Association of Polygenic Risk Scores for Multiple Cancers in a Phenome-wide Study: Results from The Michigan Genomics Initiative

Genomic Prediction of 16 Complex Disease Risks Including Heart Attack, Diabetes, Breast and Prostate Cancer

We construct risk predictors using polygenic scores (PGS) computed from common Single Nucleotide Polymorphisms (SNPs) for a number of complex disease conditions, using L1-penalized regression (also known as LASSO) on case-control data from UK Biobank. Among the disease conditions studied are Hypothyroidism, (Resistant) Hypertension, Type 1 and 2 Diabetes, Breast Cancer, Prostate Cancer, Testicular Cancer, Gallstones, Glaucoma, Gout, Atrial Fibrillation, High Cholesterol, Asthma, Basal Cell Carcinoma, Malignant Melanoma, and Heart Attack. We obtain values for the area under the receiver operating characteristic curves (AUC) in the range ~0.58-0.71 using SNP data alone. Substantially higher predictor AUCs are obtained when incorporating additional variables such as age and sex. Some SNP predictors alone are sufficient to identify outliers (e.g., in the 99th percentile of polygenic score, or PGS) with 3-8 times higher risk than typical individuals. We validate predictors out-of-sample using the eMERGE dataset, and also with different ancestry subgroups within the UK Biobank population. Our results indicate that substantial improvements in predictive power are attainable using training sets with larger case populations. We anticipate rapid improvement in genomic prediction as more case-control data become available for analysis.

Identifying Cancer Targets Based on Machine Learning Methods via Chou's 5-steps Rule and General Pseudo Components

In recent years, the successful implementation of human genome project has made people realize that genetic, environmental and lifestyle factors should be combined together to study cancer due to the complexity and various forms of the disease. The increasing availability and growth rate of 'big data' derived from various omics, opens a new window for study and therapy of cancer. In this paper, we will introduce the application of machine learning methods in handling cancer big data including the use of artificial neural networks, support vector machines, ensemble learning and naïve Bayes classifiers.

Penetrance and Pleiotropy of Polygenic Risk Scores for Schizophrenia in 106,160 Patients Across Four Health Care Systems

OBJECTIVE

Individuals at high risk for schizophrenia may benefit from early intervention, but few validated risk predictors are available. Genetic profiling is one approach to risk stratification that has been extensively validated in research cohorts. The authors sought to test the utility of this approach in clinical settings and to evaluate the broader health consequences of high genetic risk for schizophrenia.

METHODS

The authors used electronic health records for 106,160 patients from four health care systems to evaluate the penetrance and pleiotropy of genetic risk for schizophrenia. Polygenic risk scores (PRSs) for schizophrenia were calculated from summary statistics and tested for association with 1,359 disease categories, including schizophrenia and psychosis, in phenome-wide association studies. Effects were combined through meta-analysis across sites.

RESULTS

PRSs were robustly associated with schizophrenia (odds ratio per standard deviation increase in PRS, 1.55; 95% CI=1.4, 1.7), and patients in the highest risk decile of the PRS distribution had up to 4.6-fold higher odds of schizophrenia compared with those in the bottom decile (95% CI=2.9, 7.3). PRSs were also positively associated with other phenotypes, including anxiety, mood, substance use, neurological, and personality disorders, as well as suicidal behavior, memory loss, and urinary syndromes; they were inversely related to obesity.

CONCLUSIONS

The study demonstrates that an available measure of genetic risk for schizophrenia is robustly associated with schizophrenia in health care settings and has pleiotropic effects on related psychiatric disorders as well as other medical syndromes. The results provide an initial indication of the opportunities and limitations that may arise with the future application of PRS testing in health care systems.

Estimation of DNA contamination and its sources in genotyped samples

Array genotyping is a cost-effective and widely used tool that enables assessment of up to millions of genetic markers in hundreds of thousands of individuals. Genotyping array data are typically highly accurate but sensitive to mixing of DNA samples from multiple individuals before or during genotyping. Contaminated samples can lead to genotyping errors and consequently cause false positive signals or reduce power of association analyses. Here, we propose a new method to identify contaminated samples and the sources of contamination within a genotyping batch. Through analysis of array intensity and genotype data from intentionally mixed samples and 22,366 samples of the Michigan Genomics Initiative, an ongoing biobank-based study, we show that our method can reliably estimate contamination. We also show that identifying sources of contamination can implicate problematic sample processing steps and guide process improvements. Compared to existing methods, our approach can estimate the proportion of contaminating DNA more accurately, eliminate the need for external databases of allele frequencies, and provide contamination estimates that are more robust to the ancestral origin of the contaminating sample.

