In Methods in molecular biology (Clifton, N.J.) ; h5-index 0.0
We describe a statistical method for prioritizing candidate causal noncoding single nucleotide polymorphisms (SNPs) in regions of the genome that are detected as trait-associated in a population-based genome-wide association study (GWAS). Our method's key step is to combine, within a naïve Bayes-like framework, three quantities for each SNP: (1) the p-value for the association test between the SNP's genotype and the trait; (2) the p-value for the SNP's cis-expression quantitative trait locus (cis-eQTL) association test; and (3) a model-based prediction score for the SNP's potential to be a regulatory SNP (rSNP). The method is flexible with respect to the source of the model-based rSNP prediction score; we demonstrate the method using scores obtained using the previously published machine-learning-based rSNP prediction method, CERENKOV2. Because it requires only the GWAS trait association test p-value for each SNP and not full genotype information, our method is applicable for GWAS secondary analysis in the common situation where only summary data (and not full genotype data) are readily available. We illustrate how the method works in step-by-step fashion.
Ramsey Stephen A, Liu Zheng, Yao Yao, Weeder Benjamin
Bioinformatics, Computational, GWAS, Likelihood, Noncoding, SNP, eQTL, rSNP