On Genetic Correlation Estimation With Summary Statistics From Genome-Wide Association Studies

Area of Science:

• Genetics
• Neuroscience
• Biostatistics

Background:

• Cross-trait polygenic risk score (PRS) methods are popular for assessing genetic correlations using genome-wide association study (GWAS) summary statistics.
• Existing methods exhibit bias, where significant cross-trait PRS explain minimal genetic variance (<1%) in independent tests.

Purpose of the Study:

• To investigate and address the bias phenomenon in cross-trait PRS.
• To develop a corrected PRS estimator for more accurate genetic correlation estimates.

Main Methods:

• Developed a consistent cross-trait PRS estimator to correct asymptotic bias.
• Investigated the impact of single nucleotide polymorphism (SNP) screening via GWAS p-values.
• Analyzed the effect of overlapping samples in GWAS.
• Applied methods to GWAS summary statistics for reaction time and brain imaging features from the Pediatric Imaging, Neurocognition, and Genetics study.

Main Results:

• Raw cross-trait PRS estimators significantly underestimated genetic similarity between cognitive function and brain structures (mean R² = 1.32%).
• The bias-corrected PRS estimators revealed a substantial genetic overlap (mean R² = 22.42%).

Conclusions:

• The proposed bias-corrected PRS method accurately estimates genetic correlations between complex traits.
• This approach uncovers significant genetic overlap between cognitive traits and brain structure, advancing our understanding of their heritability.