Power comparisons between similarity-based multilocus association methods, logistic regression, and score tests for...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genomic distance-based regression offers a novel approach for multilocus association analyses, utilizing locus or haplotype scoring.
• Its comparative power against traditional methods in case-control studies, particularly concerning various genetic properties, remains under-explored.

Purpose of the Study:

• To compare the statistical power of the genomic distance-based regression approach against traditional methods for case-control association analyses.
• To evaluate how factors like marker informativity, number of markers, allele frequency, haplotype prevalence, and linkage disequilibrium influence the power of different association methods.

Main Methods:

• Compared seven different association methods, including locus-based logistic regression, global score tests for haplotypes, and genomic distance-based regression with locus and haplotype scoring.
• Assessed method power based on five key properties: marker informativity, marker count, causal allele frequency, common high-risk haplotype frequency, and correlation between causal SNPs and flanking markers.

Main Results:

• Traditional methods like locus-based logistic regression and global haplotype score tests lost power with increased marker numbers due to higher degrees of freedom.
• The genomic distance-based approach demonstrated less vulnerability to a larger number of markers or haplotypes.
• A genotype counting measure was sensitive to marker informativity and linkage disequilibrium with the causal SNP.

Conclusions:

• Genomic distance-based regression, particularly using a matching measure for diplotypes, emerged as the most powerful and robust method on average across the evaluated scenarios.
• This method offers a promising alternative for genetic association studies, especially when dealing with complex genetic architectures and large datasets.