Genome-wide association studies (GWAS) have identified thousands of variants robustly associated with complex traits. However, the biological mechanisms underlying these associations are, in general, not well understood. We propose a gene-based association method called PrediXcan that directly tests the molecular mechanisms through which genetic variation affects phenotype. The approach estimates the component of gene expression determined by an individual’s genetic profile and correlates the “imputed” gene expression with the phenotype under investigation to identify genes involved in the etiology of the phenotype. The genetically regulated gene expression is estimated using whole-genome tissue-dependent prediction models trained with reference transcriptome datasets. PrediXcan enjoys the benefits of gene-based approaches such as reduced multiple testing burden and a principled approach to the design of follow-up experiments. Our results demonstrate that PrediXcan can detect known and novel genes associated with disease traits and provide insights into the mechanism of these associations.

Importance With cure rates of childhood acute lymphoblastic leukemia (ALL) exceeding 85%, there is compelling need to mitigate treatment toxicities that can compromise quality of life. Peripheral neuropathy is the major dose-limiting toxicity of the microtubule inhibitor vincristine, an anticancer agent given to every child with ALL. Objective Identify genetic germline variants associated with the occurrence or severity of vincristine-induced peripheral neuropathy in children with ALL. Design, Setting and Participants All patients had been enrolled in one of two prospective clinical trials for childhood ALL that included treatment with 36–39 doses of vincristine. Genome-wide single nucleotide polymorphism (SNP) analysis and vincristine-induced peripheral neuropathy were assessed in all patients from whom DNA was available (n=321 patients); 222 patients (median age at 6.0 years, range 0.1–18.8 years) enrolled between 1994–1998 on the St. Jude Children’s Research Hospital protocol Total XIIIB (St. Jude cohort) with toxicity followed through January 2001, and 99 patients (median age 11.4 years, range 3.0–23.8 years) enrolled between 2007–2010 on the Children’s Oncology Group protocol AALL0433 (COG cohort) with toxicity followed through May 2011. Human leukemia cells and induced pluripotent stem cell neurons were used to assess the effects of lower CEP72 expression on vincristine sensitivity. Exposures Treatment with vincristine at a dosage of 1.5 or 2.0 mg/m2 as a component of protocol directed chemotherapy for childhood ALL. Main Outcomes and Measures Vincristine-induced peripheral neuropathy was assessed at each clinic visit using the National Cancer Institute Common Terminology Criteria for Adverse Events and prospectively graded as mild (grade 1), moderate (grade 2), serious/disabling (grade 3), or life-threatening (grade 4). Results Grade 2–4 vincristine-induced neuropathy during continuation therapy occurred in 28.8% of patients (n=64 of 222) in the St. Jude cohort and in 22.2% of patients (n=22 of 99) in the COG cohort. A SNP in the promoter region of the CEP72 gene, which encodes a centrosomal protein involved in microtubule formation, had a significant association with vincristine neuropathy (meta p =6.3 × 10−9). This SNP had a minor allele frequency of 37% (235/642), with 50 of 321 patients (16%, 95% CI 11.6%–19.5%) homozygous for the risk allele (TT at rs924607). Among patients with the high-risk CEP72 genotype (TT at rs924607), 28 of 50 patients (56%, 95% CI 41.2–70.0) developed at least one episode of grade 2–4 neuropathy, a higher rate than in patients with the CEP72 CC or CT genotype (58 of 271 patients; 21.4%, 95% CI 16.9–26.7); p=2.4×10−6. The severity (grade) of neuropathy was greater (2.4-fold by Poisson regression (p<0.0001), 2.7-fold based on mean grade of neuropathy (1.23 [95% CI 0.74 – 1.72] versus 0.45 [95% CI 0.3 – 0.6]; t test p=0.004)) in patients homozygous for the CEP72 risk allele (TT genotype), compared to patients with the CC or CT genotype. The CEP72 promoter SNP was show...

