Performance of a blockwise approach in variable selection using linkage disequilibrium information.

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) face high-dimensional regression challenges due to numerous single nucleotide polymorphisms (SNPs) and linkage disequilibrium (LD).
• Single marker analyses (SMA) may miss causal variants within LD blocks.
• Existing methods struggle with the complex dependency structure of genetic data.

Purpose of the Study:

• To develop a novel three-step approach leveraging LD structure for improved variant identification in GWAS.
• To identify common variants potentially missed by SMA.
• To compare the proposed method against established regression techniques.

Main Methods:

• Hierarchical clustering of SNPs based on LD to define groups.
• Model selection to identify LD blocks from the clustering hierarchy.
• Group Lasso regression applied to the inferred LD blocks.

Main Results:

• The proposed method outperforms state-of-the-art approaches when multiple causal SNPs exist within an LD block.
• Demonstrated relevance and robustness on semi-simulated and real HIV genetic data.
• The method is available as the R package BALD (Blockwise Approach using Linkage Disequilibrium).

Conclusions:

• The blockwise approach effectively infers LD block structure and identifies significant SNPs.
• Tailored integration of biological knowledge (LD structure) improves high-dimensional genomic studies.
• The method enhances the efficiency of GWAS by considering SNP groupings.