Meta-MultiSKAT: Multiple phenotype meta-analysis for region-based association test

The power of genetic association analyses can be increased by jointly meta-analyzing multiple correlated phenotypes. Here, we develop a meta-analysis framework, Meta-MultiSKAT, that uses summary statistics to test for association between multiple continuous phenotypes and variants in a region of interest. Our approach models the heterogeneity of effects between studies through a kernel matrix and performs a variance component test for association. Using a genotype kernel, our approach can test for rare-variants and the combined effects of both common and rare-variants. To achieve robust power, within Meta-MultiSKAT, we developed fast and accurate omnibus tests combining different models of genetic effects, functional genomic annotations, multiple correlated phenotypes, and heterogeneity across studies. In addition, Meta-MultiSKAT accommodates situations where studies do not share exactly the same set of phenotypes or have differing correlation patterns among the phenotypes. Simulation studies confirm that Meta-MultiSKAT can maintain the type-I error rate at the exome-wide level of 2.5 × 10^-6 . Further simulations under different models of association show that Meta-MultiSKAT can improve the power of detection from 23% to 38% on average over single phenotype-based meta-analysis approaches. We demonstrate the utility and improved power of Meta-MultiSKAT in the meta-analyses of four white blood cell subtype traits from the Michigan Genomics Initiative (MGI) and SardiNIA studies.

Genome-wide association studies and polygenic risk scores for skin cancer: clinically useful yet?

BACKGROUND

Genome-wide association studies (GWAS) have identified thousands of susceptibility variants, although most have been associated with small individual risk estimates that offer little predictive value. However, combining multiple variants into polygenic risk scores (PRS) may be more informative. Multiple studies have developed PRS composed of GWAS-identified variants for cutaneous cancers. This review highlights data from these studies.

OBJECTIVES

To review published GWAS and PRS studies for melanoma, cutaneous squamous cell carcinoma (cSCC) and basal cell carcinoma (BCC), and discuss their potential clinical utility.

METHODS

We searched PubMed and the National Human Genome Research Institute-European Bioinformatics Institute GWAS catalogue to identify relevant studies.

RESULTS

Results from 21 GWAS (11 melanoma, 3 cSCC, 7 BCC) and 11 PRS studies are summarized. Six loci in pigmentation genes overlap between these three cancers (ASIP/RALY, IRF4, MC1R, OCA2, SLC45A2 and TYR). Additional loci overlap for cSCC/BCC and BCC/melanoma, but no other loci are shared between cSCC and melanoma. PRS for melanoma show roughly two-to-threefold increases in risk and modest improvements in risk prediction (2-7% increases). PRS are associated with twofold and threefold increases in risk of cSCC and BCC, respectively, with small improvements (2% increase) in predictive ability.

CONCLUSIONS

Existing data indicate that PRS may offer small, but potentially meaningful, improvements to risk prediction. Additional research is needed to clarify the potential utility of PRS in cutaneous carcinomas. Clinical translation will require well-powered validation studies incorporating known risk factors to evaluate PRS as tools for screening. What's already known about this topic? Over 50 susceptibility loci for melanoma, basal cell carcinoma (BCC) and cutaneous squamous cell carcinoma (cSCC) have been identified in genome-wide association studies (GWAS). Polygenic risk scores (PRS) using variants identified from GWAS have also been developed for melanoma, BCC and cSCC, and investigated with respect to clinical risk prediction. What does this study add? This review provides an overview of GWAS findings and the potential clinical utility of PRS for melanoma, BCC and cSCC.