Purpose Cisplatin is widely used but highly ototoxic. Effects of cumulative cisplatin dose on hearing loss have not been comprehensively evaluated in survivors of adult-onset cancer. Patients and Methods Comprehensive audiological measures were conducted on 488 North American male germ cell tumor (GCT) survivors in relation to cumulative cisplatin dose, including audiograms (0.25 to 12 kHz), tests of middle ear function, and tinnitus. American Speech-Language-Hearing Association criteria defined hearing loss severity. The geometric mean of hearing thresholds (0.25 to 12 kHz) summarized overall hearing status consistent with audiometric guidelines. Patients were sorted into quartiles of hearing thresholds of age- and sex-matched controls. Results Increasing cumulative cisplatin dose (median, 400 mg/m2; range, 200 to 800 mg/m2) was significantly related to hearing loss at 4, 6, 8, 10, and 12 kHz (P trends, .021 to < .001): every 100 mg/m2 increase resulted in a 3.2-dB impairment in age-adjusted overall hearing threshold (4 to 12 kHz; P < .001). Cumulative cisplatin doses > 300 mg/m2 were associated with greater American Speech-Language-Hearing Association–defined hearing loss severity (odds ratio, 1.59; P = .0066) and worse normative-matched quartiles (odds ratio, 1.33; P = .093) compared with smaller doses. Almost one in five (18%) patients had severe to profound hearing loss. Tinnitus (40% patients) was significantly correlated with reduced hearing at each frequency (P < .001). Noise-induced damage (10% patients) was unaffected by cisplatin dose (P = .59). Hypertension was significantly related (P = .0066) to overall hearing threshold (4 to 12 kHz) in age- and cisplatin dose–adjusted analyses. Middle ear deficits occurred in 22.3% of patients but, as expected, were not related to cytotoxic drug dosage. Conclusion Follow-up of adult-onset cancer survivors given cisplatin should include routine inquiry for hearing status and tinnitus, referral to audiologists as clinically indicated, and hypertension control. Patients should be urged to avoid noise exposure, ototoxic drugs, and other factors that further damage hearing.

13Integration of genome-wide association studies (GWAS) and expression quantitative trait loci (eQTL) 14 studies is needed to improve our understanding of the biological mechanisms underlying GWAS hits, and 15 our ability to identify therapeutic targets. Gene-level association test methods such as PrediXcan can 16 prioritize candidate targets. However, limited eQTL sample sizes and absence of relevant developmental 17 and disease context restricts our ability to detect associations. Here we propose an efficient statistical 18 method that leverages the substantial sharing of eQTLs across tissues and contexts to improve our ability 19 to identify potential target genes: MulTiXcan. MulTiXcan integrates evidence across multiple panels 20 while taking into account their correlation. We apply our method to a broad set of complex traits available 21 from the UK Biobank and show that we can detect a larger set of significantly associated genes than 22 using each panel separately. To improve applicability, we developed an extension to work on summary 23 statistics: S-MulTiXcan, which we show yields highly concordant results with the individual level version. 24 Results from our analysis as well as software and necessary resources to apply our method are publicly 25 available. 26 Author summary 27 We develop a new method, MulTiXcan, to test the effect of gene expression regulation on complex traits, 28 integrating information available across multiple tissue studies. We show this approach has higher power 29 than traditional single-tissue methods. We extend this method to use only summary-statistics from public 30 GWAS. We apply these methods to over 200 complex traits available in the UK Biobank cohort, and 100 31 complex traits from public GWAS and discuss the findings. 32 Introduction 33Recent technological advances allow interrogation of the genome to a high level of coverage and precision, 34 enabling experimental studies that query the effect of genotype on both complex and molecular traits. 35Among these, GWAS have successfully associated genetic loci to human complex traits. GWAS meta-36 analyses with ever increasing sample sizes allow the detection of associated variants with smaller effect 37 sizes [1-3]. However, understanding the mechanism underlying these associations remains a challenging 38 problem, requiring follow-up studies and a wide array of techniques such as prioritization [4] and pathway 39 analysis [5]. 40Another approach is the study of quantitative trait loci (eQTLs), measuring association between 41 genotype and gene expression. These studies provide a wealth of biological information but tend to have 42 smaller sample sizes. A similar observation applies to QTL studies of other traits such methylation, 43 metabolites, or protein levels. 44The importance of gene expression regulation in complex traits [6][7][8][9] has motivated the integration 45 of eQTL studies and GWAS. To examine these mechanisms we developed PrediXcan [10], a method that 46 tests the mediating role of gene e...

