ABSTRACT
BACKGROUND: Recent advances in the detection and treatment of prostate cancer (PCa) have reduced morbidity and mortality from this common cancer. Despite these improvements, PCa remains the second leading cause of cancer death in men in the United States. Further understanding of the genetic underpinnings of lethal PCa is required to drive risk detection and prevention and ultimately reduce mortality. We therefore set out to identify germline variants associated with cases of lethal prostate cancer (LPCa). METHODS: Using a two-stage study design, we compared whole-exome sequencing data of 550 LPCa patients to 488 healthy male controls. Men were classified as having LPCa based on medical record review. Candidate genes were identified using gene- and gene-set-based rare truncating variant association tests. Case-control burden testing through Firth's penalized logistic regression and case-gnomAD allelic burden testing through a one-sided mid-p Fisher's exact test were conducted. Each gene's p-values from these tests were combined into an omnibus p-value for candidate gene selection. In the subsequent validation stage, genes were assessed using the UK Biobank and Firth's penalized logistic regression for each ancestry, combined through meta-analysis. RESULTS: Gene-based rare variant association tests identified 12 genes nominally associated with LPCa. Rare-variant association tests identified a gene set with a significantly higher burden of truncating germline mutations in LPCa patients than controls. Combining gene- and gene-set test results, four nominally significant genes (PPP1R3A, TG, PPFIBP2, and BTN3A3) were selected as candidates. Subsequent validation using the UK Biobank found that PPP1R3A was significantly associated with LPCa risk (odds ratio 2.34, CI 1.20-4.59). Specifically, pGln662ArgfsTer7 was identified as the predominant variant in PPP1R3A among LPCa patients in our dataset. CONCLUSIONS: Both individual gene and gene-set analyses identified candidates associated with LPCa. The novel association of PPP1R3A and LPCa risk merits further investigation.
ABSTRACT
BACKGROUND: Abdominal aortic aneurysm (AAA) is a complex disease with environmental and genetic risk factors. Polygenic risk scores (PRSs) based on disease-specific risk-associated single nucleotide variants (SNVs) have demonstrated effectiveness in stratifying individual-level disease risk for cardiovascular diseases. This prospective cohort study assessed associations of PRS of AAA (PRSAAA) with risk of incident AAA, analyzed the effectiveness of a combined clinical-genetic risk model, and explored the clinical utility of the model in identifying high-risk individuals for AAA screening. METHODS: PRSAAA was calculated using 911,440 SNVs and PRS of coronary artery disease was calculated using 2,324,683 SNVs derived from mixed ancestry genome-wide association studies. The UK Biobank was used as the study cohort. All individuals with complete genetic data available and no diagnosis of AAA at the time of recruitment were included in the analysis and followed prospectively to assess for incident AAA. A PRS-informed clinical model, Prob-AAA, was developed using clinically significant variables and PRSAAA. RESULTS: Four hundred eighty-one thousand one hundred 5 individuals were included in the analysis with 2,668 incident AAA cases. Incident AAA increased from 0.30 to 0.93% between the lowest and highest decile of PRSAAA; similarly, severe AAA, requiring surgery and/or presenting with rupture, increased from 23 to 39% of incident AAA cases across deciles. PRSAAA was a predictor of incident AAA diagnosis (hazard ratio 2.06 [1.70-2.48]) independent of other clinical risk factors including male sex, older age, and smoking history. Prob-AAA was an independent predictor of incident AAA (hazard ratio 1.92 [1.69-2.20]), and identified 9.6% of cases of incident AAA compared to only 4.2% by PRSAAA. Current screening guidelines captured 5.7% of the overall cohort, with an incident AAA rate of approximately 3.2%. Among males not included by current guidelines, Prob-AAA identified an additional cohort, approximately 2% of the overall cohort, with a similar rate of incident AAA. CONCLUSIONS: Prob-AAA, a PRS-informed clinical model for AAA, improved upon the predictive power of current, clinical risk factor-informed, screening guidelines for AAA.
ABSTRACT
BACKGROUND: Current risk assessment for ischemic stroke (IS) is limited to clinical variables. We hypothesize that polygenic scores (PGS) of IS (PGSIS) and IS-associated diseases such as atrial fibrillation (AF), venous thromboembolism (VTE), coronary artery disease (CAD), hypertension (HTN), and Type 2 diabetes (T2D) may improve the performance of IS risk assessment. METHODS: Incident IS was followed for 479,476 participants in the UK Biobank who did not have an IS diagnosis prior to the recruitment. Lifestyle variables (obesity, smoking and alcohol) at the time of study recruitment, clinical diagnoses of IS-associated diseases, PGSIS, and five PGSs for IS-associated diseases were tested using the Cox proportional-hazards model. Predictive performance was assessed using the C-statistic and net reclassification index (NRI). RESULTS: During a median average 12.5-year follow-up, 8374 subjects were diagnosed with IS. Known clinical variables (age, gender, clinical diagnoses of IS-associated diseases, obesity, and smoking) and PGSIS were all independently associated with IS (P < 0.001). In addition, PGSIS and each PGS for IS-associated diseases was also independently associated with IS (P < 0.001). Compared to the clinical model, a joint clinical/PGS model improved the C-statistic for predicting IS from 0.71 to 0.73 (P < 0.001) and significantly reclassified IS risk (NRI = 0.017, P < 0.001), and 6.48% of subjects were upgraded from low to high risk. CONCLUSIONS: Adding PGSs of IS and IS-associated diseases to known clinical risk factors statistically improved risk assessment for IS, demonstrating the supplementary value of inherited susceptibility measurement . However, its clinical utility is likely limited due to modest improvements in predictive values.