An Analysis Pipeline for Genome-wide Association Studies

Abstract

We developed an efficient pipeline to analyze genome-wide association study single nucleotide polymorphism scan results. Perl scripts were used to convert genotypes called using the BRLMM algorithm into a modified PB format. We computed summary statistics characteristic of our case and control populations including allele counts, missing values, heterozygosity, measures of compliance with Hardy-Weinberg equilibrium, and several population difference statistics. In addition, we computed association tests, including exact tests of association for genotypes, alleles, the Cochran-Armitage linear trend test, and dominant, recessive, and overdominant models at every single nucleotide polymorphism (SNP). In addition, pairwise linkage disequilbrium statistics were elaborated, using the command line version of HaploView, which was possible by writing a reformatting script. Additional Perl scripts permit loading the results into a MySQL database conjoined with a Generic Genome Browser (gbrowse) for comprehensive visualization. This browser incorporates a download feature that provides actual case and control genotypes to users in associated genomic regions. Thus, re-analysis “on the fly” is possible for casual browser users from anywhere on the Internet.