ComPaSS-GWAS: A method to reduce type I error in genome-wide association studies when replication data are not...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Traditional genome-wide association studies (GWAS) rely on replication in independent datasets, which can be problematic due to phenotype comparability issues or unavailability of replication samples.
• Secondary phenotypes often lack comparability across studies, hindering traditional replication efforts.
• The frequency or nature of certain phenotypes may preclude the availability of a dedicated replication sample.

Purpose of the Study:

• To propose Complementary Pairs Stability Selection for Genome-Wide Association Study (ComPaSS-GWAS) as an alternative to traditional replication.
• To evaluate the statistical properties, including type I error rate and power, of ComPaSS-GWAS.
• To validate the ComPaSS-GWAS approach using real-world data.

Main Methods:

• ComPaSS-GWAS involves randomly splitting the sample into two conditionally independent halves multiple times (resamples).
• A GWAS is performed on each half within each resample.
• Marker significance is corroborated if its p-value is significant in both halves of a resample, mimicking independent discovery and replication.

Main Results:

• Simulation experiments demonstrated that increasing the number of resamples decreased the type I error rate of ComPaSS-GWAS.
• A small reduction in statistical power was observed with increased resamples, but overall performance was comparable to traditional GWAS.
• Validation using blood levels of vitamin pyridoxal 5'-phosphate from the Trinity Student Study (TSS) showed results consistent with previously published independent replication data.

Conclusions:

• ComPaSS-GWAS provides a viable ad-hoc alternative to replication in GWAS, particularly when traditional replication is not feasible.
• The method effectively controls the type I error rate and maintains comparable power to traditional GWAS.
• The approach is validated and shows consistency with existing replication and functional study findings.