Ailith Pirie

Investigation of Exomic Variants Associated with Overall Survival in Ovarian Cancer

BACKGROUND:While numerous susceptibility loci for epithelial ovarian cancer (EOC) have been identified, few associations have been reported with overall survival. In the absence of common prognostic genetic markers, we hypothesize that rare coding variants may be associated with overall EOC survival and assessed their contribution in two exome-based genotyping projects of the Ovarian Cancer Association Consortium (OCAC). METHODS:The primary patient set (Set 1) included 14 independent EOC studies (4,293 patients) and 227,892 variants, and a secondary patient set (Set 2) included six additional EOC studies (1,744 patients) and 114,620 variants. Because power to detect rare variants individually is reduced, gene-level tests were conducted. Sets were analyzed separately at individual variants and by gene, and then combined with meta-analyses (73,203 variants and 13,163 genes overlapped). RESULTS:No individual variant reached genome-wide statistical significance. A SNP previously implicated to be associated with EOC risk and, to a lesser extent, survival, rs8170, showed the strongest evidence of association with survival and similar effect size estimates across sets (Pmeta = 1.1E-6, HRSet1 = 1.17, HRSet2 = 1.14). Rare variants in ATG2B, an autophagy gene important for apoptosis, were significantly associated with survival after multiple testing correction (Pmeta = 1.1E-6; Pcorrected = 0.01). CONCLUSIONS:Common variant rs8170 and rare variants in ATG2B may be associated with EOC overall survival, although further study is needed. IMPACT:This study represents the first exome-wide association study of EOC survival to include rare variant analyses, and suggests that complementary single variant and gene-level analyses in large studies are needed to identify rare variants that warrant follow-up study. Cancer Epidemiol Biomarkers Prev; 25(3); 446-54. ©2016 AACR.Background: While numerous susceptibility loci for epithelial ovarian cancer (EOC) have been identified, few associations have been reported with overall survival. In the absence of common prognostic genetic markers, we hypothesize that rare coding variants may be associated with overall EOC survival and assessed their contribution in two exome-based genotyping projects of the Ovarian Cancer Association Consortium (OCAC). Methods: The primary patient set (Set 1) included 14 independent EOC studies (4,293 patients) and 227,892 variants, and a secondary patient set (Set 2) included six additional EOC studies (1,744 patients) and 114,620 variants. Because power to detect rare variants individually is reduced, gene-level tests were conducted. Sets were analyzed separately at individual variants and by gene, and then combined with meta-analyses (73,203 variants and 13,163 genes overlapped). Results: No individual variant reached genome-wide statistical significance. A SNP previously implicated to be associated with EOC risk and, to a lesser extent, survival, rs8170, showed the strongest evidence of association with survival and similar effect size estimates across sets (Pmeta = 1.1E−6, HRSet1 = 1.17, HRSet2 = 1.14). Rare variants in ATG2B, an autophagy gene important for apoptosis, were significantly associated with survival after multiple testing correction (Pmeta = 1.1E−6; Pcorrected = 0.01). Conclusions: Common variant rs8170 and rare variants in ATG2B may be associated with EOC overall survival, although further study is needed. Impact: This study represents the first exome-wide association study of EOC survival to include rare variant analyses, and suggests that complementary single variant and gene-level analyses in large studies are needed to identify rare variants that warrant follow-up study. Cancer Epidemiol Biomarkers Prev; 25(3); 446–54. ©2016 AACR.

The effect of rare variants on inflation of the test statistics in case-control analyses

BackgroundThe detection of bias due to cryptic population structure is an important step in the evaluation of findings of genetic association studies. The standard method of measuring this bias in a genetic association study is to compare the observed median association test statistic to the expected median test statistic. This ratio is inflated in the presence of cryptic population structure. However, inflation may also be caused by the properties of the association test itself particularly in the analysis of rare variants. We compared the properties of the three most commonly used association tests: the likelihood ratio test, the Wald test and the score test when testing rare variants for association using simulated data.ResultsWe found evidence of inflation in the median test statistics of the likelihood ratio and score tests for tests of variants with less than 20 heterozygotes across the sample, regardless of the total sample size. The test statistics for the Wald test were under-inflated at the median for variants below the same minor allele frequency.ConclusionsIn a genetic association study, if a substantial proportion of the genetic variants tested have rare minor allele frequencies, the properties of the association test may mask the presence or absence of bias due to population structure. The use of either the likelihood ratio test or the score test is likely to lead to inflation in the median test statistic in the absence of population structure. In contrast, the use of the Wald test is likely to result in under-inflation of the median test statistic which may mask the presence of population structure.

