We applied creQTL mapping to a study of 52 endometrial tumor samples with genotype and expression data available (NCBI GEO 14860; [27]). We downloaded the raw genotype and normalized microarray data from GSE14860. Genotype data were from Affymetrix 100K SNP chips, and calls were made with the 'crlmm' function from the 'oligo' library (v1.10.0) [32] for Bioconductor (v2.5; [33]) and R (v2.10.0; [34]). Prior to association testing, SNP markers were prefiltered to satisfy the following stringent conditions for each marker: a call confidence value of at least 95% in 51/52 samples, no significant deviations from Hardy-Weinberg equilibrium at p = 0.05 (with 'snpMatrix' v1.8.0 [32]), and at least two different genotypes observed with at least 5 samples for each genotype. This was analogous to requiring a SNP minor allele frequency of 5% if two genotypes were observed, and 14% if all three genotypes were seen. To enrich the microarray data for abundant and strongly-hybridized transcripts, we calculated the variance of each probe and excluded the lowest 50% from our analyses. Microarray data were ranked and normalized to mean 0 and
