Benchmarking of local genetic correlation estimation methods using summary statistics from genome-wide association...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Local genetic correlation analyzes additive genetic effects across traits at specific genomic loci.
• It provides insights into trait genetic similarities beyond whole-genome correlations.
• Existing methods like rho-hess, SUPERGNOVA, and LAVA require comprehensive evaluation.

Purpose of the Study:

• To benchmark and compare the performance of three local genetic correlation methods: rho-hess, SUPERGNOVA, and LAVA.
• To identify practical guidelines for selecting appropriate methods for local genetic correlation analysis.
• To investigate the impact of sample overlap and local linkage disequilibrium (LD) estimation on method performance.

Main Methods:

• Extensive simulations were performed to assess method performance under various conditions.
• Real data analyses were conducted using 31 complex traits to validate simulation findings.
• The study focused on challenges related to sample overlap and external LD panel usage.

Main Results:

• Method performance, including identifying correlated regions and accuracy, is highly sensitive to local LD matrix estimation.
• Simulations indicated potential for incorrectly identifying correlated regions based on LD accuracy.
• Real data analyses confirmed the distinct results produced by different methods in post-GWAS studies.

Conclusions:

• Local genetic correlation estimates are sensitive to the precision of local LD estimation.
• The choice of method significantly impacts results in post-GWAS local correlation studies.
• There is a critical need for continued methodological refinement in local genetic correlation estimation.