A genome-wide association study of prostate cancer in Latinos

Latinos represent <1% of samples analyzed to date in genome‐wide association studies of cancer. The clinical value of genetic information in guiding personalized medicine in populations of non‐European ancestry will require additional discovery and risk locus characterization efforts across populations. In the present study, we performed a GWAS of prostate cancer (PrCa) in 2,820 Latino PrCa cases and 5,293 controls to search for novel PrCa risk loci and to examine the generalizability of known PrCa risk loci in Latino men. We also conducted a genetic admixture‐mapping scan to identify PrCa risk alleles associated with local ancestry. Genome‐wide significant associations were observed with 84 variants all located at the known PrCa risk regions at 8q24 (128.484–128.548) and 10q11.22 (MSMB gene). In admixture mapping, we observed genome‐wide significant associations with local African ancestry at 8q24. Of the 162 established PrCa risk variants that are common in Latino men, 135 (83.3%) had effects that were directionally consistent as previously reported, among which 55 (34.0%) were statistically significant with p < 0.05. A polygenic risk model of the known PrCa risk variants showed that, compared to men with average risk (25th–75th percentile of the polygenic risk score distribution), men in the top 10% had a 3.19‐fold (95% CI: 2.65, 3.84) increased PrCa risk. In conclusion, we found that the known PrCa risk variants can effectively stratify PrCa risk in Latino men. Larger studies in Latino populations will be required to discover and characterize genetic risk variants for PrCa and improve risk stratification for this population.

What's new?

There is strong evidence for a genetic predisposition to prostate cancer (PrCa). Most of this information has come from European ancestry populations, with Latinos representing less than 1% of samples in cancer genome‐wide association studies (GWAS). In this study, the majority of established PrCa risk variants (83.3%) were consistently associated with PrCa risk in Latinos. A polygenic risk score comprised of GWAS‐identified risk variants could identify 10% of Latino men with a ~three‐fold increase in PrCa risk. These findings suggest that common germline variants for PrCa can stratify risk in Latino men, which has implications for targeted screening and prevention.

Exploring various polygenic risk scores for skin cancer in the phenomes of the Michigan genomics initiative and the UK Biobank with a visual catalog: PRSWeb

Polygenic risk scores (PRS) are designed to serve as single summary measures that are easy to construct, condensing information from a large number of genetic variants associated with a disease. They have been used for stratification and prediction of disease risk. The primary focus of this paper is to demonstrate how we can combine PRS and electronic health records data to better understand the shared and unique genetic architecture and etiology of disease subtypes that may be both related and heterogeneous. PRS construction strategies often depend on the purpose of the study, the available data/summary estimates, and the underlying genetic architecture of a disease. We consider several choices for constructing a PRS using data obtained from various publicly-available sources including the UK Biobank and evaluate their abilities to predict not just the primary phenotype but also secondary phenotypes derived from electronic health records (EHR). This study was conducted using data from 30,702 unrelated, genotyped patients of recent European descent from the Michigan Genomics Initiative (MGI), a longitudinal biorepository effort within Michigan Medicine. We examine the three most common skin cancer subtypes in the USA: basal cell carcinoma, cutaneous squamous cell carcinoma, and melanoma. Using these PRS for various skin cancer subtypes, we conduct a phenome-wide association study (PheWAS) within the MGI data to evaluate PRS associations with secondary traits. PheWAS results are then replicated using population-based UK Biobank data and compared across various PRS construction methods. We develop an accompanying visual catalog called PRSweb that provides detailed PheWAS results and allows users to directly compare different PRS construction methods.

Sex-specific and pleiotropic effects underlying kidney function identified from GWAS meta-analysis

Chronic kidney disease (CKD) is a growing health burden currently affecting 10–15% of adults worldwide. Estimated glomerular filtration rate (eGFR) as a marker of kidney function is commonly used to diagnose CKD. We analyze eGFR data from the Nord-Trøndelag Health Study and Michigan Genomics Initiative and perform a GWAS meta-analysis with public summary statistics, more than doubling the sample size of previous meta-analyses. We identify 147 loci (53 novel) associated with eGFR, including genes involved in transcriptional regulation, kidney development, cellular signaling, metabolism, and solute transport. Additionally, sex-stratified analysis identifies one locus with more significant effects in women than men. Using genetic risk scores constructed from these eGFR meta-analysis results, we show that associated variants are generally predictive of CKD with only modest improvements in detection compared with other known clinical risk factors. Collectively, these results yield additional insight into the genetic factors underlying kidney function and progression to CKD.

Estimated glomerular filtration rate (eGFR) is a measure of kidney function and used to characterize chronic kidney disease. Here, Graham et al. identify 53 novel loci for eGFR in a GWAS meta-analysis, a subset of which are associated with other common diseases, such as diabetes and hypertension, based on PheWAS.

