Polygenic modeling of genome-wide association studies: an application to prostate and breast cancer

A review of studies examining the association between genetic biomarkers (short tandem repeats and single-nucleotide polymorphisms) and risk of prostate cancer: the need for valid predictive biomarkers

Prostate cancer (PCa) is a challenging polygenic disease because the genes that cause PCa remain largely elusive and are affected by several causal factors. Consequently, research continuously strives to identify a genetic marker which could be used as an indicator to predict the most vulnerable (i.e., predisposed) segments of the population to the disease or for the gene which may be directly responsible for PCa. To enhance the genetic etiology of PCa, this research sought to discover the key studies conducted in this field using data from the main journal publication search engines, as it was hoped that this could shed light on the main research findings from these studies, which in turn could assist in determining these genes or markers. From the research highlighted, the studies primarily used two kinds of markers: short tandem repeats or single-nucleotide polymorphisms. These markers were found to be quite prevalent in all the chromosomes within the research carried out. It also became apparent that the studies differed in both quantity and quality, as well as being conducted in a variety of societies. Links were also determined between the degree and strength of the relationship between these markers and the occurrence of the disease. From the studies identified, most recommended a larger and more diverse survey for the parameters which had not been studied before, as well as an increase in the size of the community (i.e., the population) being studied. This is an indication that work in this field is far from complete, and thus, current research remains committed toward finding genetic markers that can be used clinically for the diagnosis and screening of patients with PCa.

Prediction of inherited genomic susceptibility to 20 common cancer types by a supervised machine-learning method

Prevention and early intervention are the most effective ways of avoiding or minimizing psychological, physical, and financial suffering from cancer. However, such proactive action requires the ability to predict the individual's susceptibility to cancer with a measure of probability. Of the triad of cancer-causing factors (inherited genomic susceptibility, environmental factors, and lifestyle factors), the inherited genomic component may be derivable from the recent public availability of a large body of whole-genome variation data. However, genome-wide association studies have so far showed limited success in predicting the inherited susceptibility to common cancers. We present here a multiple classification approach for predicting individuals' inherited genomic susceptibility to acquire the most likely phenotype among a panel of 20 major common cancer types plus 1 "healthy" type by application of a supervised machine-learning method under competing conditions among the cohorts of the 21 types. This approach suggests that, depending on the phenotypes of 5,919 individuals of "white" ethnic population in this study, (i) the portion of the cohort of a cancer type who acquired the observed type due to mostly inherited genomic susceptibility factors ranges from about 33 to 88% (or its corollary: the portion due to mostly environmental and lifestyle factors ranges from 12 to 67%), and (ii) on an individual level, the method also predicts individuals' inherited genomic susceptibility to acquire the other types ranked with associated probabilities. These probabilities may provide practical information for individuals, heath professionals, and health policymakers related to prevention and/or early intervention of cancer.

Quantifying the Genetic Correlation between Multiple Cancer Types

Background: Many cancers share specific genetic risk factors, including both rare high-penetrance mutations and common SNPs identified through genome-wide association studies (GWAS). However, little is known about the overall shared heritability across cancers. Quantifying the extent to which two distinct cancers share genetic origin will give insights to shared biological mechanisms underlying cancer and inform design for future genetic association studies.Methods: In this study, we estimated the pair-wise genetic correlation between six cancer types (breast, colorectal, lung, ovarian, pancreatic, and prostate) using cancer-specific GWAS summary statistics data based on 66,958 case and 70,665 control subjects of European ancestry. We also estimated genetic correlations between cancers and 14 noncancer diseases and traits.Results: After adjusting for 15 pair-wise genetic correlation tests between cancers, we found significant (P < 0.003) genetic correlations between pancreatic and colorectal cancer (rg = 0.55, P = 0.003), lung and colorectal cancer (rg = 0.31, P = 0.001). We also found suggestive genetic correlations between lung and breast cancer (rg = 0.27, P = 0.009), and colorectal and breast cancer (rg = 0.22, P = 0.01). In contrast, we found no evidence that prostate cancer shared an appreciable proportion of heritability with other cancers. After adjusting for 84 tests studying genetic correlations between cancer types and other traits (Bonferroni-corrected P value: 0.0006), only the genetic correlation between lung cancer and smoking remained significant (rg = 0.41, P = 1.03 × 10-6). We also observed nominally significant genetic correlations between body mass index and all cancers except ovarian cancer.Conclusions: Our results highlight novel genetic correlations and lend support to previous observational studies that have observed links between cancers and risk factors.Impact: This study demonstrates modest genetic correlations between cancers; in particular, breast, colorectal, and lung cancer share some degree of genetic basis. Cancer Epidemiol Biomarkers Prev; 26(9); 1427-35. ©2017 AACR.

