Deletion of chromosome 6q14–q22 is common in multiple human cancers including prostate cancer, and chromosome 6 transferred into cancer cells induces senescence and reduces cell growth, tumorigenicity and metastasis, indicating the existence of one or more tumor-suppressor genes in 6q. To identify the 6q tumor-suppressor gene, we first narrowed the common region of deletion to a 2.5 Mb interval at 6q14–15. Of the 11 genes located in this minimal deletion region and expressed in normal prostates, only snoRNA U50 was mutated, demonstrated transcriptional downregulation and inhibited colony formation in prostate cancer cells. The mutation, a homozygous 2 bp (TT) deletion, was found in two of 30 prostate cancer cell lines/xenografts and nine of 89 localized prostate cancers (eleven of 119 or 9% cancers). Two of 89 (2%) patients with prostate cancer also showed the same mutation in their germline DNA, but none of 104 cancer-free control men did. The homozygous deletion abolished U50 function in a colony formation assay. Analysis of 1371 prostate cancer cases and 1371 matched control men from a case–control study nested in a prospective cohort showed that, although a germline heterozygous genotype of the deletion was detected in both patients and controls at similar frequencies, the homozygosity of the deletion was significantly associated with clinically significant prostate cancer (odds ratio 2.9; 95% confidence interval 1.17–7.21). These findings establish snoRNA U50 as a reasonable candidate for the 6q tumor-suppressor gene in prostate cancer and likely in other types of cancers.