A novel and robust feature selection method with FDR control for omics-wide association analysis

Area of Science:

• Genomics
• Biostatistics
• Computational Biology

Background:

• Omics-wide association analysis is crucial for understanding human health but faces challenges with high-dimensional data, unknown distributions, and complex relationships.
• Existing methods often rely on restrictive model assumptions and lack robust control over the false discovery rate (FDR).

Purpose of the Study:

• To develop a novel, robust feature selection method for omics data that addresses limitations of existing approaches.
• To enable accurate fine-mapping of risk features while effectively controlling the false positive rate.

Main Methods:

• Application of a single index model to accommodate unknown monotonic link functions and arbitrary distributions of covariates and errors.
• Integration of a rank-based approach with a symmetrized data aggregation strategy for feature selection.
• Theoretical validation and simulation studies to assess method performance.

Main Results:

• The proposed method demonstrates robust and effective performance across various scenarios in simulated data.
• Analysis of real-world omics datasets identified novel risk features not previously reported.
• The method successfully controls the false positive rate in feature selection.

Conclusions:

• The novel feature selection method offers a powerful and flexible tool for omics-wide association studies.
• This approach enhances the reliability of identifying genetic risk factors in complex diseases.
• The method's robustness makes it suitable for diverse and challenging omics datasets.