Subphenotyping and Classification of Orofacial Clefts: Need for Orofacial Cleft Subphenotyping Calls for Revised Classification

Nonsyndromic orofacial clefting (OFC) describes a range of phenotypes that represent the most common craniofacial birth defects in humans, with an overall birth prevalence of 1:700 live births. Because of the lifelong negative implications on health and well-being associated with OFC and the numbers of people affected, quality research into its etiology, diagnosis, treatment outcomes, and preventative strategies is essential. A range of different methods is used for recording and classifying OFC subphenotypes, one of which is the International Classification of Diseases (ICD) system. However, there is a general perception that research is being hampered by a lack of sensitivity and specificity in grouping those with OFC into subphenotypes, with potential heterogeneity and confounding in epidemiologic, genetic, and genotype-phenotype correlation studies. This article provides a background to the necessity of OFC research, discusses current controversies within cleft subphenotyping, and provides a brief overview of current OFC classifications as well as their limitations. The LAHSHAL classification is described in the context of a potentially useful tool for OFC that could complement the ICD-10/ICD-11 Beta coding systems to become a simply understood, universally accepted, clinically friendly, and research-sensitive instrument. Empowering registries, clinicians, and researchers to use a common classification system would have significant implications for OFC research across the world at a time when accurate subphenotyping is crucial and health care research is becoming increasingly tailored toward the individual.

The PLCO Biorepository: Creating, Maintaining, and Administering a Unique Biospecimen Resource

Inclusion of biospecimens in population-based studies is an integral part of understanding disease etiology, identifying biomarkers and developing prevention and treatment strategies. The Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial collected, processed and stored biospecimens from participants to create a biorepository of specimens which serves as a useful resource for a broad research community. PLCO collected blood samples from consented screening arm participants at six screening rounds and a buccal sample from consented control arm participants. In addition, formalin-fixed paraffin embedded tumor tissue specimens were collected for participants in both arms for selected cancer sites. Collection of biospecimens at multiple timepoints (i.e. serial samples) and prior to cancer diagnosis, paired with rich epidemiologic and screening data, makes the PLCO collection of biospecimens a uniquely valuable resource. As such, access to the PLCO biorepository is granted to investigators by a rigorous scientific review process and guided by a steering committee which is responsible for developing and implementing the biospecimen use policies. Here, we describe the procedures for biospecimen collection, processing, storage, requisition, and distribution, as well as data management employed in PLCO. We also provide examples of how the biospecimens have been used to advance cancer research and describe relevant lessons learned to help inform cohorts wishing to add or modify biospecimen collection.

Shared genetics underlying epidemiological association between endometriosis and ovarian cancer

Epidemiological studies have demonstrated associations between endometriosis and certain histotypes of ovarian cancer, including clear cell, low-grade serous and endometrioid carcinomas. We aimed to determine whether the observed associations might be due to shared genetic aetiology. To address this, we used two endometriosis datasets genotyped on common arrays with full-genome coverage (3194 cases and 7060 controls) and a large ovarian cancer dataset genotyped on the customized Illumina Infinium iSelect (iCOGS) arrays (10 065 cases and 21 663 controls). Previous work has suggested that a large number of genetic variants contribute to endometriosis and ovarian cancer (all histotypes combined) susceptibility. Here, using the iCOGS data, we confirmed polygenic architecture for most histotypes of ovarian cancer. This led us to evaluate if the polygenic effects are shared across diseases. We found evidence for shared genetic risks between endometriosis and all histotypes of ovarian cancer, except for the intestinal mucinous type. Clear cell carcinoma showed the strongest genetic correlation with endometriosis (0.51, 95% CI = 0.18-0.84). Endometrioid and low-grade serous carcinomas had similar correlation coefficients (0.48, 95% CI = 0.07-0.89 and 0.40, 95% CI = 0.05-0.75, respectively). High-grade serous carcinoma, which often arises from the fallopian tubes, showed a weaker genetic correlation with endometriosis (0.25, 95% CI = 0.11-0.39), despite the absence of a known epidemiological association. These results suggest that the epidemiological association between endometriosis and ovarian adenocarcinoma may be attributable to shared genetic susceptibility loci.

Prediction of individual genetic risk to prostate cancer using a polygenic score

BACKGROUND

Polygenic risk scores comprising established susceptibility variants have shown to be informative classifiers for several complex diseases including prostate cancer. For prostate cancer it is unknown if inclusion of genetic markers that have so far not been associated with prostate cancer risk at a genome-wide significant level will improve disease prediction.

