Gennady V. Roshchupkin

The paper presents a pipeline for analyzing image-based complex traits using AI phenotyping, combined-GWAS, and federated learning, applied to human facial data to identify genetic loci associated with facial shape.

The paper develops methods to detect non-linear genetic interactions using visible neural networks and applies them to identify significant epistasis pairs in inflammatory bowel disease.

This study identifies genetic variants influencing miRNA expression and links them to disease associations using genome-wide association studies and Mendelian randomization.

The paper introduces four VAE-based deconfounding frameworks for multi-omics clustering, showing they reduce confounding effects and improve biological clustering.

The paper develops interpretable neural networks for multi-omics data to predict smoking status, age, and LDL levels, showing improved performance and generalizability across cohorts.

The paper develops BRAIN-MAGNET, a convolutional neural network that predicts non-coding regulatory element activity and helps prioritize disease-causing variants in neurogenetic disorders.

The paper introduces mEthAE, an autoencoder for methylation data that reduces dimensions 400-fold while maintaining accuracy and identifies biologically significant CpG sites through perturbation-based interpretation.

GenNet predicts phenotypes from genetic variants using interpretable deep learning, identifying genes and pathways associated with traits like hair color and schizophrenia.

This study identifies genetic variants linked to longitudinal brain structure changes across the lifespan using MRI data from 15,640 individuals.

The paper maps cognitive performance to brain structure using voxel-based morphometry and subcortical shape analysis in a large population sample.

This study identifies genetic loci specific to periventricular and deep white matter hyperintensities, revealing distinct genetic architectures and potential biological mechanisms.

The paper identifies 161 genome-wide significant genetic associations for cortical structure traits across 34 brain regions, highlighting key biological pathways and gene functions.

The paper identifies genetic associations with cortical surface area using PCA and GWAS, linking DAAM1 to visual cortex development through mitochondrial pathways and thalamocortical input.