Lang Wu

Transcriptome-wide association studies accounting for colocalization using Egger regression

Integrating genome‐wide association (GWAS) and expression quantitative trait locus (eQTL) data into transcriptome‐wide association studies (TWAS) based on predicted expression can boost power to detect novel disease loci or pinpoint the susceptibility gene at a known disease locus. However, it is often the case that multiple eQTL genes colocalize at disease loci, making the identification of the true susceptibility gene challenging, due to confounding through linkage disequilibrium (LD). To distinguish between true susceptibility genes (where the genetic effect on phenotype is mediated through expression) and colocalization due to LD, we examine an extension of the Mendelian randomization (MR) egger regression method that allows for LD while only requiring summary association data for both GWAS and eQTL. We derive the standard TWAS approach in the context of MR and show in simulations that the standard TWAS does not control type I error for causal gene identification when eQTLs have pleiotropic or LD‐confounded effects on disease. In contrast, LD‐aware MR‐Egger (LDA MR‐Egger) regression can control type I error in this case while attaining similar power as other methods in situations where these provide valid tests. However, when the direct effects of genetic variants on traits are correlated with the eQTL associations, all of the methods we examined including LDA MR‐Egger regression can have inflated type I error. We illustrate these methods by integrating gene expression within a recent large‐scale breast cancer GWAS to provide guidance on susceptibility gene identification.

Assessing the genetic effect mediated through gene expression from summary eQTL and GWAS data

Integrating genome-wide association (GWAS) and expression quantitative trait locus (eQTL) data can boost power to detect novel disease loci or pinpoint the susceptibility gene at a known disease locus. However, it is often the case that multiple eQTL genes co-localize at disease loci (an effect of linkage disequilibrium, LD), making the identification of the true susceptibility gene challenging. To distinguish between true susceptibility genes (i.e. when the genetic effect on phenotype is mediated through expression) and spurious co-localizations, we developed an approach to quantify the genetic effect mediated through expression. Our approach can be viewed as an extension of the standard Mendelian randomization Egger technique to incorporate LD among variants while only requiring summary association data (both GWAS and eQTL) along with LD from reference panels. Through simulations we show that when eQTLs have pleiotropic or LD-confounded effects on disease, our approach provides adequate control of Type I error, more power, and less bias than previously-proposed methods. When there is no effect of gene expression on disease, our method has the desired Type I Error, while LD-aware Mendelian randomization, which assumes no pleiotropy, can have inflated Type I Error. In the presence of direct effect of genetic variants on traits, our approach attained up to 3x greater power than the standard approaches while properly controlling Type I error. To illustrate our method, we analyzed recent large scale breast cancer GWAS with gene expression in breast tissue from GTEx.Integrating genome-wide association (GWAS) and expression quantitative trait locus (eQTL) data into transcriptome-wide association studies (TWAS) based on predicted expression can boost power to detect novel disease loci or pinpoint the susceptibility gene at a known disease locus. However, it is often the case that multiple eQTL genes colocalize at disease loci, making the identification of the true susceptibility gene challenging, due to confounding through linkage disequilibrium (LD). To distinguish between true susceptibility genes (where the genetic effect on phenotype is mediated through expression) and colocalization due to LD, we examine an extension of the Mendelian Randomization Egger regression method that allows for LD while only requiring summary association data for both GWAS and eQTL. We derive the standard TWAS approach in the context of Mendelian Randomization and show in simulations that the standard TWAS does not control Type I error for causal gene identification when eQTLs have pleiotropic or LD-confounded effects on disease. In contrast, LD Aware MR-Egger regression can control Type I error in this case while attaining similar power as other methods in situations where these provide valid tests. However, when the direct effects of genetic variants on traits are correlated with the eQTL associations, all of the methods we examined including LD Aware MR-Egger regression can have inflated Type I error. We illustrate these methods by integrating gene expression within a recent large-scale breast cancer GWAS to provide guidance on susceptibility gene identification.

