Discovering non-additive heritability using additive GWAS summary statistics

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• LD score regression (LDSC) estimates heritability from GWAS summary statistics but does not account for genetic interactions.
• Complex traits are influenced by non-additive genetic effects, including interactions between variants.
• Existing methods may underestimate the total genetic contribution to complex traits.

Purpose of the Study:

• To introduce interaction-LD score (i-LDSC) regression, an extension of LDSC that incorporates genetic interactions.
• To evaluate the performance of i-LDSC in simulations and real-world genetic data.
• To identify genetic variance missed by standard LDSC due to interactions.

Main Methods:

• Developed the interaction-LD score (i-LDSC) regression framework.
• Conducted simulations using various generative models to test i-LDSC.
• Applied i-LDSC to analyze 25 quantitative phenotypes from UK Biobank (349,468 individuals) and BioBank Japan (up to 159,095 individuals).

Main Results:

• Simulations demonstrated i-LDSC's ability to recover genetic variance from interactions.
• Re-analysis of 25 traits revealed that i-LDSC detects additional genetic variation not captured by standard LDSC.
• The inclusion of cis-interaction scores significantly improved heritability estimates for all analyzed traits.

Conclusions:

• i-LDSC regression is a powerful extension of LDSC for estimating heritability, particularly when considering genetic interactions.
• The i-LDSC software successfully identifies contributions from non-additive genetic effects in large-scale biobank data.
• This method advances the understanding of the genetic architecture of complex traits by accounting for variant interactions.