False discovery rate estimation for stability selection: application to genome-wide association studies.

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Stability Selection is a variable selection algorithm combining penalized regression and subsampling, particularly useful for ultra-high dimensional data.
• Genome-Wide Association Studies (GWAS) require robust methods for variable selection due to the massive number of comparisons.
• Current Stability Selection decision rules often rely on Family-Wise Error Rate (FWER) control, which can be overly conservative.

Purpose of the Study:

• To propose and evaluate an alternative decision rule for Stability Selection in GWAS using the False Discovery Rate (FDR) criterion.
• To compare the performance of the proposed FDR-based method against the traditional FWER-based approach.
• To demonstrate the application of the FDR-based Stability Selection in a real-world GWAS analysis.

Main Methods:

• Implementing Stability Selection with a permutation-based procedure for False Discovery Rate (FDR) estimation.
• Conducting simulation studies with various genetic data correlation structures to assess performance.
• Comparing the FDR method's results against the theoretical upper bound of the Family-Wise Error Rate (FWER).

Main Results:

• The proposed FDR-based procedure is less conservative than the FWER upper bound.
• The FDR method successfully identifies a greater number of true genetic signals.
• The methodology is effectively illustrated on a GWAS of high-density lipoprotein (HDL) levels.

Conclusions:

• The FDR-based decision rule enhances Stability Selection's power in GWAS by being less conservative.
• This approach offers a more liberal yet statistically sound alternative for variable selection in genetic studies.
• The proposed method provides a valuable tool for identifying genetic associations in large-scale datasets like GWAS.