This was calculated using the –indep option in PLINK that prunes SNPs based on the variance inflation factor, VIF (http://pngu.mgh.harvard.edu/~purcell/plink/summary.shtml#prune; default parameters were used). The genotypes of the CEU population from HapMap version 19 were used, since the GWA samples analyzed in this work are of European descendent. This yielded 310,399 independent autosomal SNPs; (4) Number of recombination hotspots spanning gene g per kb, hg. Recombination hotspot positions were taken from [56]; (5) Genetic distance of each gene g, cg in centi-Morgan (cM) per kb units calculated based on a fine-scale map of recombination rates [56]; and (6) Linkage disequilibrium units (LDU) per kb for each gene g, lg, calculated based on an LDU map downloaded from http://cedar.genetics.soton.ac.uk/pub/PROGRAMS/LDMAP [57]. All variables were calculated based on the extended gene boundaries. All variables but gene size, dg were transformed to ‘per kilobase’ units: variables ng, ug, and hg, were divided by dg, and variables cg and lg were divided by the physical distance between the most extreme genetic markers within the gene boundaries for which genetic distance or LDU data were available. All six variables showed a significant correlation with for all genes g, using 1,000 DGI study permutations, both before and
