Genome-wide association studies (GWAS) hold great promise as a method for the discovery of new host genetic variants that affect susceptibility to HIV infection or AIDS progression. The analysis of a GWAS presents many challenges including the processing of large data sets and the interpretation of the results of large numbers of statistical tests. To facilitate an HIV/AIDS GWAS, the most informative samples were selected from eight HIV/AIDS natural history cohorts and genotypes were ascertained using the Affymetrix Genome-Wide Human SNP Array 6.0. Samples from individual subjects were checked for the proportion of missing genotypes, consistency of sex estimated from genomic data with reported sex, sex chromosomal aneuploidies, and identity or close relatedness to other samples. The proportion of missing genotypes, the minor allele frequency, the Hardy-Weinberg equilibrium P-value, and the Mendel error rate were computed for all single nucleotide polymorphisms (SNPs). Association statistics were computed for 51 categorical analysis tests and 72 survival analysis tests. The hypotheses tested included infection, progression to AIDS outcomes (CD4 <200, AIDS-1993, AIDS-1987, and AIDS-related death), progression to specific AIDS-defining illnesses (sequelae), and efficacy of highly active antiretroviral therapy (HAART). The NIH Biowulf cluster allowed the 700,000 SNPs in the GWAS to be processed in batches running simultaneously in parallel. These results were augmented with two principal components analysis P-values for combined progression hypotheses and six viral load or CD4 trajectory statistics to give a total of 131 tests. Several techniques were developed to visualize the results of these tests. Monotypic (Manhattan) plots display the P-values for all SNPs for a single genetic association test while the ARG-Trax report displays the results of multiple tests for a single SNP. ARG-Array displays the results of over 100 association tests for over 80 SNPs in a single page using a grid of squares whose colors indicate the P-value. In the ARG-Highway plot, the height of blocks indicates the P-values. A red block color indicates that the minor allele is protective and a green block color indicates that the minor allele increases susceptibility or accelerates progression. The intensity of the color indicates the strength of the association. Lighter colors indicate an odds or hazard ratio near one while more intense (darker) colors indicate odds or hazard ratios further away from one. The ARG-Highway plot may be polarized such that when the minor allele of an index SNP tracks the major allele instead of the minor allele of a given SNP, the colors are reversed. All together, these plots allow investigators to make informed decisions as to which observed affects are biological important.
