Reporting and interpretation in genome-wide association studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Critique of common methods for flagging associations in genome-wide association studies (GWAS).
• P-values require careful calibration and do not account for test power.
• Q-values control the false discovery rate (FDR).

Purpose of the Study:

• To evaluate and advocate for improved statistical methods in GWAS.
• To present the Bayes factor as a superior measure for hypothesis testing in genetic association studies.
• To demonstrate the utility of Bayes factors for data combination and power calculations.

Main Methods:

• Critique of P-values and their limitations in GWAS.
• Introduction of the q-value as a frequentist FDR control method.
• Advocacy and description of Bayes factor calculation and application.

Main Results:

• Bayes factors provide a robust summary of evidence for or against hypotheses.
• A recently proposed Bayes factor calculation method is easily implemented.
• Bayes factors facilitate straightforward data combination across studies and power calculations.

Conclusions:

• Bayes factors and q-values offer complementary information to P-values.
• Combined use of P-values, q-values, and Bayes factors can decrease the rate of non-reproducible findings.
• These methods enhance the reliability of results from genetic association studies.