Weighted False Discovery Rate Control in Large-Scale Multiple Testing.

Area of Science:

• Statistical Inference
• Bioinformatics
• Genomics

Background:

• Large-scale inference often benefits from incorporating prior domain knowledge.
• Traditional multiple testing methods may not fully leverage existing information.
• Weighted testing offers a framework to integrate prior knowledge effectively.

Purpose of the Study:

• To develop and validate weighted multiple testing procedures within a decision-theoretic framework.
• To maximize the expected number of true positives while controlling the weighted false discovery rate.
• To assess the impact of incorporating domain knowledge on interpretability and precision.

Main Methods:

• Development of oracle and data-driven weighted multiple testing procedures.
• Theoretical establishment of asymptotic validity and optimality.
• Simulation studies to evaluate performance against existing methods.
• Application to a genome-wide association study dataset.

Main Results:

• Proposed methods demonstrate asymptotic validity and optimality.
• Simulation studies confirm error rate control and substantial power gains over existing approaches.
• Incorporating domain knowledge via weights enhances interpretability and precision of inference.

Conclusions:

• Weighted multiple testing is an effective strategy for integrating prior domain knowledge in large-scale inference.
• The developed procedures offer improved statistical power and interpretability.
• The approach is particularly valuable for complex analyses such as genome-wide association studies.