Summary results for the six SNPs selected for replication in oral cancer GWAS. Ranking was based on the Bayesian False Discovery Probability (BFDP).

a)Total number of cases and controls included in the final GWA analysis (Table S2).
b)Total number of cases and controls included in the replication analysis.
c)Major and minor alleles, with corresponding allele frequencies in controls.
d)OR, 95% CI and p-values were estimated for the per-rare-allele log-additive genetic model by unconditional logistic regression, adjusting for sex and country (see methods).
e)Prior probability of association (prior for the alternative hypothesis H0) based on the ADAPT literature search (see methods).
f)GWAS ranking based on p-values.
g)The Bayesian False Discovery Probability (BFDP) was estimated based on the association results and the prior probability of association (see methods). The point BFDP estimate corresponds to 100 true susceptibility SNPs assumed to be included in the dataset that are evenly distributed across the prior categories. The range refers to a sensitivity analysis of the BFDP by varying the assumed number of true susceptibility SNPs in the dataset. The bottom and upper boundaries were estimated by assuming 500 and 50 true susceptibility SNPs, respectively.
h)GWAS ranking based on BFDP estimates.
i)OR, 95% CI and p-values were estimated for the per-rare-allele log-additive genetic model by unconditional logistic regression, adjusting for sex and study center (see methods).
j)P-heterogeneity indicates differences in OR between the discovery and replication phases, and was derived from the Cochran's Q test.