Varmole

Varmole prioritizes disease-associated genomic variants and genes by embedding multi-omic networks, including gene expression, into a biologically inspired drop-connect deep neural network to link genotype to phenotype and reveal molecular mechanisms.

Key Features:

• Interpretable Deep Learning Methodology: Employs a biologically inspired drop-connect deep neural network model that facilitates interpretation of connections among variants, genes, and phenotypes.
• Multi-Omic Data Integration: Embeds multi-omic networks, such as gene expression data, into the neural architecture to connect variants to individual genes.
• Prioritization of Variants and Genes: Prioritizes genomic variants, genes, and regulatory linkages relevant to disease mechanisms.
• No Prior Feature Selection Required: Learns directly from omics and genotype data without requiring prior feature selection.

Scientific Applications:

• Genome-Wide Association Studies (GWAS): Assists GWAS and population studies in identifying and prioritizing genomic variants linked to human diseases by integrating functional omics with genotype and phenotype data.
• Mechanistic Interpretation of Genotype–Phenotype Associations: Helps elucidate molecular mechanisms and regulatory pathways by revealing functional links between variants and genes.

Methodology:

Embeds multi-omic networks (e.g., gene expression) into a deep neural network using a biologically inspired drop-connect approach to prioritize variants, genes, and regulatory linkages and predict phenotypes from genotypes.