SCAMPI: A scalable statistical framework for genome-wide interaction testing harnessing cross-trait correlations

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Family-based heritability estimates for complex traits often exceed SNP-based heritability.
• This discrepancy may stem from non-additive genetic effects, such as gene-gene or gene-environment interactions.
• Current variance-based methods efficiently screen for SNP interactions but are limited to single traits.

Purpose of the Study:

• To address the limitations of univariate methods, we introduce SCAMPI (Scalable Cauchy Aggregate test using Multiple Phenotypes to test Interactions).
• SCAMPI is designed to screen for genetic variants exhibiting interaction effects across multiple correlated traits.
• The goal is to enhance the power to detect interaction effects by leveraging pleiotropy.

Main Methods:

• SCAMPI utilizes the observation that SNPs with pleiotropic interaction effects alter trait correlation patterns across genotype categories.
• It employs a computationally scalable approach to analyze these patterns across multiple traits.
• The method does not require pre-specification of the interacting variable.

Main Results:

• SCAMPI demonstrates improved performance compared to traditional univariate variance-based methods in detecting interaction effects.
• Application to UK Biobank data for four lipid-related traits identified multiple gene regions missed by existing methods.
• The method is computationally scalable for large-scale biobank data analysis.

Conclusions:

• SCAMPI offers a powerful and scalable strategy for identifying genetic variants with pleiotropic interaction effects.
• This approach can help bridge the gap between family-based and SNP-based heritability estimates.
• The developed software is publicly available for broader research use.