DeepPerVar

DeepPerVar predicts genome-wide quantitative epigenetic signals from paired whole-genome sequencing and epigenetic functional assays to functionally interpret non-coding genetic variants while accounting for personal genetic variation and traits.

Key Features:

• Multimodal deep learning framework: Integrates genomic and epigenomic data using a deep learning architecture to analyze the functional impact of genetic variants.
• Integration of paired whole-genome sequencing and epigenetic functional assays: Leverages paired WGS and epigenetic functional assays within population studies to link genetic variation to epigenetic signals.
• Personalized analysis: Accounts for individual genetic heterogeneity and personal traits in predictions across a population.
• Quantitative epigenetic signal prediction: Predicts genome-wide quantitative epigenetic signals to evaluate the functional consequences of non-coding variants.
• Allelic difference quantification: Quantifies allelic differences in epigenetic signals to assess variant-specific effects.
• Enhanced heritability analysis and GWAS prioritization: Improves partitioning heritability analysis and aids prioritization of putative causal variants within GWAS risk loci.
• Application to disease studies: Identifies key genomic regions, AD causal genes, and learns canonical regulatory motifs relevant to Alzheimer's disease in the ROSMAP cohort.

Scientific Applications:

• Functional interpretation of non-coding variants: Assigns quantitative epigenetic effects to non-coding genetic variants to infer potential regulatory impact.
• Prioritization of GWAS loci: Supports partitioning heritability and prioritizing putative causal variants within GWAS risk loci.
• Study of Alzheimer's disease (AD): Applied to the ROSMAP cohort to predict epigenetic signals, identify AD-associated genomic regions and regulatory motifs linked to AD causal genes.

Methodology:

Uses a multimodal deep learning framework trained on paired whole-genome sequencing and epigenetic functional assays from population studies to predict genome-wide quantitative epigenetic signals, account for personal genetic variation and traits, learn canonical regulatory motifs, and quantify allelic differences.