Keratinocyte Carcinomas: Current Concepts and Future Research Priorities

Cutaneous squamous cell carcinoma (cSCC) and basal cell carcinoma (BCC) are keratinocyte carcinomas, the most frequently diagnosed cancers in fair-skinned populations. Ultraviolet radiation (UVR) is the main driving carcinogen for these tumors, but immunosuppression, pigmentary factors, and aging are also risk factors. Scientific discoveries have improved the understanding of the role of human papillomaviruses (HPV) in cSCC as well as the skin microbiome and a compromised immune system in the development of both cSCC and BCC. Genomic analyses have uncovered genetic risk variants, high-risk susceptibility genes, and somatic events that underlie common pathways important in keratinocyte carcinoma tumorigenesis and tumor characteristics that have enabled development of prediction models for early identification of high-risk individuals. Advances in chemoprevention in high-risk individuals and progress in targeted and immune-based treatment approaches have the potential to decrease the morbidity and mortality associated with these tumors. As the incidence and prevalence of keratinocyte carcinoma continue to increase, strategies for prevention, including effective sun-protective behavior, educational interventions, and reduction of tanning bed access and usage, are essential. Gaps in our knowledge requiring additional research to reduce the high morbidity and costs associated with keratinocyte carcinoma include better understanding of factors leading to more aggressive tumors, the roles of microbiome and HPV infection, prediction of response to therapies including immune checkpoint blockade, and how to tailor both prevention and treatment to individual risk factors and needs.

Systematic evaluation of cancer-specific genetic risk score for 11 types of cancer in The Cancer Genome Atlas and Electronic Medical Records and Genomics cohorts

BACKGROUND

Genetic risk score (GRS) is an odds ratio (OR)-weighted and population-standardized method for measuring cumulative effect of multiple risk-associated single nucleotide polymorphisms (SNPs). We hypothesize that GRS is a valid tool for risk assessment of most common cancers.

METHODS

Utilizing genotype and phenotype data from The Cancer Genome Atlas (TCGA) and Electronic Medical Records and Genomics (eMERGE), we tested 11 cancer-specific GRSs (bladder, breast, colorectal, glioma, lung, melanoma, ovarian, pancreatic, prostate, renal, and thyroid cancer) for association with the respective cancer type. Cancer-specific GRSs were calculated, for the first time in these cohorts, based on previously published risk-associated SNPs using the Caucasian subjects in these two cohorts.

RESULTS

Mean cancer-specific GRS in the population controls of eMERGE approximated the expected value of 1.00 (between 0.98 and 1.02) for all 11 types of cancer. Mean cancer-specific GRS was consistently higher in respective cancer patients than controls for all 11 types of cancer (P < 0.05). When subjects were categorized into low-, average-, and high-risk groups based on cancer-specific GRS (<0.5, 0.5-1.5, and >1.5, respectively), significant dose-response associations of higher cancer-specific GRS with higher OR of respective type of cancer were found for nine types of cancer (P_-trend  < 0.05). More than 64% subjects in the population controls of eMERGE can be classified as high risk for at least one type of these cancers.

CONCLUSION

Validity of GRS for predicting cancer risk is demonstrated for most types of cancer. If confirmed in larger studies, cancer-specific GRS may have the potential for developing personalized cancer screening strategy.

Methods for the Analysis and Interpretation for Rare Variants Associated with Complex Traits

With the advent of Next Generation Sequencing (NGS) technologies, whole genome and whole exome DNA sequencing has become affordable for routine genetic studies. Coupled with improved genotyping arrays and genotype imputation methodologies, it is increasingly feasible to obtain rare genetic variant information in large datasets. Such datasets allow researchers to gain a more complete understanding of the genetic architecture of complex traits caused by rare variants. State-of-the-art statistical methods for the statistical genetics analysis of sequence-based association, including efficient algorithms for association analysis in biobank-scale datasets, gene-association tests, meta-analysis, fine mapping methods that integrate functional genomic dataset, and phenome-wide association studies (PheWAS), are reviewed here. These methods are expected to be highly useful for next generation statistical genetics analysis in the era of precision medicine. © 2019 by John Wiley & Sons, Inc.

Risk-modeling of dog osteosarcoma genome scans shows individuals with Mendelian-level polygenic risk are common

Background

Despite the tremendous therapeutic advances that have stemmed from somatic oncogenetics, survival of some cancers has not improved in 50 years. Osteosarcoma still has a 5-year survival rate of 66%. We propose the natural canine osteosarcoma model can change that: it is extremely similar to the human condition, except for being highly heritable and having a dramatically higher incidence. Here we reanalyze published genome scans of osteosarcoma in three frequently-affected dog breeds and report entirely new understandings with immediate translational indications.

