Interactions with polygenic background impact quantitative traits in the UK Biobank

Area of Science:

• Human Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Association studies identify genetic variants linked to phenotypes, but biological mechanisms remain challenging to elucidate.
• Genes function in complex networks, suggesting genetic interactions (epistasis) are expected but difficult to detect due to vast search spaces and small effect sizes.

Purpose of the Study:

• To develop a powerful statistical method for detecting interactions between single-nucleotide polymorphisms (SNPs) and groups of variants aggregated in polygenic scores (PGS).
• To identify novel genetic interaction networks influencing quantitative human traits.
• To explore functional partitioning of PGSs based on transcription factor binding sites to uncover regulatory interactions.

Main Methods:

• Developed a novel statistical test for SNP-by-PGS interactions applicable to any quantitative trait.
• Applied the method to 97 quantitative phenotypes in UK Biobank European samples.
• Developed methods for refining signals and detecting pairwise SNP-SNP interactions, including functional partitioning of PGSs using the HOCOMOCO database.

Main Results:

• Identified 144 independent interactions affecting 52 traits, including known disease risk variants in genes like APOE, FTO, and TCF7L2.
• Detected 38 pairwise SNP interactions, including known interactions for alkaline phosphatase levels and a novel interaction for eosinophil levels.
• Identified 12 interactions involving functionally partitioned PGSs, including a regulatory interaction between TCF7L2 and KDM2A affecting glycated haemoglobin.

Conclusions:

• The developed method significantly increases power to detect genetic interaction networks.
• The study substantially expands the repertoire of known epistatic effects for human phenotypes.
• Statistical interactions effectively reflect underlying biological interdependencies between genes and regulatory elements.