Comparison of GWAS ranking of validated lung cancer susceptibility SNPs by p-values and BFDPs for known susceptibility loci of lung cancer.

1)Odds ratios were estimated based on the complete dataset (100%).2)Pr(H1) refers to the prior probability of the alternative hypothesis (prior probability of SNP being associated with lung cancer) and was calculated based on the AdAPT webervice being run on Entrez gene riftexts and Pubmed Abstracts, respectively. The priors were calculated in three categories (low/mid/high). See further details on the statistical framework for performing these calculations in the Methods section.3)50% and 75% of data were randomly sampled from the complete dataset 100 times.4)For each randomly sampled sub-dataset we performed logistic regression and estimated odds ratios along with 95% confidence intervals, p-values and approximate bayes factors. These were subsequently used to estimate BFDPs in order to compare the ranking of known susceptibility SNPs of lung cancer using two ranking methods, by p-values and by BFPDs.5)P-values were estimated using logistic regression models.6)Median and mean ranking were based on the results from 100 randomly sampled datasets, Δ indicates the change in ranking compared to p-value based ranking, the range refers to the highest and lowest ranking observed, respectively.7)BFDP (PubMed abstracts) were calculated using priors that were estimated by running the AdAPT web service on Pubmed abstracts (published before January 2008).