Publications

Computational methods to predict Z-DNA regions are in high demand to understand the functional role of Z-DNA. The previous state-of-the-art method Z-Hunt is based on statistical mechanical and energy considerations about B- to Z-DNA transition using sequence information. Z-DNA CHiP-seq experiment results showed little overlap with Z-Hunt predictions implying that sequence information only is not sufficient to explain emergence of Z-DNA at different genomic locations. Adding epigenetic and other functional genomic mark-ups to DNA sequence level can help revealing the functional Z-DNA sites. Here we take advantage of the deep learning approach that can analyze and extract information from large volumes of molecular biology data. We developed a machine learning approach DeepZ that aggregates information from genome-wide maps of epigenetic markers, transcription factor and RNA polymerase binding sites, and chromosome accessibility maps. With the developed model we not only verify the experimental Z-DNA predictions, but also generate the whole-genome annotation, introducing new possible Z-DNA regions, which have not yet been found in experiments and can be of interest to the researchers from various fields.

Computational methods to predict Z-DNA regions are in high demand to understand the functional role of Z-DNA. The previous state-of-the-art method Z-Hunt is based on statistical mechanical and energy considerations about B- to Z-DNA transition using sequence information. Z-DNA CHiP-seq experiment results showed little overlap with Z-Hunt predictions implying that sequence information only is not sufficient to explain emergence of Z-DNA at different genomic locations. Adding epigenetic and other functional genomic mark-ups to DNA sequence level can help revealing the functional Z-DNA sites. Here we take advantage of the deep learning approach that can analyze and extract information from large volumes of molecular biology data. We developed a machine learning approach DeepZ that aggregates information from genome-wide maps of epigenetic markers, transcription factor and RNA polymerase binding sites, and chromosome accessibility maps. With the developed model we not only verify the experimental Z-DNA predictions, but also generate the whole-genome annotation, introducing new possible Z-DNA regions, which have not yet been found in experiments and can be of interest to the researchers from various fields.

The aim of this study was to determine the prevalence of CYP2C9, VKORC1, CYP2C19, ABCB1, CYP2D6 and SLCO1B1 genes polymorphisms among residents of the Volga region (Chuvash and Mari) and northern Caucasus (Kabardins and Ossetians). Materials & methods: The study involved 845 apparently healthy volunteers of both sexes of the four different ethnic groups living in the Russian Federation: 238 from the Chuvash ethnic group, 206 Mari, 157 Kabardins and 244 Ossetians. Results: Significant differences were identified in allele frequency of CYP2C9, VKORC1, CYP2C19, ABCB1, CYP2D6 and SLCO1B1 genes polymorphisms between the Chuvash and Kabardins, Chuvash and Ossetians, Mari and Kabardians, Mari and Ossetians.

Background

Eukaryotic protein-coding genes consist of exons and introns. Exon–intron borders are conserved between species and thus their changes might be observed only on quite long evolutionary distances. One of the rarest types of change, in which intron relocates over a short distance, is called "intron sliding", but the reality of this event has been debated for a long time. The main idea of a search for intron sliding is to use the most accurate genome annotation and genome sequence, as well as high-quality transcriptome data. We applied them in a search for sliding introns in mammals in order to widen knowledge about the presence or absence of such phenomena in this group.

Results

We didn’t find any significant evidence of intron sliding in the primate group (human, chimpanzee, rhesus macaque, crab-eating macaque, green monkey, marmoset). Only one possible intron sliding event supported by a set of high quality transcriptomes was observed between EIF1AX human and sheep gene orthologs. Also, we checked a list of previously observed intron sliding events in mammals and showed that most likely they are artifacts of genome annotations and are not shown in subsequent annotation versions as well as are not supported by transcriptomic data.

Conclusions

We assume that intron sliding is indeed a very rare evolutionary event if it exists at all. Every case of intron sliding needs a lot of supportive data for detection and confirmation.

Over the past few years, genome research using machine and in-depth learning techniques has become increasingly popular, and researchers are being provided with sophisticated data analysis tools. Recognition of patterns of DNA secondary structures and genomic functional elements are still poorly investigated, despite the fact that research in this area has the potential to contribute greatly to the development of medicine and pharmacology. This study aims to explore machine and deep learning methods that have proven to be successful in natural language processing with respect to the task of DNA sequence recognition. Two deep learning models based on CNN and LSTM architectures were developed. Each model was tested on multiple classification tasks for recognition of DNA sequences containing quadruplexes with potential function of nucleosome barriers. Additionally, model interpretation analysis was performed in the form of extraction of CNN significant filters and their transformation into DNA-motifs.

