Kernelized approach enables explainable gene prioritizations for complex traits

Area of Science:

• Genetics and Genomics
• Computational Biology
• Statistical Genetics

Background:

• Genome-wide association studies (GWAS) identify genetic variants linked to traits, but pinpointing causal effector genes within these loci is difficult.
• Existing methods like PoPS prioritize effector genes using functional profiles but lack transparency in their predictions.
• The biological plausibility of single effector gene nominations from GWAS loci remains a challenge.

Purpose of the Study:

• Introduce Kernelized Polygenic Priority Score (K-PoPS), a novel method for interpretable effector gene prioritization in GWAS.
• Enable gene-centric explanations by decomposing predictions into contributions from training genes.
• Quantify support for nominated genes using an anchor score based on user-defined trait-relevant genes.

Main Methods:

• Developed K-PoPS as a kernelized reformulation of the PoPS algorithm.
• Implemented K-PoPS using a full-feature Ordinary Least Squares (OLS) approach.
• Evaluated K-PoPS performance across 38 Pan-UK Biobank traits and 25 traits with curated anchor sets.

Main Results:

• K-PoPS demonstrated improved closest-gene enrichment over default PoPS for 26 of 37 evaluable traits.
• Predictions supported by K-PoPS anchor scores showed greater enrichment for closest-gene proxies compared to unsupported predictions.
• K-PoPS provided convincing explanations for nominating SCARB1 over UBC for blood apolipoprotein B levels and identified multiple plausible genes for dilated cardiomyopathy.

Conclusions:

• K-PoPS offers a powerful post hoc framework for examining and interpreting GWAS effector-gene nominations.
• The method enhances biological plausibility assessment and provides transparent explanations for gene prioritization.
• K-PoPS facilitates a more nuanced understanding of genetic architecture by identifying multiple potential effector genes within GWAS loci.