Flaw or discovery? Calculating exact p-values for genome-wide association studies in inbred populations

MotivationGenome-wide association studies have been conducted in inbred populations where the sample size is small. The ordinary association p-values and multiple testing correction therefore become questionable, as the detected genetic effect may or may not be due to chance, depending on the minor allele frequency distribution across the genome. Instead of permutation testing, marker-specific false positive rate can be analytically calculated in inbred populations without heterozygotes.\n\nResultsSolutions of exact p-values for genome-wide association studies in inbred populations were derived and implemented. An example is presented to illustrate that the marker-specific experiment-wise p-value varies as the genome-wide minor allele frequency distribution changes. A simulation using real Arabidopsis thaliana genome indicates that the use of exact p-values improves detection power and reduces inflation due to population structure. An analysis of a defense-related case-control phenotype using the exact p-values revealed the causal locus, where markers with higher MAFs had smaller p-values than the top variants with lower MAFs in ordinary genome-wide association analysis.\n\nAvailability and ImplementationProject URL: https://r-forge.r-project.org/projects/statomics/. The R package p.exact: https://r-forge.r-project.org/R/?group_id=2030.\n\nContactxia.shen@ki.se