the curation status. It utilizes Weighted Gene Correlation Analysis (WGCNA) unsupervised clusterization algorithm that identifies groups of genes that are co-regulated across the samples of each study. Despite being the first search engine spanning virtually all of published microarray studies for human, mouse, and rat, an obvious limitation of GeneQuery was its inability to search RNA-seq data, which became the method of choice for gene expression profiling during the last 10 years. Thus, this work features an update of GeneQuery that would allow us to search most of the published RNA-seq data. We also discuss experimental validation of some targets discovered using GeneQuery. In our earlier studies, GeneQuery revealed an unexpected connection between the transcriptional signatures of TREM2-deficient microglia and a portion of the aging-associated expression signature consisting of genes responsive to α/γ-tocopherol treatment of the mouse brain. In this work we find additional evidence of a specific transcriptional signature of TREM2-dependent microglia inflammation that is upregulated in aging murine brain and can be reversed by α/γ-tocopherol treatment. The obtained results allowed us to rethink the previous design of validation experiments. Expression signature analysis presented in this thesis started experiments to assess the efficacy of administering α/γ-tocopherol to TREM2(–/–) microglia cell culture (a model of Alzheimer's disease exacerbated by TREM2-deficiency) for mitigating pyroptosis induced by damage-associated molecules.

Presentation_B. Shpak (slides)

New studies of genome rearrangements cannot be provided without chromosome-level assemblies. The contiguity of genome scaffolds allows better understanding of the organization of chromatin inside the cell nucleus. Possibility to sequence long repeat regions provides insights into the organization of heterochromatin, large centromere, and telomere regions. However, only long reads sequencing will probably not achieve this level of genome contiguity. It can be that sequencer cannot read particular regions at all. In that case, we need good scaffolding. If we have a reference genome, there are no problems with this, but it is more complicated if there is no reference - we have to use an additional source of information. In the past, the best choice was to use mate-pairs reads. Now we have an incredible source of information about proximities in genome Hi-C. Hi-C method is excellent for scaffolding but has some issues with low signal regions and ambiguity in haplotype regions. After the finish of assembly and scaffolding, genome assemblies must be validated to avoid misassembles and misjoints. The present thesis is about all of these stages of chromosome-scale genome assembly during execution of two genome assembly projects - Mosquitos and Barncles projects.

Presentation_A. Zamyatin (slides)

Lack of Russian variant compendium represents a major gap on the genetic map of the world. Having such a compendium can greatly enrich our understanding of variation in global populations. The Genome Russia Project is unlikely to get completed soon. For the time being efforts are directed towards releasing a draft variant database using a few hundred russian exomes. A draft of the database has already been formed with the data analysis based on the Genome Analysis Toolkit (GATK3), but uniform reanalysis of samples with newer tools (i.e., GATK4) is necessary. During this project, a variant analysis pipeline based on GATK4 Best Practices has been developed. The pipeline is deployable on an HPC cluster within a containerized environment. The constructed pipeline was used for re-analysis of 1276 exome samples. The resulting variant dataset was used to compute allele frequencies, which were compared with other data sources such as the Genome Aggregation Database (gnomAD). Furthermore, statistical analyses were done for the monogenic disease prevalence in Russian population based on known pathogenic variants. Finally, we established a variant browser to make the data publicly available. This will be the first step towards developing a database similar to gnomAD comprising exome germline variants for the Russian population.

Presentation_M. Vashisth (slides)

Mutations in Mendelian diseases are located within the single genetic locus, they have low frequency but high effect size. One of the methods for finding such mutations can be RNA-sequencing analysis. It enables expression comparison between individual sample versus control samples, thus it can reveal expression outliers and imbalances in allele expression. Transcriptional level information in RNA-sequencing data can help in the discovery of novel splicing events. Validation of coding changes that impact RNA expression and splicing usually is done with RNA sequencing analysis among many other functional tests. And variant calling is also available. Thus, RNA sequencing can serve as another complementary method to confirm the diagnosis, as well as an independent method with a number of advantages. Thus, the main point of this work was to create an automated reproducible pipeline of tools that are most suitable for analyzing RNA- sequencing data in order to obtain a list of a prioritized candidate or even causative genes for help in the diagnosis of rare Mendelian diseases.

