RegVar

RegVar: Deep neural network for tissue-specific prioritization of noncoding regulatory variants

RegVar applies a deep neural network (DNN) to prioritize tissue-specific noncoding regulatory variants by predicting the regulatory probabilities of single nucleotide polymorphisms (SNPs) on target genes across 17 human tissues.

Key Features:

• Multi-omic Integration: Integrates sequential, epigenetic, and evolutionary genomic profiles of SNPs with potential target genes.
• Tissue-Specific Prediction: Estimates regulatory effects of noncoding variants on gene expression in 17 distinct human tissues.
• Variant–Gene Association Modeling: Learns from large-scale variant-gene expression associations to improve prioritization accuracy.

Scientific Applications:

• Human Genetics and Complex Trait Analysis: Identifies functional noncoding genomic variants and links them to target genes influencing gene expression and disease-associated traits.

Methodology:

RegVar trains a deep neural network on extensive cross-tissue variant-gene expression association data, integrating sequential, epigenetic, and evolutionary features of single nucleotide polymorphisms to model tissue-specific regulatory impacts on target genes.