Identification of disease-associated loci using machine learning for genotype and network data integration

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome-Wide Association Studies (GWAS) often miss loci with small effect sizes, contributing to the 'missing heritability' in complex diseases.
• Integrating diverse omics data (genotype, phenotype, gene networks) is crucial for identifying comprehensive biological signatures.
• Novel machine learning approaches are needed to merge multi-omics data for improved locus prioritization.

Purpose of the Study:

• To develop and validate a novel integrative algorithm, cNMTF (corrected non-negative matrix tri-factorization), for identifying loci-trait associations.
• To enhance the prioritization of weak genetic loci by merging genotype data with other omics information.
• To improve the understanding of complex diseases and facilitate personalized medicine.

Main Methods:

• Developed cNMTF, an integrative algorithm utilizing clustering techniques for biological data analysis.
• Assessed inter-relationships between genotypes, phenotypes, variant effects, and gene networks.
• Applied cNMTF to prioritize genes associated with lipid traits in two population cohorts, accounting for population structure and ancestry.

Main Results:

• cNMTF successfully prioritized genes associated with lipid traits, replicating 129 known GWAS findings.
• 85% of the identified genes (226/265) were supported by evidence, including novel associations and known lipid regulators.
• The algorithm demonstrated robustness against population structure, effectively accounting for individual ancestry.

Conclusions:

• cNMTF is an effective integrative algorithm for identifying loci-trait associations by merging multi-omics data.
• The method enhances the discovery of weak genetic loci, contributing to understanding complex diseases.
• cNMTF offers a flexible and adaptable approach for personalized medicine research.