The ranking of the p-value of the true causal single nucleotide polymorphism in the ordered list of individual SNP p-values is an important factor for achieving success in the ultimate objective of association studies - identifying deleterious genetic variants. Thus, we undertake a study to assess the implications of complex, multimarker correlation structure, sample size and disease models on the ranking of the causal SNP. We carry out an extensive family-based candidate gene simulation study to analyze the position of the disease susceptibility locus in the complete list of individual SNP p-values ordered according to their statistical significance. We simulate data based on the haplotype distributions of ten randomly selected genes extracted from the HapMap database, various sample sizes (600,1000 and 2000) that current association studies employ, and disease models that mimic the characteristics of complex human disorders. We conclude that the average ranking of the causal SNP for sample sizes 600, 100 and 200 of 10.97, 9.65, and 8.34 are dramatically distant from the most significant and intuitively appropriate top position. This result is even more pronounced for genes with high average correlation and large number of common SNPs. Moreover, the gain of the DSL ranking when comparing sample sizes 600 to 1000 and 1000 to 2000, averaged over disease models, causal SNPs and genes, was approximately 1.3. These outcomes both reveal the importance of the sample size and quantify the magnitude required to unequivocally determine the identity of the DSL in family-based candidate gene studies. Our results show the overwhelming importance of large sample sizes in the localization of deleterious SNPs even under simple disease models. These conclusions possess pronounced importance for the design and result interpretation of candidate gene, next generation high-density genome-wide association studies, as well as for the construction and implementation of association tests based on the distribution of the most significant (minimum p-value) test statistics.
Brown, Lisa A. and Rakovski, Cyril
"On the ranking of the disease susceptibility locus in family-based candidate gene studies: a simulation-based analysis,"
e-Research: A Journal of Undergraduate Work: Vol. 1
, Article 3.
Available at: http://digitalcommons.chapman.edu/e-Research/vol1/iss2/3