Over the last 20 years whole-genome sequencing of cancer genomes supported the phenomenon of cancer mutation heterogeneity both for point and structural variants. Alongside with the whole-genome sequencing projects many next-generation sequencing experiments including ChIP-seq for histone modifications and transcription factors, DNase-seq, MeDIP-Seq, Hi-C, and others were collected for thousands of cancer genomes. Machine learning approach became an efficient method of predictive modeling because machine learning algorithms are able to consider multiple factors and their interactions and range them in an order of importance. Machine learning models, predicting cancer point mutations at 1Mb scale and using as predictors state of the chromatin, epigenetic factors and non-B DNA structures, achieved a good predictive power. However, predicting cancer breakpoints appeared to be a more difficult task than predicting point mutations. Machine learning models, that were successfully used to predict cancer point mutations, using the same features, could not achieve high performance in predicting cancer breakpoints. Nevertheless, the available models demonstrate that aggregating information from omics experiments increases the model prediction power. Here we review state-of-the art machine learning approaches to predict cancer breakpoints and discuss current understanding of the determinants of cancer breakpoint formation.

BACKGROUND: Chlamydia are ancient intracellular pathogens with reduced, though strikingly conserved genome. Despite their parasitic lifestyle and isolated intracellular environment, these bacteria managed to avoid accumulation of deleterious mutations leading to subsequent genome degradation characteristic for many parasitic bacteria. RESULTS: We report pan-genomic analysis of sixteen species from genus Chlamydia including identification and functional annotation of orthologous genes, and characterization of gene gains, losses, and rearrangements. We demonstrate the overall genome stability of these bacteria as indicated by a large fraction of common genes with conserved genomic locations. On the other hand, extreme evolvability is confined to several paralogous gene families such as polymorphic membrane proteins and phospholipase D, and likely is caused by the pressure from the host immune system. CONCLUSIONS: This combination of a large, conserved core genome and a small, evolvable periphery likely reflect the balance between the selective pressure towards genome reduction and the need to adapt to escape from the host immunity.

Aromatic compounds are a common carbon and energy source for many microorganisms, some of which can even degrade toxic chloroaromatic xenobiotics. This comparative study of aromatic metabolism in 32 Betaproteobacteria species describes the links between several transcription factors (TFs) that control benzoate (BenR, BenM, BoxR, BzdR), catechol (CatR, CatM, BenM), chlorocatechol (ClcR), methylcatechol (MmlR), 2,4-dichlorophenoxyacetate (TfdR, TfdS), phenol (AphS, AphR, AphT), biphenyl (BphS), and toluene (TbuT) metabolism. We characterize the complexity and variability in the organization of aromatic metabolism operons and the structure of regulatory networks that may differ even between closely related species. Generally, the upper parts of pathways, rare pathway variants, and degradative pathways of exotic and complex, in particular, xenobiotic compounds are often controlled by a single TF, while the regulation of more common and/or central parts of the aromatic metabolism may vary widely and often involves several TFs with shared and/or dual, or cascade regulation. The most frequent and at the same time variable connections exist between AphS, AphR, AphT, and BenR. We have identified a novel LysR-family TF that regulates the metabolism of catechol (or some catechol derivative) and either substitutes CatR(M)/BenM, or shares functions with it. We have also predicted several new members of aromatic metabolism regulons, in particular, some COGs regulated by several different TFs.

DNAsecondary structures are important functional elements thatmay influence cellular processes. One of theirpossible functions is regulation of nucleosome positioning. Here MNAse-seq and ssDNA-seq data were used to define patterns of positional relationship of DNA structures such as Z-DNA, H-DNA and G-quadruplexes with nucleosomes. Three types of patterns werefound: a structure is surrounded by nucleosomes from both sides, from one side, or nucleosome free region. Machine-learning models based on Random forest algorithm and XGBoost weretrained to recognize DNA region of 500 bp length containing a pattern of nucleosome positioning for three types of DNA struc-tures (Z-DNA, H-DNA and G-quadruplexes) based on DNAsequence composi-tional properties. The best performance (more than 86% for ROC-AUC, accu-racy, recall and presicion scores) wasreached for G-quadruplexes. 500 bp re-gions containing G-quadruplexes have distinct compositional properties and point to the preferential locations of the defined patterns, which regulatory functions require further investigation. For other DNA structures a region com-position is less powerful predictive factor and one should take into account oth-er physical and structural DNA properties to improve nucleosome-DNA-structure pattern recognition.

BACKGROUND:

The genus Streptococcus comprises pathogens that strongly influence the health of humans and animals. Genome sequencing of multiple Streptococcus strains demonstrated high variability in gene content and order even in closely related strains of the same species and created a newly emerged object for genomic analysis, the pan-genome. Here we analysed the genome evolution of 25 strains of Streptococcus suis, 50 strains of Streptococcus pyogenes and 28 strains of Streptococcus pneumoniae.

