rs17036101

This paper applies a bias correction method for odds ratio estimates from GWAS discovery data, demonstrating that the 'winner's curse' affects initial effect size estimates. The authors applied conditional maximum likelihood estimation to correct bias in GWAS findings from multiple complex diseases (breast cancer, colorectal cancer, lung cancer, prostate cancer, type I and II diabetes) and show that bias-adjusted odds ratios are substantially more consistent with subsequent replication studies, with selection-adjusted confidence intervals providing better uncertainty quantification than uncorrected estimates.

Replication study of 7 meta-analysis-derived type 2 diabetes susceptibility SNPs in 6,244 Japanese individuals across 3 independent populations. Only rs864745 in JAZF1 showed nominal association (OR 1.148, 95% CI 1.034-1.275, p=0.0098) but not after Bonferroni correction; other loci did not reach statistical significance, suggesting these European-identified variants have minor or absent effects in Japanese populations.