Seminars 2025

15.01.2025 [online] Study of experimental data from the EQ database

Zoom

Speaker: Konovalov Dmitry, trainee researcher

The report presented the results of the analysis of the EndoQuad quadruplex database data. The database contains data from more than a thousand Chip-Seq and CuT&Tag experiments, but many of them were performed on model cell lines. It was found that most of the experimental support comes from several cell lines and tissue types. In this regard, bias in the model trained on this data is possible. In the future, it is necessary to analyze the model's predictions.

29.01.2025 [online] Generation of binders for Z-DNA binding proteins

Zoom

Speaker: Gromak Dmitry, trainee researcher

The report presented the adaptation of the pipeline consisting of RFdiffusion, ProteinMPNN and AlphaFold for generating binders to ADAR1 and ZBP1. The input data were the target protein domain structures from the PDB database, as well as information on key amino acid residues obtained from the literature and our own observations. Binders with the highest affinity were obtained for the Zalpha domain of the ADAR1 protein and analyzed using the ZDOCK program. These results require further study.

05.02.2025 [online] Analysis of intersections of DNA secondary structures with chromatin elements: genomic loops, TAD boundaries, loops of the polycomb-repressive complex

Zoom

Speaker: Rakhimov Bulat, trainee researcher

The report presents the results of the analysis of intersections of DNA secondary structures (Z-DNA and G-quadruplexes) with chromatin elements - polycomb loops, TAD boundaries and genomic loops in neurons and glial cells. These structures were obtained from Hi-C data of the human brain. A block of graphs with the distribution of secondary structures by chromosomes is presented, as well as the proportion of polycomb regions containing intersections of secondary structures with promoters. Statistically significant enrichment of secondary structures in polycomb loops and depletion at TAD boundaries and in genomic loops are shown. Genes with secondary structures in their promoters were identified and GO analysis was performed separately for transcription factors and other genes. The results revealed tissue specificity for nervous tissue and the involvement of genes in morphogenesis, development and expression regulation. These results require further study.

26.02.2025 [online] Statistical analysis of enrichment of polycomb loops in Z-DNA and G-quadruplex regions

Zoom

Speaker: Rakhimov Bulat, trainee researcher

The report presented the results of a permutation test assessing the enrichment and depletion of secondary DNA structures at intersections with genes encoding transcription factors (TFs) and with other genes. It was shown that enrichment in polycomb loops is determined by the loop regions themselves, rather than by the promoter regions of genes, for both TFs and non-TFs. Gene ontology graphs for G-quadruplexes and Z-DNA were also presented, revealing a similar composition of genes and their involvement in related biological processes. Colocalization of G-quadruplexes and Z-DNA in the same or adjacent gene regions was noted. The results obtained require further analysis.

12.03.2025 [online] Determining the bias of the quadruplex prediction model

Zoom

Speaker: Konovalov Dmitry, trainee researcher

Any classifier can make mistakes and it is important to understand what its errors are related to.

The report presented the results of the analysis of false-negative predictions of quadruplexes by the GQ-DNABERT model.

Various sources of bias were considered: the level of experimental support, nucleotide composition, tissue type, the presence of loops and mismatches in the structure, tissue type and cell line. Clustering was carried out. An increase in the number of errors was noted with a decrease in the level of experimental support, which indicates the ability of the model to predict the most confirmed structures. These results require further study.

26.03.2025 [online] Development of deep learning model architecture for predicting secondary structures associated with polycomb loops

Zoom

Speaker: Rakhimov Bulat, trainee researcher

The report presented the results of developing deep learning models for predicting DNA secondary structures associated with polycomb loop regions in neurons. DNA regions containing secondary structures and demonstrating statistically significant enrichment within polycomb loops were used for training. Various omics profiles of neuronal cells were used as input data, including ChIP-seq, ATAC-seq, and RNA-seq data. Architectures using convolutional (CNN) and recurrent (RNN) neural networks were tested. These models require further hyperparameter optimization to improve the quality of predictions.

02.04.2025 [online] Analysis of binders to find new proteins from the ADAR1 interactome

Zoom

Speaker: Gromak Dmitry, trainee researcher

The report presented the results of the structure and sequence analysis of the generated binders. Using the BLASTP tool, a search was performed in the UniProtKB Swiss-Prot database to identify new proteins from the ADAR1 interactome. Several matches with proteins from various organisms with statistically significant E-value were obtained. Their structures were analyzed for binding to the Zalpha domain using the Alphafold Multimer model. The data require further study.

23.04.2025 [online] Study of the role of quadruplexes in enhancer-promoter interactions

Zoom

Speaker: Konovalov Dmitry, trainee researcher

It has been shown previously that quadruplexes are often found in promoters and enhancers, and can also interact with each other via stacking or intertwining. The report presented the results of the analysis of 10x genomics Chromium (RNA-seq + ATAC-seq) and BG4 Cut&Tag data. The frequencies of quadruplexes in promoters, enhancers, and their co-occurrence were studied. BG4 CuT&Tag analysis was performed taking into account chromatin openness data. A list of genes whose expression is potentially regulated by quadruplex interactions was obtained.