Citations

Abstract Polygenic risk scores (PRSs), which quantify inherited susceptibility to complex traits and diseases, have emerged as valuable tools for risk stratification and precision medicine. Despite their promise, PRS developed on European cohorts often demonstrate substantially reduced predictive accuracy in non-European populations, due to differences in genetic architecture. The disproportionate representation of European ancestry cohorts in genome-wide association studies (GWAS) leads to inequitable deployment of PRS technologies across diverse populations. Here, we introduce PRANA (Polygenic Risk Adaptation via Neural-network Architecture), a deep learning framework that adapts an existing PRS developed on one population to other ancestries. Unlike methods that require large-scale GWAS in the target population, PRANA leverages pre-trained PRS models derived from European cohorts and adapts them using modestly sized cohorts from the target population. We evaluated PRANA on seven complex traits in South Asian, East Asian and Ashkenazi Jewish populations, as well as in selected smaller East Asian subpopulations where the scarcity of training data poses a particular challenge. PRANA mostly improved predictive performance of the baseline PRS models by 5%-20% in terms of effect size (β) and Nagelkerke’s R², and, in most cases, outperformed existing cross-ancestry multi-PRS approaches. These results highlight PRANA as a scalable and practical strategy to reduce disparities in genomic risk prediction and advance the equitable application of PRS in diverse populations.

Background: Breast cancer polygenic risk scores (PRS) and traditional risk models (e.g., the Gail model [Gail]) are known to contribute largely independent information, but it is unclear how the overlap varies by ancestry, age, disease type (invasive breast cancer, DCIS), and risk threshold. Methods: In a retrospective case–control study, we evaluated risk prediction performance in 180,398 women (161,849 of European ancestry; 18,549 of Asian ancestry). Odds ratios (ORs) from logistic regression models and the area under the receiver operating characteristic curve (AUC) were estimated. Results: PRS for invasive disease showed a stronger association in younger (<50 years) women (OR = 2.51, AUC = 0.622) than in women ≥ 50 years (OR = 2.06, AUC = 0.653) of European ancestry. PRS performance in Asians was lower (OR range = 1.62–1.64, AUC = 0.551–0.600). Gail performance was modest across groups and poor in younger Asian women (OR = 0.94–0.99, AUC = 0.523–0.533). Age interactions were observed for both PRS (p < 0.001) and Gail (p < 0.001) in Europeans, whereas in Asians, age interaction was observed only for Gail (invasive: p < 0.001; DCIS: p = 0.002). PRS identified more high-risk individuals than Gail in Asian populations, especially ≥50 years, while Gail identified more in Europeans. Overlap between PRS, Gail, and family history was limited at higher thresholds. Calibration analysis, comparing empirical and model-based ROC curves, showed divergence for both PRS and Gail (p < 0.001), which indicates miscalibration. In Europeans, family history and prior biopsies drove Gail discrimination. In younger Asians, age at first live birth was influential. Conclusions: PRS adds value to risk stratification beyond traditional tools, especially in younger women and Asian ancestry populations.

AbstractBackgroundKnown risk alleles for epithelial ovarian cancer (EOC) account for approximately 40% of the heritability for EOC. Copy number variants (CNVs) have not been investigated as EOC risk alleles in a large population cohort.MethodsSingle nucleotide polymorphism array data from 13 071 EOC cases and 17 306 controls of White European ancestry were used to identify CNVs associated with EOC risk using a rare admixture maximum likelihood test for gene burden and a by-probe ratio test. We performed enrichment analysis of CNVs at known EOC risk loci and functional biofeatures in ovarian cancer–related cell types.ResultsWe identified statistically significant risk associations with CNVs at known EOC risk genes; BRCA1 (PEOC = 1.60E-21; OREOC = 8.24), RAD51C (Phigh-grade serous ovarian cancer [HGSOC] = 5.5E-4; odds ratio [OR]HGSOC = 5.74 del), and BRCA2 (PHGSOC = 7.0E-4; ORHGSOC = 3.31 deletion). Four suggestive associations (P &lt; .001) were identified for rare CNVs. Risk-associated CNVs were enriched (P &lt; .05) at known EOC risk loci identified by genome-wide association study. Noncoding CNVs were enriched in active promoters and insulators in EOC-related cell types.ConclusionsCNVs in BRCA1 have been previously reported in smaller studies, but their observed frequency in this large population-based cohort, along with the CNVs observed at BRCA2 and RAD51C gene loci in EOC cases, suggests that these CNVs are potentially pathogenic and may contribute to the spectrum of disease-causing mutations in these genes. CNVs are likely to occur in a wider set of susceptibility regions, with potential implications for clinical genetic testing and disease prevention.

