在All of Us研究计划中，非洲血统男性13q12染色体带上与前列腺癌风险相关的非连续单倍型模式的证据

Genetic ancestry is known to be an important predictor of prostate cancer (PCa) risk. In particular, PCa incidence and mortality can be up to two times higher in men of African ancestry. While polygenic risk scores based on genome-wide significant polymorphisms can stratify the population into well-defined risk groups, a large portion of the genetic risk remains unaccounted for. Admixture mapping is a technique used to determine the likely genetic origin of a segment of chromosome. We hypothesized that an increase or decrease in local African ancestry in PCa cases vs. controls would implicate regions associated with increased risk, and reveal underlying combinatorial patterns of polymorphisms. To this end, we identified 5,246 PCa cases and controls aged 55 years or older in All of Us Research Program (Controlled Tier v. 7), who were of at least 80% global African ancestry, and no two of whom were closer than third-degree relatives. After performing genome-wide admixture mapping using RFMix, we identified a suggestive peak of increased local European ancestry in controls (by 3.2 percentage points) at chromosome band 13q12, near the genes PCOTH and MIPEP (which we previously found to be more highly expressed in prostate tumors in 51 matched samples from The Cancer Genome Atlas). To study the haplotype structure of this region more closely, we extracted haplotypes from the subset of 926 chromosomes that had a haplotype of unambiguous local European origin-associated with decreased PCa risk (OR = 0.70, P = 0.026) compared to haplotypes of non-European origin-and clustered the resulting haplotypes using affinity propagation. Although no haplotype cluster was individually statistically significantly associated with PCa risk in a conditional analysis of carriers of each cluster compared to all other clusters, we found that, after pooling the risk-increasing and risk-decreasing clusters, and comparing the combined clusters to each other, a statistically significant signal could be identified (OR = 3.7, P = 1.5 × 10 −4 ), functional validation of which is being pursued by eQTL analysis in TCGA. The pooled risk-clusters revealed the presence of noncontiguous blocks of variants, suggesting the possibility that combinatorial haplotype patterns could account for increased PCa risk among African-ancestry individuals who locally have a normally protective European-ancestry haplotype. To test this hypothesis, and determine the extent to which higher-order combinatorial interactions can explain additional PCa risk, we are developing a Docker version of our previously published Chromosome Overlap algorithm for use in the All of Us Researcher Workbench, to see if the haplotype patterns we reveal mirror those detected by admixture mapping. The authors of this study gratefully acknowledge All of Us participants, and the NIH All of Us Research Program.