Abstract 4636: Investigation of exome variants associated with overall survival in ovarian cancer

Many germline genetic variants have been found to be associated with susceptibility to epithelial ovarian cancer (EOC), but currently few associations have been identified for EOC outcomes such as overall survival. In the absence of evidence for common germline markers, it is possible that rare variants not captured in GWAS arrays may be associated with overall survival. We hypothesize that rare variants, either individually or combined across a gene, may be associated with overall survival in EOC. We utilized data from two genotyping projects of the Ovarian Cancer Association Consortium (OCAC) that used commercial exome-based arrays. The primary sample, genotyped on the Affymetrix Axiom Exome Array at 227,892 standard and custom variants, consisted of 14 independent EOC studies (4293 cases; 2257 deaths). The secondary sample, genotyped on the Illumina Infinium HumanExome BeadChip at 114,620 variants, consisted of six additional EOC studies (1744 cases; 1027 deaths). Both sets were analyzed using Cox proportional hazards regression to determine the association of each variant with overall survival time. Meta-analysis was conducted across samples at 73,203 overlapping variants. Gene-level tests of over 18,000 genes were also conducted using burden tests and Sequencing Kernel Association Tests (SKAT) adapted for Cox proportional hazards models. Gene-level tests were conducted separately for the primary and secondary cases and then were meta-analyzed. Models were adjusted for age, site, and three principle components. Six individual variants had meta-analysis p-values With over 6000 EOC cases in the primary and secondary samples, this is the largest genome-wide assessment of rare variant genetic associations with EOC overall survival. Variant rs8170 in BABAM1, a necessary component of the BRCA1 complex, is associated with overall survival. Single variant and gene-level analyses provide complementary approaches, suggesting additional candidates that warrant follow-up study. Citation Format: Stacey J. Winham, Brooke L. Fridley, Melissa C. Larson, Zachary Fogarty, Andrew Berchuck, Yian Ann Chen, Hui-Yi Lin, Georgia Chenevix-Trench, Jenny Permuth-Wey, Thomas A. Sellers, Ailith Pirie, Ellen L. Goode, Ovarian Cancer Association Consortium. Investigation of exome variants associated with overall survival in ovarian cancer. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4636. doi:10.1158/1538-7445.AM2015-4636

The effect of height, BMI and serum lipid levels on ovarian cancer prognosis in over 12,000 women: a Mendelian randomization study

Introduction Previous observational studies investigating height, body mass index and serum lipid levels as prognostic factors in ovarian cancer have been inconclusive. In addition to possible influences of reverse causation, it is possible that factors such as diet, socio-economic status and other lifestyle factors are confounding true associations. Mendelian Randomization (MR) utilises genotype data for variants associated with phenotypes of interest to create genetic risk scores for these modifiable exposures. One advantage of using genetic markers as proxies is that they are determined from birth and are therefore unaffected by confounding variables. We aim to use MR to investigate the association between height, BMI and serum lipid levels (high-density lipoprotein (HDL), low-density lipoprotein (LDL) and triglycerides) and ovarian cancer prognosis in the absence of confounding variables. Methods We used data from 12,908 invasive ovarian cancer cases with 5,813 events from 26 studies in the Ovarian Cancer Association Consortium. All individuals were of European ancestry. We calculated genetic risk scores for each individual for height, BMI and serum lipids by taking the sum of the alleles associated with the trait, weighted by the size of their effect on the trait. The genetic risk scores were then included in a Cox proportional hazards model adjusted for study and two principal components to test for association with prognosis. For the analysis of height, we included 422 uncorrelated single nucleotide polymorphisms identified by the Genetic Investigation of Anthropometric Traits (GIANT) consortium as associated with height at genome-wide significance. In the analysis of BMI, we included 32 SNPs associated with BMI in analyses by the GIANT consortium. In order to account for the pleiotropy between the three lipid types we included the genetic risk scores for each of the three traits in a joint analysis. SNPs identified by the Global Lipids Genetics Consortium as associated with lipid levels were included: 95 with HDL, 82 with LDL and 64 with triglycerides. Results We found no evidence of association between the five genetic risk scores and ovarian cancer prognosis. The genetic risk score for height had an estimated hazard ratio of 1.01, 95% confidence interval 0.94 - 1.08, p-value = 0.82. The hazard ratio for BMI was 1.00, 95% CI 0.95 - 1.05, p-value = 0.99. The hazard ratios for HDL, LDL and triglycerides were 1.03(0.94-1.13), 1.02(0.94-1.12) and 1.08(0.96-1.21) respectively with p-values = 0.53, 0.58 and 0.19. Conclusion Our study does not provide any evidence of association between height, BMI and serum lipid levels and ovarian cancer prognosis. Citation Format: Ailith Pirie, Suzanne C. Dixon, Penelope M. Webb, Wei Zheng, Paul D. P. Pharoah, on behalf of the Ovarian Cancer Association Consortium. The effect of height, BMI and serum lipid levels on ovarian cancer prognosis in over 12,000 women: a Mendelian randomization study. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4637. doi:10.1158/1538-7445.AM2015-4637

