Biology subjects

Slobodyanyuk, M.

Publications and source records attributed to Slobodyanyuk, M..

4 recordsLinked to original sources

bioRxiv2025

FlashRNA: An Efficient Model for Regulatory Genomics

Transformer-based genomic sequence-to-function models effectively capture long-range genomic interactions but incur high computational costs due to the quadratic complexity of their self-attention layers. In this work, we introduce FlashRNA, which significantly improves computational and memory efficiency through FlashAttention, advancements in model architecture, and optimized training setup. FlashRNA achieves comparable or slightly improved predictive performance compared to similar sized Borzoi or Flashzoi models, notably without depending on pre-trained weights - a major limitation of Flashzoi. Remarkably, we trained FlashRNA from scratch in one day on a single GPU, significantly accelerating training and inference speed. These improvements can facilitate further developments in models for regulatory genomics by reducing computational cost. We demonstrate this in two downstream applications: 1) we train a large ensemble of 16 FlashRNA models and distill them into a single model to improve performance while maintaining efficiency, and 2) we fine-tune FlashRNA on three prediction tasks - ChIP-seq, RNA half-life, and translation efficiency - achieving performance matching or exceeding state-of-the-art task-specific models. Code: https://github.com/deepgenomics/flashrna

genomics↗

bioRxiv2024

Combinations of genomic alterations and immune microenvironmental features associate with patient survival in multiple cancer types

Oncogenesis and tumor progression are shaped by somatic alterations in the cancer genome and features of the tumor immune microenvironment (TME). How interactions of these two systems influence tumor development and clinical outcomes remains incompletely understood. To address this challenge, we developed the multi-omics analysis framework PACIFIC to systematically integrate genetic cancer drivers and infiltration profiles of immune cells with clinical information. In an analysis of 8500 cancer samples, we report 34 immunogenomic interactions (IGXs) in 13 cancer types in which context-specific combinations of genomic alterations and immune cell activities associate with disease outcomes. Risk associations of IGXs are potentially explained by tumor-intrinsic and microenvironmental metrics of immunogenicity and differential expression of therapeutic targets. In luminal-A breast cancer, MEN1 deletion combined with reduced neutrophils is associated with poor prognosis and deregulation of immune signalling pathways. These findings help elucidate how cancer drivers interact with TME to contribute to tumorigenesis.

cancer biology↗

bioRxiv2023

Directional integration and pathway enrichment analysis for multi-omics data

Omics techniques generate comprehensive profiles of biomolecules in cells and tissues. However, a holistic understanding of the data requires joint multi-omics analyses that are challenging. Here we present DPM, a data fusion method for combining multiple omics datasets using directionality and significance estimates of genes, transcripts, or proteins. DPM allows users to define how the input datasets are expected to interact directionally, reflecting the initial experimental design or regulatory relationships between the datasets. DPM statistically prioritises genes and pathways that change consistently across the datasets, while penalising those violating the constraints. Joint analyses of transcriptomic, proteomic, DNA methylation, and clinical datasets of cancer samples demonstrate how directional integration identifies genes and pathways modulated across omics datasets, highlights those with inconsistent evidence, and reveals candidate biomarkers with prognostic signals in multiple datasets. DPM is implemented in the ActivePathways method and provides a general framework for testing detailed hypotheses in multi-omics data.

bioinformatics↗

bioRxiv2023

Mutational processes of tobacco smoking and APOBEC activity generate protein-truncating mutations in cancer genomes

Mutational signatures represent a footprint of tumor evolution and its endogenous and exogenous mutational processes. However, their functional impact on the proteome remains incompletely understood. We analysed the protein-coding impact of single base substitution signatures in 12,341 cancer genomes from 18 cancer types. Stop-gain mutations (SGMs) were strongly enriched in the signatures of tobacco smoking, APOBEC cytidine deaminases, and reactive oxygen species. These mutational processes affect specific trinucleotide contexts to substitute serine and glutamic acid residues with stop codons. SGMs are enriched in cancer hallmark pathways and tumor suppressors such as TP53, FAT1, and APC. Tobacco-driven SGMs in lung cancer correlate with lifetime smoking history and highlight a preventable determinant of these harmful mutations. Our study exposes SGM expansion as a genetic mechanism by which endogenous and carcinogenic mutational processes contribute to protein loss-of-function, oncogenesis, and tumor heterogeneity, providing potential translational and mechanistic insights.

cancer biology↗