Stability SCAD: a powerful approach to detect interactions in large-scale genomic study.

Area of Science:

• Genetics and Bioinformatics
• Statistical Genomics

Background:

• Complex diseases may involve interactions between single nucleotide polymorphisms (SNPs).
• Detecting SNP-SNP interactions is challenging in high-dimensional, small-sample genomic studies (small-n-large-p).
• Existing methods like stability least absolute shrinkage and selection operator (SLASSO) can have low power or high false positives.

Purpose of the Study:

• To introduce a novel statistical procedure, stability smoothly clipped absolute deviation (SSCAD), for robust SNP-SNP interaction detection.
• To evaluate the performance of SSCAD against established methods in simulated and real genomic data.

Main Methods:

• SSCAD applies a smoothly clipped absolute deviation (SCAD) algorithm to multiple sub-samples.
• It identifies clusters of interactions across these sub-samples to enhance detection reliability.
• The method was rigorously compared with SLASSO and traditional penalized regression techniques via simulations.

Main Results:

• Intensive simulations demonstrated that SSCAD achieves higher statistical power and a lower false discovery rate (FDR) compared to SLASSO and other penalized methods.
• Application to a genome-wide association study (GWAS) of lung cancer validated SSCAD's efficacy.
• SSCAD successfully identified all interactions previously found by SLASSO and uncovered two novel significant interactions.

Conclusions:

• SSCAD is a powerful and effective procedure for detecting SNP-SNP interactions in large-scale genomic datasets.
• The method offers improved sensitivity and specificity over existing approaches.
• SSCAD holds significant promise for advancing genetic research in complex diseases.