Abstract 881: Association of adult body mass index and height with risk of ovarian cancer in 39,000 women: Results of a Mendelian randomization study

INTRODUCTION: Observational studies have reported positive associations between higher body mass index (BMI) and risks of borderline ovarian tumors and non-high grade serous ovarian cancer (non-HGSC), but the lack of association observed for HGSC may be due to bias arising from weight loss before diagnosis. Observational studies also suggest a positive association between greater height and ovarian cancer risk, but confounding factors may be obscuring the true relationship. The Mendelian randomization (MR) technique uses genetic markers as proxies for environmental risk factors, and can overcome the limitations of bias and confounding which affect observational studies. AIM / METHODS: This study used MR to elucidate the relationship between body size (BMI and height) and risk of ovarian cancer. We pooled data from 16,395 cases and 23,003 controls, all genetically European, from 39 studies in the Ovarian Cancer Association Consortium. We constructed a weighted genetic risk score (GRS) for each trait, summing trait-increasing alleles at 31 single nucleotide polymorphisms (SNPs) associated with BMI and 609 SNPs associated with height in genome-wide association studies, weighting alleles by published β-coefficients for their effect on the trait. Each GRS was a strong instrument for the trait (F-statistics 33.8 [BMI] and 516 [height]). In a two-stage predictor substitution MR approach, we used multivariate logistic regression to model case-control status on body size predicted by each GRS. Study-specific estimates per 5-unit increase in predicted BMI or height were pooled to generate pooled odds ratios (OR) and 95% confidence intervals (CI) using random-effects meta-analysis. Our primary hypotheses were that genetic BMI would be associated with increased risk of non-HGSC but not HGSC and genetic height would be associated with increased risk of ovarian cancer overall. RESULTS: Higher genetically-predicted BMI was associated with increased risk of non-HGSC cancer (OR 1.37, 95% CI 1.02-1.83 per 5-unit increase) but not HGSC (OR 1.05, 95% CI 0.83-1.33). In secondary analyses stratified by behavior/subtype, the strongest association was seen for low grade/borderline serous cancers (OR 2.02, 95% CI 1.24-3.30). Women with greater genetically-predicted height had a modestly increased risk of all (invasive and borderline) ovarian tumors (OR 1.06, 95% CI 1.01-1.11 per 5 cm). In secondary analyses stratified by histologic subtype, the strongest association was seen for clear cell cancers (OR 1.20, 95% CI 1.04-1.38). CONCLUSION: This study is the first to apply MR to investigate ovarian cancer risk factors. These data confirm results from epidemiologic studies, suggesting that obesity is causally associated with non-HGSC, but does not increase risk of the most common HGSC subtype. They also support an association between height and ovarian cancer. Citation Format: Suzanne C. Dixon, Christina M. Nagle, Aaron P. Thrift, Paul D.P Pharoah, Ailith Pirie, Celeste Leigh Pearce, Wei Zheng, Penelope M. Webb, for the Ovarian Cancer Association Consortium. Association of adult body mass index and height with risk of ovarian cancer in 39,000 women: Results of a Mendelian randomization study. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 881. doi:10.1158/1538-7445.AM2015-881

Abstract 5545: Rare genetic variation association with neurotoxicity and infection in breast cancer patients enrolled in PG-SNPS