Results

First, meta-analysis revealed association near FGF9, which has strong biological and therapeutic relevance. Secondly, risk-modeling by multiple logistic regression shows 22 of the 34 associated loci contribute to risk and eight have large effect sizes. We validated the Greyhound stepwise model in our own, independent, case-control cohort. Lastly, we updated the gene annotation from approximately 50 genes to 175, and prioritized those using cross-species genomics data. Mostly positional evidence suggests 13 genes are likely to be associated with mapped risk (including MTMR9, EWSR1 retrogene, TANGO2 and FGF9). Previous annotation included seven of those 13 and prioritized four by pathway enrichment. Ten of our 13 priority genes are in loci that contribute to risk modeling and thus can be studied epidemiologically and translationally in pet dogs. Other new candidates include MYCN, SVIL and MIR100HG.

Conclusions

Polygenic osteosarcoma-risk commonly rises to Mendelian-levels in some dog breeds. This justifies caninized animal models and targeted clinical trials in pet dogs (e.g., using CDK4/6 and FGFR1/2 inhibitors).

Electronic supplementary material

The online version of this article (10.1186/s12864-019-5531-6) contains supplementary material, which is available to authorized users.

An atlas of polygenic risk score associations to highlight putative causal relationships across the human phenome

The age of large-scale genome-wide association studies (GWAS) has provided us with an unprecedented opportunity to evaluate the genetic liability of complex disease using polygenic risk scores (PRS). In this study, we have analysed 162 PRS (p<5×10-05) derived from GWAS and 551 heritable traits from the UK Biobank study (N = 334,398). Findings can be investigated using a web application (http:‌//‌mrcieu.‌mrsoftware.org/‌PRS‌_atlas/), which we envisage will help uncover both known and novel mechanisms which contribute towards disease susceptibility. To demonstrate this, we have investigated the results from a phenome-wide evaluation of schizophrenia genetic liability. Amongst findings were inverse associations with measures of cognitive function which extensive follow-up analyses using Mendelian randomization (MR) provided evidence of a causal relationship. We have also investigated the effect of multiple risk factors on disease using mediation and multivariable MR frameworks. Our atlas provides a resource for future endeavours seeking to unravel the causal determinants of complex disease.

Elucidating the Underlying Functional Mechanisms of Breast Cancer Susceptibility Through Post-GWAS Analyses

Genome-wide association studies (GWAS) have identified more than 170 single nucleotide polymorphisms (SNPs) associated with the susceptibility to breast cancer. Together, these SNPs explain 18% of the familial relative risk, which is estimated to be nearly half of the total familial breast cancer risk that is collectively explained by low-risk susceptibility alleles. An important aspect of this success has been the access to large sample sizes through collaborative efforts within the Breast Cancer Association Consortium (BCAC), but also collaborations between cancer association consortia. Despite these achievements, however, understanding of each variant's underlying mechanism and how these SNPs predispose women to breast cancer remains limited and represents a major challenge in the field, particularly since the vast majority of the GWAS-identified SNPs are located in non-coding regions of the genome and are merely tags for the causal variants. In recent years, fine-scale mapping studies followed by functional evaluation of putative causal variants have begun to elucidate the biological function of several GWAS-identified variants. In this review, we discuss the findings and lessons learned from these post-GWAS analyses of 22 risk loci. Identifying the true causal variants underlying breast cancer susceptibility and their function not only provides better estimates of the explained familial relative risk thereby improving polygenetic risk scores (PRSs), it also increases our understanding of the biological mechanisms responsible for causing susceptibility to breast cancer. This will facilitate the identification of further breast cancer risk alleles and the development of preventive medicine for those women at increased risk for developing the disease.

Biobank-driven genomic discovery yields new insight into atrial fibrillation biology

To identify genetic variation underlying atrial fibrillation, the most common cardiac arrhythmia, we performed a genome-wide association study of >1,000,000 people, including 60,620 atrial fibrillation cases and 970,216 controls. We identified 142 independent risk variants at 111 loci and prioritized 151 functional candidate genes likely to be involved in atrial fibrillation. Many of the identified risk variants fall near genes where more deleterious mutations have been reported to cause serious heart defects in humans (GATA4, MYH6, NKX2-5, PITX2, TBX5)¹, or near genes important for striated muscle function and integrity (for example, CFL2, MYH7, PKP2, RBM20, SGCG, SSPN). Pathway and functional enrichment analyses also suggested that many of the putative atrial fibrillation genes act via cardiac structural remodeling, potentially in the form of an 'atrial cardiomyopathy'², either during fetal heart development or as a response to stress in the adult heart.