Haplotype block partitioning as a tool for dimensionality reduction in SNP association studies

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genetics

Background:

• Genome-wide association studies (GWAS) face challenges with high dimensionality due to numerous SNPs.
• Spatial correlation of SNPs along the genome is crucial for biologically interpretable and powerful genetic association analyses.
• Genomic partitioning based on spatial correlation offers a strategy to address dimensionality and improve the identification of disease-related genes.

Purpose of the Study:

• To develop and validate a novel algorithm for clustering contiguous SNPs based on linkage disequilibrium (LD).
• To establish a statistical framework for detecting genetic associations using these SNP partitions.
• To compare the performance of the new method against existing block partitioning algorithms.

Main Methods:

• Developed the MCMC Algorithm To Identify blocks of Linkage DisEquilibrium (MATILDE) for SNP clustering.
• Created a statistical testing framework using partitions as units of analysis.
• Validated the approach using simulations based on HapMap data (region 14q11) with assigned phenotypes.

Main Results:

• MATILDE demonstrated higher accuracy in identifying disease-associated SNPs, particularly for small genetic effects, compared to Haploview and HapBlock.
• The method improves true positive findings and reduces false discoveries.
• The probabilistic approach adapts to study-specific parameters (population, technology, sample size) and provides uncertainty assessments for block boundaries.

Conclusions:

• The proposed adaptive, study-specific block partitioning approach using MATILDE is efficient for guiding the search for disease loci.
• This method is as or more effective than current LD-based approaches in realistic genetic association study scenarios.