Abstract BackgroundCopy number variants (CNVs) are pervasive in the human genome but potential disease associations with rare CNVs have not been comprehensively assessed in large datasets. We analysed rare CNVs in genes and non-coding regions for 86,788 breast cancer cases and 76,122 controls of European ancestry with genome-wide array data.ResultsGene burden tests detected the strongest association for deletions in BRCA1 (P= 3.7E-18). Nine other genes were associated with a p-value < 0.01 including known susceptibility genes CHEK2 (P= 0.0008), ATM (P= 0.002) and BRCA2 (P= 0.008). Outside the known genes we detected associations with p-values < 0.001 for either overall or subtype-specific breast cancer at nine deletion regions and four duplication regions. Three of the deletion regions were in established common susceptibility loci.ConclusionsThis is the first genome-wide analysis of rare CNVs in a large breast cancer case-control dataset. We detected associations with exonic deletions in established breast cancer susceptibility genes. We also detected suggestive associations with non-coding CNVs in known and novel loci with large effects sizes. Larger sample sizes will be required to reach robust levels of statistical significance.

Thymidylate synthase (TYMS) is a crucial enzyme for DNA synthesis. TYMS expression is regulated by its antisense mRNA, ENOSF1. Disrupted regulation may promote uncontrolled DNA synthesis and tumor growth. We sought to replicate our previously reported association between rs495139 in the TYMS-ENOSF1 3′ gene region and increased risk of mucinous ovarian carcinoma (MOC) in an independent sample. Genotypes from 24,351 controls to 15,000 women with invasive OC, including 665 MOC, were available. We estimated per-allele odds ratios (OR) and 95% confidence intervals (CI) using unconditional logistic regression, and meta-analysis when combining these data with our previous report. The association between rs495139 and MOC was not significant in the independent sample (OR = 1.09; 95% CI = 0.97–1.22; p = 0.15; N = 665 cases). Meta-analysis suggested a weak association (OR = 1.13; 95% CI = 1.03–1.24; p = 0.01; N = 1019 cases). No significant association with risk of other OC histologic types was observed (p = 0.05 for tumor heterogeneity). In expression quantitative trait locus (eQTL) analysis, the rs495139 allele was positively associated with ENOSF1 mRNA expression in normal tissues of the gastrointestinal system, particularly esophageal mucosa (r = 0.51, p = 1.7 × 10−28), and nonsignificantly in five MOC tumors. The association results, along with inconclusive tumor eQTL findings, suggest that a true effect of rs495139 might be small.

Background The rarity of mutations in PALB2, CHEK2 and ATM make it difficult to estimate precisely associated cancer risks. Population-based family studies have provided evidence that at least some of these mutations are associated with breast cancer risk as high as those associated with rare BRCA2 mutations. We aimed to estimate the relative risks associated with specific rare variants in PALB2, CHEK2 and ATM via a multicentre case-control study. Methods We genotyped 10 rare mutations using the custom iCOGS array: PALB2 c.1592delT, c.2816T>G and c.3113G>A, CHEK2 c.349A>G, c.538C>T, c.715G>A, c.1036C>T, c.1312G>T, and c.1343T>G and ATM c.7271T>G. We assessed associations with breast cancer risk (42 671 cases and 42 164 controls), as well as prostate (22 301 cases and 22 320 controls) and ovarian (14 542 cases and 23 491 controls) cancer risk, for each variant. Results For European women, strong evidence of association with breast cancer risk was observed for PALB2 c.1592delT OR 3.44 (95% CI 1.39 to 8.52, p=7.1×10−5), PALB2 c.3113G>A OR 4.21 (95% CI 1.84 to 9.60, p=6.9×10−8) and ATM c.7271T>G OR 11.0 (95% CI 1.42 to 85.7, p=0.0012). We also found evidence of association with breast cancer risk for three variants in CHEK2, c.349A>G OR 2.26 (95% CI 1.29 to 3.95), c.1036C>T OR 5.06 (95% CI 1.09 to 23.5) and c.538C>T OR 1.33 (95% CI 1.05 to 1.67) (p≤0.017). Evidence for prostate cancer risk was observed for CHEK2 c.1343T>G OR 3.03 (95% CI 1.53 to 6.03, p=0.0006) for African men and CHEK2 c.1312G>T OR 2.21 (95% CI 1.06 to 4.63, p=0.030) for European men. No evidence of association with ovarian cancer was found for any of these variants. Conclusions This report adds to accumulating evidence that at least some variants in these genes are associated with an increased risk of breast cancer that is clinically important.