Presentation_M. Romanova (slides)

Autism spectrum disorder (ASD) includes a group of neurodevelopmental disorders characterized by social defects and stereotyped behavior. It is shocking that in most cases, the etiology of ASD is unclear, but it is generally believed that ASD has a strong genetic link. There is currently no consensus on which genes have sufficient evidence to support the relationship with ASD. Between the research team and the clinical sequencing team, estimates of the number of genes related to ASD vary widely, ranging from a few to a few hundred. The purpose of this project is to discover unique mutations associated with ASD in a cohort of 194 subjects.

Presentation_E. Gibitova (slides)

CRISPR-Cas systems are adaptive immunity that is present in the majority of archaea, about 90 percent, and almost half of the bacteria. CRISPR-Cas can capture fragments which are originated from invasive DNA sequences (spacers), such as viruses, bacteriophage for bacteria or plasmids and create a sequence-based array for cleaving viral mobile elements, and also ancillary DNA that can be either taken by transformation, natural acquisition and transduction or also target self chromosome or plasmids that are presented inside the cell. Characterization and study the evolution of CRISPR-Cas systems not only provided a better understanding of defense mechanisms in prokaryotes but also is necessary knowledge for genome editing.
CRISPR-Cas systems are under rapid evolution, and due to the additional horizontal gene transfer events, there are different combinations of Cas proteins that give rise to multiple types of CRISPR-Cas systems. Therefore, it is quite challenging to study all these diversities from an evolutionary point of view. The aim of this project is to discover the diversity and distribution of different varieties of CRISPR-Cas systems based on their effector complex (Cas proteins) across the phylogenetic tree.
We were able to identify different functional clusters of the Cas-related proteins. We showed that multiple clusters are present in major phyla, implying a high degree of HGT, and at the same time we found phyla associated with single clusters that may have evolved in isolation from bacteriophages.

Presentation_S. Nassirnia (slides)

To date, viral epidemics represent a significant threat to public health. In the last decade, at least seven viral outbreaks (COVID-19, Ebola, MERS-CoV, H1N1, H7N9 and others) have occurred resulting in numerous human deaths. In order to prevent disease spread, monitoring of its current state is highly necessary. In recent years, with the introduction of next-generation sequencing, it has become much easier to obtain comprehensive data for the pathogen samples. As a result, it is now possible to establish detailed and accurate information on the outbreak source, transmission chains and viral population composition. However, despite the abundance of the software created to serve the aforementioned objectives, there are still unresolved problems, such as the absence of an adequate system for detection of recombination events and the usage of too simplified viral populations simulations. This work aimed to address the challenges mentioned above, by creating the simulation pipeline, which includes all aspects of viral evolution within a single host, such as mutations, recombinations, changes in haplotypes fitness values and size of the population. Besides, the probabilistic model that manages recombination events was developed.

Presentation_D. Nemirich (slides)

An understanding of how microbial communities interact with plants under various environmental conditions might yield insights into macroecological processes. Since the next-generation sequencing analysis has become available, a lot of statistical methods have been adapted for research in ecology to help identify microbial signatures (groups of taxa) that are associated with some ecological patterns. Interactions between plants and microorganisms are reasonably obvious around plant roots, and the evidence of long-range plants specific responses in the bulk soil is overgrowing. However, this scientific field is covered by an insufficient number of studies, mainly due to the diversity and complexity of specific plant responses in soil communities. Multiple studies have underpinned the necessity of the evaluation of host-microbiome interactions for effective crop rotation and the prevention of soil deterioration. In this regard, proper modelling of plant-microbe interactions is a crucial step toward the rational exploitation of the microbiota for agricultural management.

Presentation_K. Maximova (slides)