Abstract 2969: Genetically predicted blood protein biomarkers and prostate cancer risk: an analysis in over 140,000 European descendants

Prostate cancer (PrCa) is the second most frequently diagnosed malignancy among males in many countries. Several protein markers measured in blood have been found to be associated with PrCa risk. However, most previous studies assessed only a small number of biomarkers or included a small sample size. To search for novel protein biomarkers for PrCa risk, we performed a large study in 79,194 prostate cancer cases and 61,112 controls of European ancestry included in PRACTICAL/ELLIPSE consortia by using genetic instruments. Protein quantitative trait loci (pQTLs) for 1,478 plasma proteins identified in a large study of 3,301 European descendants were used as instruments to evaluate associations between genetically predicted protein levels and PrCa risk. For proteins showing a significant association with PrCa risk, we further evaluated whether genetically predicted mRNA expression levels of the corresponding genes were associated with PrCa risk, by using mRNA expression prediction models for blood, prostate and cross tissue built using data of the Genotype-Tissue Expression Project. We identified 31 proteins showing an association of their predicted levels with PrCa risk at a false discovery rate Our study has identified multiple proteins significantly associated with PrCa risk. Further research is needed to evaluate potential utility of the identified proteins in early detection of PrCa. Citation Format: Lang Wu, Xiang Shu, Jiandong Bao, Xingyi Guo, the PRACTICAL, CRUK, BPC3, CAPS, PEGASUS consortia, Zsofia Kote-Jarai, Christopher A. Haiman, Rosalind A. Eeles, Wei Zheng. Genetically predicted blood protein biomarkers and prostate cancer risk: an analysis in over 140,000 European descendants [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2969.

Abstract 1309: Association between insulin resistance and breast cancer risk: A Mendelian randomization analysis of data from 228,000 women of European descent

Background: Epidemiologic studies suggest that insulin resistance may be associated with breast cancer risk. We conducted Mendelian randomization (MR) analyses to reduce the biases associated with previous studies and provide evidence for causal inference. Materials and Methods: We used genetic variants identified in genome-wide association studies for circulating fasting insulin (15 variants), early insulin secretion (16 variants), fasting proinsulin (8 variants), fasting glucose (35 variants), and 2-hour glucose (8 variants) as instruments in MR analyses. To reduce possible pleiotropic effects, variants associated with obesity were removed from the instruments. We first evaluated the association of these instruments with type 2 diabetes risk in 110,452 subjects to assess instrument validity. We then investigated the association of these instruments with breast cancer risk using data obtained from 122,977 cases and 105,974 controls of European descent included in the Breast Cancer Association Consortium (BCAC). Odds ratios (OR) were calculated to measure the associations of instrumental variables with risk of overall breast cancer and its subtypes defined by estrogen-receptor [ER] status. Results: All instrumental variables constructed for this study were strongly associated with type 2 diabetes risk with ORs of 3.01 (p=7.86x10-5), 0.22 (p=3.54x10-14), 1.90 (p=8.28x10-4), 6.11 (p=3.59x10-19), and 1.91 (p=6.8x10-16) for per unit increase of fasting insulin, early insulin secretion, fasting proinsulin, fasting glucose, and 2-hour glucose levels, respectively. Statistically significant associations with overall breast cancer risk were found for fasting insulin (OR=1.36 for per unit increase, 95% CI=1.09-1.70, p=0.011) and fasting proinsulin (OR=1.21, 95% CI=1.06- 1.38, p=0.011). These associations were observed only for ER-positive breast cancer. No statistically significant association at p Conclusions: Our study provides strong support that certain insulin resistance traits may be causally associated with risk of breast cancer, particularly ER-positive breast cancer. Citation Format: Xiang Shu, Lang Wu, Nikhil K. Khankari, Kyriaki Michailidou, Manjeet K. Bolla, Jean Wang, Joe Dennis, Xiao-ou Shu, Jacques Simard, Douglas F. Easton, Wei Zheng. Association between insulin resistance and breast cancer risk: A Mendelian randomization analysis of data from 228,000 women of European descent [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1309. doi:10.1158/1538-7445.AM2017-1309