A powerful subset-based method identifies gene set associations and improves interpretation in UK Biobank.

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Gene set association tests offer insights into complex trait genetics beyond single-gene analyses.
• Existing methods lack power when few genes drive the association and cannot pinpoint causal genes.

Purpose of the Study:

• To develop a novel method, Gene set analysis Association Using Sparse Signals (GAUSS), for gene set association analysis using only GWAS summary statistics.
• To enhance the detection of gene set-phenotype associations, especially when only a small fraction of genes are involved.
• To identify specific genes within associated sets that contribute most to the observed association.

Main Methods:

• GAUSS utilizes GWAS summary statistics and pre-computed correlation structures for efficient p-value calculation.
• The method identifies subsets of genes with maximal association evidence within significant gene sets.
• Computational speed is enhanced by leveraging existing correlation structures, avoiding computationally intensive permutations or simulations.

Main Results:

• Simulations show GAUSS controls type 1 error and outperforms existing methods in power, particularly with sparse causal signals.
• Analysis of UK Biobank data identified 13,466 phenotype-gene set associations across 10,679 gene sets and 1,403 phenotypes.
• GAUSS successfully pinpointed an average of 17.2 underlying genes per significant association, demonstrating scalability and effectiveness.

Conclusions:

• GAUSS provides a powerful and computationally efficient approach for gene set association analysis using GWAS summary statistics.
• The method effectively identifies biologically relevant gene sets and the specific genes driving these associations.
• GAUSS advances the understanding of the genetic architecture of complex phenotypes by revealing pathway-level genetic influences.