The genetic architecture underlying individual variation in susceptibility to severe toxicity to chemotherapeutic agents is not yet fully understood. We used a gene-based omnibus test to investigate the role of rare genetic variation in severe neurosensory toxicity and infection. Samples from a breast cancer pharmacogenetics study (PG-SNPS), were genotyped using the Illumina Human Exome Beadchip array. Early breast cancer patients treated with adjuvant and neo-adjuvant chemotherapy trials were recruited into PG-SNPS. Clinical data relating to their chemotherapy-related toxicities was collected prospectively. The National Cancer Institute Common Toxicity Criteria (NCI CTC version 2) was used to assess toxicities. For each of the toxicities, a case-control analysis was conducted using the Rare Admixture Maximum Likelihood method(1). RAML is an omnibus test that looks at joint effects of multiple variants on a phenotype. The test considers the probability that a given variant is associated with severe toxicity, the average effect of the associated variants within a gene on severe toxicity, and the expected standard error of this effect. Patients were considered as taxane-related neurosensory cases, if they had a CTC grade of ≥ 2 and as infection cases if they have a CTC grade of ≥ 3. For both analyses variants with minor allele frequency less than 2% were included. Only patients treated with paclitaxel were included in the neurosensory analysis. In this analysis we included 1,127 patients, 333 of which were considered cases. We tested variants in 14,762 genes exome-wide. The gene LRRTM3 was associated with severe neurosensory toxicity (p-value=5.5 x 10-4). LRRTM3 is a protein-coding gene with strong links to the maintenance and development of the nervous system. Patients exposed to specific adjuvant/neo-adjuvant breast cancer combination chemotherapy regimens with paclitaxel, epirubicin, methotrexate, cyclophosphamide, 5fluorouracil or gemcitabine, were included in the infection analysis (2,3,4). These included 1,564 patients, 583 of which were considered cases. We tested 15,161 genes exome-wide for association with severe infection. The top associated gene was FKBP5 with a p-value of 7 x 10-5. This gene has a role in immunoregulation and encodes a protein which binds to the immunosuppressants FK506 and rapamycin. These genes may be of future interest as candidate genes within pharmacodynamic pathways. 1. Tyrer JP, Guo Q, Easton DF, Pharoah PD. The admixture maximum likelihood test to test for association between rare variants and disease phenotypes. BMC Bioinformatics 2013;14:177. 2. NEAT trial - ClinicalTrials.gov Identifier NCT00003577 3. tAnGo trial - ClinicalTrials.gov Identifier NCT00039546 4. Neo-tAnGo trial - ClinicalTrials.gov Identifier NCT00070278 Citation Format: Ailith Pirie, Jean Abraham, Kyriaki Michailidou, Jonathan Tyrer, Leila Dorling, Joe Dennis, Helena Earl, Carlos Caldas, Paul Pharoah. Rare genetic variation association with neurotoxicity and infection in breast cancer patients enrolled in PG-SNPS. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 5545. doi:10.1158/1538-7445.AM2014-5545

Abstract 946: Exome genotyping array identifies rare and low-frequency variants that may be associated with ovarian cancer risk

Proceedings: AACR Annual Meeting 2014; April 5-9, 2014; San Diego, CA Background: Common genetic variants identified by genome-wide association studies and follow-up genotyping initiatives only explain a small proportion of the heritability of epithelial ovarian cancer (EOC). We hypothesized that some of the missing heritability may be explained by rare (minor allele frequency (MAF) 90%. After removing 31,436 monomorphic markers, the 249,301 remaining markers were included in association analyses. We confirmed variants at previously reported EOC susceptibility loci (9p22, 3q25, 17cen-q21.3, and 8q24), with p-values between 10-5 and 10-11. Several correlated low-frequency non-synonymous variants (MAF=5%) residing in a novel EOC susceptibility gene at 3q25.31 were risk-associated. The most significant variant at 3q25.31 had an OR=1.36 and p = 5.58x10-8. Also noteworthy is a rare variant at 15q15.1-q21.1 (MAF=0.5%) that introduces a stop codon in a gene belonging to the tyrosine kinase family (OR=0.39, CI=0.26-0.58, p = 4.52x10-6). We also identified a more common synonymous coding variant in a phosphatase at 6q21.3 that was associated with increased risk (OR=1.22, CI=1.13-1.33, p = 1.84x10-6, MAF=9.6%). Population stratification and gene-based analyses are underway to further examine the impact of the newly-identified variants on EOC risk. A meta-analysis to include data from an additional 2104 EOC cases and 2516 controls genotyped with Illuminas HumanExome Beadchip rare variant array is also underway. Functional analysis of the most promising variants will follow. Conclusions: Preliminary data from this large-scale study reveals novel rare and low-frequency variants that may influence susceptibility to epithelial ovarian cancer. Citation Format: Jennifer Permuth-Wey, Y. Ann Chen, Zhihua Chen, Andrew Berchuck, Georgia Chenevix-Trench, Jennifer Doherty, Simon Gayther, Ellen L. Goode, Edwin Iversen, Alvaro N.A. Monteiro, Leigh Pearce, Paul D.P. Pharoah, Catherine M. Phelan, Ailith Pirie, Susan Ramus, Mary Ann Rossing, Joellen M. Schildkraut, Thomas A. Sellers, on behalf of the Ovarian Cancer Association Consortium. Exome genotyping array identifies rare and low-frequency variants that may be associated with ovarian cancer risk. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 946. doi:10.1158/1538-7445.AM2014-946