We downloaded genotype data for bladder cancer from dbGaP (NCI Cancer Genetic Markers of Susceptibility (CGEMS) project; accession number phs000346.v1.p1). We used samples genotyped from both Illumina 1 M chip and 610K chip for our analysis. For quality control, we removed SNPs with missing rates 0.01. We checked Hardy-Weinberg Equilibrium and excluded SNPs with p-value 0.001. SNPs with minor allele frequencies (MAF) were also removed. After quality control, 490,614 SNPs from 3,631 cases and 3,356 controls of European descent were used in the analysis. SNP-level association p-values were calculated for this bladder cancer data using logistic regression by assuming an additive genetic model. We also downloaded the uniform peak files for DHSs in 125 cell lines from the ENCODE project (http://genome.ucsc.edu/cgi-bin/hgFileUi?db=hg19&g=wgEncodeAwgDnaseUniform). Note that the DHSs for these 125 cell lines were identified with a uniform analysis workflow by the ENCODE Consortium; this facilitates fair and unbiased comparison among cell lines as annotation for our GPA model.
