PheWAS were performed for a total of ∼12,000 variants, including 11,979 missense, 58 nonsense, 68 splice site, 39 frameshift mutations and 276 CNVs The closest marker for each variant from a set of 3,804 genetic markers—each representing a unique haplotype block—was used to represent that variant in the PheWAS. We used 16 expression data sets representing different tissues of the BXD strains to explore the genetic basis of variation at mRNA levels. Similarly, we used 4,236 classic phenotypes from GeneNetwork.org (www.genenetwork.org) to study the association between variants and phenotypes. We calculated the P value of the Pearson correlation between each marker (variant) and 4,236 phenotypes and ∼40,000 transcripts for the expression data. All P values of correlation were calculated as a two-tailed test, and the q values (FDR) were calculated using QVALUE (ref. 53). We used an FDR threshold of 0.01 for associations. The analyses were performed using in-house Python scripts, and the R statistical package.
