Biology subjects

Gankin, D.

Publications and source records attributed to Gankin, D..

3 recordsLinked to original sources

bioRxiv2026

Leveraging multiplicity in biologically informed neural networks to uncover disease heterogeneity

Biologically inspired neural networks (BINNs) embed pathway, ontology, or protein-interaction structure directly into neural networks, promising interpretable disease prediction where hidden nodes map to named biological entities. Yet BINNs have been hard to train at biobank scale, and the reliability of their interpretations remains largely untested. Here we present a fast BINN implementation trained on UK Biobank genotype and plasma proteomics data from about 500,000 individuals across six common diseases. BINNs achieve competitive predictive performance, but we uncover two major limits to their interpretability. First, attribution scores are strongly biased by graph topology, because node degree and layer position influence the scores. Normalization reduces this bias but can weaken enrichment for known disease genes. Second, BINNs show substantial predictive multiplicity, that is, independently trained models with identical architecture and data reach similarly accurate solutions while prioritizing different genes and pathways. Although this multiplicity makes single-model explanations unstable, the range of interpretations can itself reveal disease biology. Across 100 replicate BINNs for type 2 diabetes, we find distinct solution clusters prioritizing either inflammatory or hepatic-metabolic pathways, mirroring known disease heterogeneity. Thus, analyzing the space of BINN explanations can turn multiplicity into a tool for studying complex disease mechanisms.

bioinformatics↗

bioRxiv2026

Unifying the genetic landscape of common and rare diseases via latent neighborhoods

Resolving disease mechanisms and identifying safer drug targets remains challenging due to difficulties in integrating inconsistent sources of genetic evidence. Here, we developed a deep-learning (DL) method for obtaining latent representations of human diseases and other terms by coupling a variational autoencoder (VAE) to network embeddings from graph representation learning on a protein interaction network. We apply this approach to map the landscape of human disease, by systematically integrating evidence for [~]42,000 traits and terms from sources such as clinical reports (ClinVar), genome-wide association studies (GWAS), mouse phenotypes and gene ontology. Similar diseases share their latent neighborhoods with the sources of evidence highlighting different aspects of the same cell biology for its role in disease. By decoding latent neighborhoods, we unify the sources of evidence and integrate across diseases to prioritize genes for common and rare diseases. Through examples for diabetes, hearing loss, and familial long QT syndrome, we illustrate how the methodology can be applied to develop hypotheses for disease mechanisms and to propose experimental models, measurements, targets and potential drugs that may improve diagnosis and treatments.

genetics↗

bioRxiv2023

Species-aware DNA language modeling

The rise of large-scale multi-species genome sequencing projects promises to shed new light on how genomes encode gene regulatory instructions. To this end, new algorithms are needed that can leverage conservation to capture regulatory elements while accounting for their evolution. Here we introduce species-aware DNA language models (LMs), which we trained on more than 800 species spanning over 500 million years of evolution. Investigating their ability to predict masked nucleotides from context, we show that DNA LMs distinguish transcription factor and RNA-binding protein motifs from background non-coding sequence. Owing to their flexibility, DNA LMs capture conserved regulatory elements over much further evolutionary distances than sequence alignment would allow. Remarkably, DNA LMs reconstruct motif instances bound in vivo better than unbound ones and account for the evolution of motif sequences and their positional constraints, showing that these models capture functional high-order sequence and evolutionary context. We further show that species-aware training yields improved sequence representations for endogenous and MPRA-based gene expression prediction, as well as motif discovery. Collectively, these results demonstrate that species-aware DNA language models are a powerful, flexible, and scalable tool to integrate information from large compendia of highly diverged genomes.

bioinformatics↗