METHODS

We built polygenic risk scores in a large training set comprising over 25,000 individuals. Initially 65 established prostate cancer susceptibility variants were selected. After LD pruning additional variants were prioritized based on their association with prostate cancer. Six-fold cross validation was performed to assess genetic risk scores and optimize the number of additional variants to be included. The final model was evaluated in an independent study population including 1,370 cases and 1,239 controls.

RESULTS

The polygenic risk score with 65 established susceptibility variants provided an area under the curve (AUC) of 0.67. Adding an additional 68 novel variants significantly increased the AUC to 0.68 (P = 0.0012) and the net reclassification index with 0.21 (P = 8.5E-08). All novel variants were located in genomic regions established as associated with prostate cancer risk.

CONCLUSIONS

Inclusion of additional genetic variants from established prostate cancer susceptibility regions improves disease prediction.

Most common 'sporadic' cancers have a significant germline genetic component

Common cancers have been demarcated into 'hereditary' or 'sporadic' ('non-hereditary') types historically. Such distinctions initially arose from work identifying rare, highly penetrant germline mutations causing 'hereditary' cancer. While rare mutations are important in particular families, most cases in the general population are 'sporadic'. Twin studies have suggested that many 'sporadic' cancers show little or no heritability. To quantify the role of germline mutations in cancer susceptibility, we applied a method for estimating the importance of common genetic variants (array heritability, h(2)g) to twelve cancer types. The following cancers showed a significant (P < 0.05) array heritability: melanoma USA set h(2)g = 0.19 (95% CI = 0.01-0.37) and Australian set h(2)g = 0.30 (0.10-0.50); pancreatic h(2)g = 0.18 (0.06-0.30); prostate h(2)g = 0.81 (0.32-1); kidney h(2)g = 0.18 (0.04-0.32); ovarian h(2)g = 0.30 (0.18-0.42); esophageal adenocarcinoma h(2)g = 0.24 (0.14-0.34); esophageal squamous cell carcinoma h(2)g = 0.19 (0.07-0.31); endometrial UK set h(2)g = 0.23 (0.01-0.45) and Australian set h(2)g = 0.39 (0.02-0.76). Three cancers showed a positive but non-significant effect: breast h(2) g = 0.13 (0-0.56); gastric h(2)g = 0.11 (0-0.27); lung h(2)g = 0.10 (0-0.24). One cancer showed a small effect: bladder h(2)g = 0.01 (0-0.11). Among these cancers, previous twin studies were only able to show heritability for prostate and breast cancer, but we can now make much stronger statements for several common cancers which emphasize the important role of genetic variants in cancer susceptibility. We have demonstrated that several 'sporadic' cancers have a significant inherited component. Larger genome-wide association studies in these cancers will continue to find more loci, which explain part of the remaining polygenic component.

Inflammation-related genetic variants predict toxicity following definitive radiotherapy for lung cancer

Definitive radiotherapy improves locoregional control and survival in inoperable non-small cell lung cancer (NSCLC) patients. However, radiation-induced toxicities (pneumonitis/esophagitis) are common dose-limiting inflammatory conditions. We therefore conducted a pathway-based analysis to identify inflammation-related SNPs associated with radiation-induced pneumonitis or esophagitis. 11,930 SNPs were genotyped in 201 stage I-III NSCLC patients treated with definitive radiotherapy. Validation was performed in an additional 220 NSCLC cases. After validation, 19 SNPs remained significant. A polygenic risk score (PRS) was generated to summarize the effect from validated SNPs. Significant improvements in discriminative ability were observed by adding the PRS into the clinical/epidemiological variable-based model. We then used 277 lymphoblastoid cell-lines to assess radiation sensitivity and eQTL relationships of the identified SNPs. Three genes (PRKCE,DDX58 and TNFSF7) were associated with radiation sensitivity. We concluded that inflammation-related genetic variants could contribute to the development of radiation-induced toxicities. These loci could assist in predicting those unfavorable events.

Strategy to control type I error increases power to identify genetic variation using the full biological trajectory

Genome-wide association studies have been successful in identifying loci that underlie continuous traits measured at a single time point. To additionally consider continuous traits longitudinally, it is desirable to look at SNP effects at baseline and over time using linear-mixed effects models. Estimation and interpretation of two coefficients in the same model raises concern regarding the optimal control of type I error. To investigate this issue, we calculate type I error and power under an alternative for joint tests, including the two degree of freedom likelihood ratio test, and compare this to single degree of freedom tests for each effect separately at varying alpha levels. We show which joint tests are the optimal way to control the type I error and also illustrate that information can be gained by joint testing in situations where either or both SNP effects are underpowered. We also show that closed form power calculations can approximate simulated power for the case of balanced data, provide reasonable approximations for imbalanced data, but overestimate power for complicated residual error structures. We conclude that a two degree of freedom test is an attractive strategy in a hypothesis-free genome-wide setting and recommend its use for genome-wide studies employing linear-mixed effects models.