ABSTRACTEpithelial‐mesenchymal transition (EMT) is a process whereby epithelial cells assume mesenchymal characteristics to facilitate cancer metastasis. However, EMT also contributes to the initiation and development of primary tumors. Prior studies that explored the hypothesis that EMT gene variants contribute to epithelial ovarian carcinoma (EOC) risk have been based on small sample sizes and none have sought replication in an independent population. We screened 15,816 single‐nucleotide polymorphisms (SNPs) in 296 genes in a discovery phase using data from a genome‐wide association study of EOC among women of European ancestry (1,947 cases and 2,009 controls) and identified 793 variants in 278 EMT‐related genes that were nominally (P < 0.05) associated with invasive EOC. These SNPs were then genotyped in a larger study of 14,525 invasive‐cancer patients and 23,447 controls. A P‐value <0.05 and a false discovery rate (FDR) <0.2 were considered statistically significant. In the larger dataset, GPC6/GPC5 rs17702471 was associated with the endometrioid subtype among Caucasians (odds ratio (OR) = 1.16, 95% CI = 1.07–1.25, P = 0.0003, FDR = 0.19), whereas F8 rs7053448 (OR = 1.69, 95% CI = 1.27–2.24, P = 0.0003, FDR = 0.12), F8 rs7058826 (OR = 1.69, 95% CI = 1.27–2.24, P = 0.0003, FDR = 0.12), and CAPN13 rs1983383 (OR = 0.79, 95% CI = 0.69–0.90, P = 0.0005, FDR = 0.12) were associated with combined invasive EOC among Asians. In silico functional analyses revealed that GPC6/GPC5 rs17702471 coincided with DNA regulatory elements. These results suggest that EMT gene variants do not appear to play a significant role in the susceptibility to EOC.

Abstract Background: Genome-wide association studies (GWAS) have so far reported 12 loci associated with serous epithelial ovarian cancer (EOC) risk. We hypothesized that some of these loci function through nearby transcription factor (TF) genes and that putative target genes of these TFs as identified by coexpression may also be enriched for additional EOC risk associations. Methods: We selected TF genes within 1 Mb of the top signal at the 12 genome-wide significant risk loci. Mutual information, a form of correlation, was used to build networks of genes strongly coexpressed with each selected TF gene in the unified microarray dataset of 489 serous EOC tumors from The Cancer Genome Atlas. Genes represented in this dataset were subsequently ranked using a gene-level test based on results for germline SNPs from a serous EOC GWAS meta-analysis (2,196 cases/4,396 controls). Results: Gene set enrichment analysis identified six networks centered on TF genes (HOXB2, HOXB5, HOXB6, HOXB7 at 17q21.32 and HOXD1, HOXD3 at 2q31) that were significantly enriched for genes from the risk-associated end of the ranked list (P &lt; 0.05 and FDR &lt; 0.05). These results were replicated (P &lt; 0.05) using an independent association study (7,035 cases/21,693 controls). Genes underlying enrichment in the six networks were pooled into a combined network. Conclusion: We identified a HOX-centric network associated with serous EOC risk containing several genes with known or emerging roles in serous EOC development. Impact: Network analysis integrating large, context-specific datasets has the potential to offer mechanistic insights into cancer susceptibility and prioritize genes for experimental characterization. Cancer Epidemiol Biomarkers Prev; 24(10); 1574–84. ©2015 AACR.