Sparse matrix factorization robust to sample sharing across GWAS reveals interpretable genetic components.

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genetics

Background:

• Complex traits exhibit extensive genetic pleiotropy, where single genetic variants influence multiple phenotypes.
• Existing multi-phenotype analysis methods, like matrix factorization (MF), struggle with confounding from sample overlap in biobank GWAS and produce dense, hard-to-interpret factors.

Purpose of the Study:

• To introduce GLEANR (GWAS latent embeddings accounting for noise and regularization), a novel MF method for detecting sparse genetic factors from GWAS summary statistics.
• To address limitations of existing methods, including sample-sharing confounding and the interpretability of dense factors.

Main Methods:

• GLEANR employs matrix factorization with regularization to estimate a data-driven number of sparse genetic factors.
• The method explicitly accounts for sample sharing between genome-wide association studies (GWAS).
• GLEANR is designed to be robust to confounding and improve the replicability of identified genetic factors.

Main Results:

• Application of GLEANR to 137 UK Biobank GWAS identified 58 sparse genetic factors.
• These factors effectively decompose the genetic architecture of traits, showing distinct signatures of negative selection and polygenicity.
• Identified factors demonstrate enrichment for disease, cell-type, and pathway information, with specific examples in platelet phenotypes.

Conclusions:

• GLEANR provides a robust and interpretable approach for dissecting the genetic underpinnings of complex traits using GWAS summary statistics.
• The method facilitates the discovery of both shared and trait-specific genetic pathways.
• GLEANR enhances the biological interpretation of genetic associations by identifying sparse, meaningful factors.