RESULTS:

Fractions of the pan-genome, unique, periphery, and universal genes differ in size, functional composition, the level of nucleotide substitutions, and predisposition to horizontal gene transfer and genomic rearrangements. The density of substitutions in intergenic regions appears to be correlated with selection acting on adjacent genes, implying that more conserved genes tend to have more conserved regulatory regions. The total pan-genome of the genus is open, but only due to strain-specific genes, whereas other pan-genome fractions reach saturation. We have identified the set of genes with phylogenies inconsistent with species and non-conserved location in the chromosome; these genes are rare in at least one species and have likely experienced recent horizontal transfer between species. The strain-specific fraction is enriched with mobile elements and hypothetical proteins, but also contains a number of candidate virulence-related genes, so it may have a strong impact on adaptability and pathogenicity. Mapping the rearrangements to the phylogenetic tree revealed large parallel inversions in all species. A parallel inversion of length 15 kB with breakpoints formed by genes encoding surface antigen proteins PhtD and PhtB in S. pneumoniae leads to replacement of gene fragments that likely indicates the action of an antigen variation mechanism.

CONCLUSIONS:

Members of genus Streptococcus have a highly dynamic, open pan-genome, that potentially confers them with the ability to adapt to changing environmental conditions, i.e. antibiotic resistance or transmission between different hosts. Hence, integrated analysis of all aspects of genome evolution is important for the identification of potential pathogens and design of drugs and vaccines.

How the nuclear lamina (NL) impacts on global chromatin architecture is poorly understood. Here, we show that NL disruption in Drosophila S2 cells leads to chromatin compaction and repositioning from the nuclear envelope. This increases the chromatin density in a fraction of topologically-associating domains (TADs) enriched in active chromatin and enhances interactions between active and inactive chromatin. Importantly, upon NL disruption the NL-associated TADs become more acetylated at histone H3 and less compact, while background transcription is derepressed. Two-colour FISH confirms that a TAD becomes less compact following its release from the NL. Finally, polymer simulations show that chromatin binding to the NL can per se compact attached TADs. Collectively, our findings demonstrate a dual function of the NL in shaping the 3D genome. Attachment of TADs to the NL makes them more condensed but decreases the overall chromatin density in the nucleus by stretching interphase chromosomes.

The role of 3’-end stem-loops in transposition was experimentally demonstrated for transposons of various species, where LINE-SINE transposons share the same 3’-end sequences, containing a stem-loop. We have discovered that 62-68% of processed pseduogenes and mRNAs also have 3’-end stem-loops. We investigated the properties of 3’-end stem-loops of human L1s, Alus, processed pseudogenes and mRNAs that do not share the same sequences, but all have 3’-end stem-loops. We have built sequence-based and structure-based machine-learning models that are able to recognize 3’-end L1, Alu, processed pseudogene and mRNA stem-loops with high performance. The sequence-based models use only sequence information and capture compositional bias in 3’-ends. The structure-based models consider physical, chemical and geometrical properties of dinucleotides composing a stem and position-specific nucleotide content of a loop and a bulge. The most important parameters include shift, tilt, rise, and hydrophilicity. The obtained results clearly point to the existence of structural constrains for 3’-end stem-loops of L1 and Alu, which are probably important for transposition, and reveal the potential of mRNAs to be recognized by the L1 machinery. The constructed models are freely available at github (https://github.com/AlexShein/transposons/) and can be used for de novo discovery of transposon-related stem-loops.

We trained Random Forest model to recognize patterns of nucleosome and non-B DNA structures, considered as potential nucleosome barriers in the mouse genome. We showed that among four types of structures – Z-DNA, H-DNA, G-Quadruplexes and SIDD regions – recognition of G-Quadruplexes and H-DNA showed the best performance.

We built and evaluated two types of models: sequence-based and structure-based for recognition of 3’-end stem- loops of human L1s and Alus and found most important parameters contributing to recognition: Shift, Tilt and Rise, and aslo hydrophilicity.

Background: Chromosomal rearrangements are the typical phenomena in cancer genomes causing gene disruptions and fusions, corruption of regulatory elements, damage to chromosome integrity. Among the factors contributing to genomic instability are non-B DNA structures with stem-loops and quadruplexes being the most prevalent. We aimed at investigating the impact of specifically these two classes of non-B DNA structures on cancer breakpoint hotspots using machine learning approach.

Methods: We developed procedure for machine learning model building and evaluation as the considered data are extremely imbalanced and it was required to get a reliable estimate of the prediction power. We built logistic regression models predicting cancer breakpoint hotspots based on the densities of stem-loops and quadruplexes, jointly and separately. We also tested Random Forest models varying different resampling schemes (leave-one-out cross validation, train-test split, 3-fold cross-validation) and class balancing techniques (oversampling, stratification, synthetic minority oversampling).

