Genome-wide Analysis of Rare Haplotypes Associated with Breast Cancer Risk: Discovery, Replication, and Generalizability Evaluation

Abstract

AbstractWhile numerous common variants have been linked to breast cancer (BCa) risk, they explain only partially the total BCa heritability. Inference from the Nordic population-based twin data indicates that rare high-risk loci are the chief determinant of BCa risk. Here, we use haplotypes, rather than single variants, to identify rare high-risk loci for BCa. With computationally phased genotypes from 181,034 white British women in the UK Biobank, we conducted a genome-wide haplotype-BCa association analysis using sliding windows of 5-500 consecutive array-genotyped variants. In the discovery stage, haplotype associations with BCa risk were evaluated retrospectively in the pre-study-enrollment portion of data including 5,487 BCa cases. BCa hazard ratios (HRs) for additive haplotypic effects were estimated using Cox regression. Our replication analysis included women free of BCa at enrollment, of whom 3,524 later developed BCa. This two-stage analysis detected 13 rare loci (frequency <1%), each associated with an appreciable BCa risk increase (discovery: HRs=2.84-6.10, P-value<5×10−8; replication: HRs=2.08-5.61, P-value<0.01). In contrast, the variants that formed these rare haplotypes individually exhibited much smaller effects. Functional annotation revealed extensive cis-regulatory DNA elements in BCa-related cells underlying the replicated rare haplotypes. Using phased, imputed genotypes from 30,064 cases and 25,282 controls in the DRIVE OncoArray case-control study, six of the 13 rare-loci associations proved generalizability (odds ratio estimates: 1.48-7.67, P-value<0.05). This study demonstrates the complementary advantage of utilizing rare haplotypes to capture novel risk loci and possible discoveries of more genetic elements contributing to BCa heritability once large, germline whole-genome sequencing data become available.