Power and predictive accuracy of polygenic risk scores

Polygenic scores have recently been used to summarise genetic effects among an ensemble of markers that do not individually achieve significance in a large-scale association study. Markers are selected using an initial training sample and used to construct a score in an independent replication sample by forming the weighted sum of associated alleles within each subject. Association between a trait and this composite score implies that a genetic signal is present among the selected markers, and the score can then be used for prediction of individual trait values. This approach has been used to obtain evidence of a genetic effect when no single markers are significant, to establish a common genetic basis for related disorders, and to construct risk prediction models. In some cases, however, the desired association or prediction has not been achieved. Here, the power and predictive accuracy of a polygenic score are derived from a quantitative genetics model as a function of the sizes of the two samples, explained genetic variance, selection thresholds for including a marker in the score, and methods for weighting effect sizes in the score. Expressions are derived for quantitative and discrete traits, the latter allowing for case/control sampling. A novel approach to estimating the variance explained by a marker panel is also proposed. It is shown that published studies with significant association of polygenic scores have been well powered, whereas those with negative results can be explained by low sample size. It is also shown that useful levels of prediction may only be approached when predictors are estimated from very large samples, up to an order of magnitude greater than currently available. Therefore, polygenic scores currently have more utility for association testing than predicting complex traits, but prediction will become more feasible as sample sizes continue to grow.

Recently there has been much interest in combining multiple genetic markers into a single score for predicting disease risk. Even if many of the individual markers have no detected effect, the combined score could be a strong predictor of disease. This has allowed researchers to demonstrate that some diseases have a strong genetic basis, even if few actual genes have been identified, and it has also revealed a common genetic basis for distinct diseases. These analyses have so far been performed opportunistically, with mixed results. Here I derive formulae based on the heritability of disease and size of the study, allowing researchers to plan their analyses from a more informed position. I show that discouraging results in some previous studies were due to the low number of subjects studied, but a modest increase in study size would allow more successful analysis. However, I also show that, for genetics to become useful for predicting individual risk of disease, hundreds of thousands of subjects may be needed to estimate the gene effects. This is larger than most existing studies, but will become more common in the near future, so that gene scores will become more useful for predicting disease than has appeared to date.

Exploring the genetic architecture of circulating 25-hydroxyvitamin D

The primary circulating form of vitamin D is 25-hydroxy vitamin D (25(OH)D), a modifiable trait linked with a growing number of chronic diseases. In addition to environmental determinants of 25(OH)D, including dietary sources and skin ultraviolet B (UVB) exposure, twin- and family-based studies suggest that genetics contribute substantially to vitamin D variability with heritability estimates ranging from 43% to 80%. Genome-wide association studies (GWAS) have identified single nucleotide polymorphisms (SNPs) located in four gene regions associated with 25(OH)D. These SNPs collectively explain only a fraction of the heritability in 25(OH)D estimated by twin- and family-based studies. Using 25(OH)D concentrations and GWAS data on 5,575 subjects drawn from five cohorts, we hypothesized that genome-wide data, in the form of (1) a polygenic score comprised of hundreds or thousands of SNPs that do not individually reach GWAS significance, or (2) a linear mixed model for genome-wide complex trait analysis, would explain variance in measured circulating 25(OH)D beyond that explained by known genome-wide significant 25(OH)D-associated SNPs. GWAS identified SNPs explained 5.2% of the variation in circulating 25(OH)D in these samples and there was little evidence additional markers significantly improved predictive ability. On average, a polygenic score comprised of GWAS-identified SNPs explained a larger proportion of variation in circulating 25(OH)D than scores comprised of thousands of SNPs that were on average, nonsignificant. Employing a linear mixed model for genome-wide complex trait analysis explained little additional variability (range 0-22%). The absence of a significant polygenic effect in this relatively large sample suggests an oligogenetic architecture for 25(OH)D.

Pharmacogenetics of aromatase inhibitors

Aromatase inhibitors (AIs) are an important class of endocrine drugs used in the treatment of early and advanced breast cancer in postmenopausal women. A number of studies have taken candidate approaches to assess the role of variants in genes encoding enzymes important in AI metabolism, notably CYP19A1 (aromatase), in AI response. These studies have shown conflicting, but interesting, results suggesting that CYP19A1 variants may be important in both the efficacy and toxicity of AIs. A recent genome-wide association study has identified a variant, creating an estrogen response element in TCL1A, which is associated with an increased risk of the musculoskeletal side effects associated with AI use. As breast cancer incidence increases, predictive biomarkers of response to AIs will become more important to ensure the most effective use of endocrine treatments.