Results: We performed analysis of 487,425 breakpoints from 2234 samples covering 10 cancer types available from the International Cancer Genome Consortium. We showed that distribution of breakpoint hotspots in different types of cancer are not correlated, confirming the heterogeneous nature of cancer. It appeared that stem-loop- based model best explains the blood, brain, liver, and prostate cancer breakpoint hotspot profiles while quadruplex- based model has higher performance for the bone, breast, ovary, pancreatic, and skin cancer. For the overall cancer profile and uterus cancer the joint model shows the highest performance. For particular datasets the constructed models reach high predictive power using just one predictor, and in the majority of the cases, the model built on both predictors does not increase the model performance.

Conclusion: Despite the heterogeneity in breakpoint hotspots’ distribution across different cancer types, our results demonstrate an association between cancer breakpoint hotspots and stem-loops and quadruplexes. Approximately for half of the cancer types stem-loops are the most influential factors while for the others these are quadruplexes. This fact reflects the differences in regulatory potential of stem-loops and quadruplexes at the tissue-specific level, which yet to be discovered at the genome-wide scale. The performed analysis demonstrates that influence of stem- loops and quadruplexes on breakpoint hotspots formation is tissue-specific.

Background. Restriction-modification (R-M) systems protect bacteria and archaea from attacks by bacteriophages and archaeal viruses. An R-M system specifically recognizes short sites in foreign DNA and cleaves it, while such sites in the host DNA are protected by methylation. Prokaryotic viruses have developed a number of strategies to overcome this host defense. The simplest anti-restriction strategy is the elimination of recognition sites in the viral genome: no sites, no DNA cleavage. Even a decrease of the number of recognition sites can help a virus to overcome this type of host defense. Recognition site avoidance has been a known anti-restriction strategy of prokaryotic viruses for decades. However, recognition site avoidance has not been systematically studied with the currently available sequence data. We analyzed the complete genomes of almost 4000 prokaryotic viruses with known host species and more than 17,000 restriction endonucleases with known specificities in terms of recognition site avoidance.

Results. We observed considerable limitations of recognition site avoidance as an anti-restriction strategy. Namely, the avoidance of recognition sites is specific for dsDNA and ssDNA prokaryotic viruses. Avoidance is much more pronounced in the genomes of non-temperate bacteriophages than in the genomes of temperate ones. Avoidance is not observed for the sites of Type I and Type IIG systems and is very rarely observed for the sites of Type III systems. The vast majority of avoidance cases concern recognition sites of orthodox Type II restriction-modification systems. Even under these constraints, complete or almost complete elimination of sites is observed for approximately one-tenth of viral genomes and a significant under-representation for approximately one-fourth of them.

Conclusions. Avoidance of recognition sites of restriction-modification systems is a widespread but not universal anti-restriction strategy of prokaryotic viruses.

Riboswitches are conserved RNA structures located in non-coding regions of mRNA and able to bind small molecules (e.g. metabolites) changing conformation upon binding. This feature enables them to function as regulators of gene expression. The thiamin pyrophosphate (TPP) riboswitch is the only type of riboswitches found not only in bacteria, but also in eukaryotes - in plants, green algae, protists, and fungi. Two main mechanisms of fungal TPP riboswitch action, involving alternative splicing, have been established so far. Here, we report a large-scale bioinformatic study of riboswitch structural features, action mechanisms, and distribution along the fungal taxonomy groups. For each putatively regulated gene, we reconstruct the riboswitch structure, identify other components of the regulation machinery, and establish mechanisms of riboswitch-mediated regulation. In addition to three genes known to be regulated by TPP riboswitches, thiazole synthase THI4, hydroxymethilpyrimidine-syntase NMT1, and putative transporter NCU01977, we identify two new genes, a putative thiamin transporter THI9 and a transporter of unknown specificity. While the riboswitch sequence and structure remain highly conserved in all species and genes, the mode of riboswitch-mediated regulation varies between regulated genes. The riboswitch usage varies strongly between fungal taxa, with the largest number of riboswitch-regulated genes found in Pezizomycotina and no riboswitch-mediated regulation established in Saccaromycotina.

While most endosymbiotic bacteria are transmitted only vertically, Holospora spp., an alphaproteobacterium from the Rickettsiales order, can desert its host and invade a new one. All bacteria from the genus Holospora are intranuclear symbionts of ciliates Paramecium spp. with strict species and nuclear specificity. Comparative metabolic reconstruction based on the newly sequenced genome of Holospora curviuscula, a macronuclear symbiont of Paramecium bursaria, and known genomes of other Holospora species shows that even though all Holospora spp. can persist outside the host, they cannot synthesize most of the essential small molecules, such as amino acids, and lack some central energy metabolic pathways, including glycolysis and the citric acid cycle. As the main energy source, Holospora spp. likely rely on nucleotides pirated from the host. Holospora-specific genes absent from other Rickettsiales are possibly involved in the lifestyle switch from the infectious to the reproductive form and in cell invasion.