Unidentified genetic variants influence pancreatic cancer risk: an analysis of polygenic susceptibility in the PanScan study

Genome-wide association (GWA) studies have identified several pancreatic cancer (PanCa) susceptibility loci. Methods for assessment of polygenic susceptibility can be employed to detect the collective effect of additional association signals for PanCa. Using data on 492,651 autosomal single nucleotide polymorphisms (SNPs) from the PanScan GWA study (2,857 cases, 2,967 controls), we employed polygenic risk score (PRS) cross-validation (CV) methods to (a) confirm the existence of unidentified association signals, (b) assess the predictive value of PRSs, and (c) assess evidence for polygenic effects in specific genomic locations (genic vs. intergenic). After excluding SNPs in known PanCa susceptibility regions, we constructed PRS models using a training GWA dataset and then tested the model in an independent testing dataset using fourfold CV. We also employed a "power-replication" approach, where power to detect SNP associations was calculated using a training dataset, and power was tested for association with "replication status" in a testing dataset. PRS scores constructed using ≥ 10% of genome-wide SNPs showed significant association with PanCa (P< 0.05) across the majority of CV analyses. Associations were stronger for PRSs restricted to genic SNPs compared to intergenic PRSs. The power-replications approach produced weaker associations that were not significant when restricting to SNPs with low pairwise linkage disequilibrium, whereas PRS results were robust to such restrictions. Although the PRS approach will not dramatically improve PanCa prediction, it provides strong evidence for unidentified association signals for PanCa. Our results suggest that focusing association studies on genic regions and conducting larger GWA studies can reveal additional PanCa susceptibility loci.

Mammographic breast density and breast cancer: evidence of a shared genetic basis

Percent mammographic breast density (PMD) is a strong heritable risk factor for breast cancer. However, the pathways through which this risk is mediated are still unclear. To explore whether PMD and breast cancer have a shared genetic basis, we identified genetic variants most strongly associated with PMD in a published meta-analysis of five genome-wide association studies (GWAS) and used these to construct risk scores for 3,628 breast cancer cases and 5,190 controls from the UK2 GWAS of breast cancer. The signed per-allele effect estimates of single-nucleotide polymorphisms (SNP) were multiplied with the respective allele counts in the individual and summed over all SNPs to derive the risk score for an individual. These scores were included as the exposure variable in a logistic regression model with breast cancer case-control status as the outcome. This analysis was repeated using 10 different cutoff points for the most significant density SNPs (1%-10% representing 5,222-50,899 SNPs). Permutation analysis was also conducted across all 10 cutoff points. The association between risk score and breast cancer was significant for all cutoff points from 3% to 10% of top density SNPs, being most significant for the 6% (2-sided P = 0.002) to 10% (P = 0.001) cutoff points (overall permutation P = 0.003). Women in the top 10% of the risk score distribution had a 31% increased risk of breast cancer [OR = 1.31; 95% confidence interval (CI), 1.08-1.59] compared with women in the bottom 10%. Together, our results show that PMD and breast cancer have a shared genetic basis that is mediated through a large number of common variants.

Evaluation of polygenic risk scores for predicting breast and prostate cancer risk

Recently, polygenic risk scores (PRS) have been shown to be associated with certain complex diseases. The approach has been based on the contribution of counting multiple alleles associated with disease across independent loci, without requiring compelling evidence that every locus had already achieved definitive genome-wide statistical significance. Whether PRS assist in the prediction of risk of common cancers is unknown. We built PRS from lists of genetic markers prioritized by their association with breast cancer (BCa) or prostate cancer (PCa) in a training data set and evaluated whether these scores could improve current genetic prediction of these specific cancers in independent test samples. We used genome-wide association data on 1,145 BCa cases and 1,142 controls from the Nurses' Health Study and 1,164 PCa cases and 1,113 controls from the Prostate Lung Colorectal and Ovarian Cancer Screening Trial. Ten-fold cross validation was used to build and evaluate PRS with 10 to 60,000 independent single nucleotide polymorphisms (SNPs). For both BCa and PCa, the models that included only published risk alleles maximized the cross-validation estimate of the area under the ROC curve (0.53 for breast and 0.57 for prostate). We found no significant evidence that PRS using common variants improved risk prediction for BCa and PCa over replicated SNP scores.