Integration of genome-wide association studies (GWAS) and expression quantitative trait loci (eQTL) studies is needed to improve our understanding of the biological mechanisms underlying GWAS hits, and our ability to identify therapeutic targets. Gene-level association methods such as PrediXcan can prioritize candidate targets. However, limited eQTL sample sizes and absence of relevant developmental and disease context restrict our ability to detect associations. Here we propose an efficient statistical method (MultiXcan) that leverages the substantial sharing of eQTLs across tissues and contexts to improve our ability to identify potential target genes. MultiXcan integrates evidence across multiple panels using multivariate regression, which naturally takes into account the correlation structure. We apply our method to simulated and real traits from the UK Biobank and show that, in realistic settings, we can detect a larger set of significantly associated genes than using each panel separately. To improve applicability, we developed a summary result-based extension called S-MultiXcan, which we show yields highly concordant results with the individual level version when LD is well matched. Our multivariate model-based approach allowed us to use the individual level results as a gold standard to calibrate and develop a robust implementation of the summary-based extension. Results from our analysis as well as software and necessary resources to apply our method are publicly available.

Genetic variation influences the response of an individual to drug treatments. Understanding this variation has the potential to make therapy safer and more effective by determining selection and dosing of drugs for an individual patient. In the context of cancer, tumours may have specific disease-defining mutations, but a patient’s germline genetic variation will also affect drug response (both efficacy and toxicity), and here we focus on how to study this variation. Advances in sequencing technologies, statistical genetics analysis methods and clinical trial designs have shown promise for the discovery of variants associated with drug response. We discuss the application of germline genetics analysis methods to cancer pharmacogenomics with a focus on the special considerations for study design.

Understanding the genetic architecture of gene expression traits is key to elucidating the underlying mechanisms of complex traits. Here, for the first time, we perform a systematic survey of the heritability and the distribution of effect sizes across all representative tissues in the human body. We find that local h2 can be relatively well characterized with 59% of expressed genes showing significant h2 (FDR < 0.1) in the DGN whole blood cohort. However, current sample sizes (n ≤ 922) do not allow us to compute distal h2. Bayesian Sparse Linear Mixed Model (BSLMM) analysis provides strong evidence that the genetic contribution to local expression traits is dominated by a handful of genetic variants rather than by the collective contribution of a large number of variants each of modest size. In other words, the local architecture of gene expression traits is sparse rather than polygenic across all 40 tissues (from DGN and GTEx) examined. This result is confirmed by the sparsity of optimal performing gene expression predictors via elastic net modeling. To further explore the tissue context specificity, we decompose the expression traits into cross-tissue and tissue-specific components using a novel Orthogonal Tissue Decomposition (OTD) approach. Through a series of simulations we show that the cross-tissue and tissue-specific components are identifiable via OTD. Heritability and sparsity estimates of these derived expression phenotypes show similar characteristics to the original traits. Consistent properties relative to prior GTEx multi-tissue analysis results suggest that these traits reflect the expected biology. Finally, we apply this knowledge to develop prediction models of gene expression traits for all tissues. The prediction models, heritability, and prediction performance R2 for original and decomposed expression phenotypes are made publicly available (https://github.com/hakyimlab/PrediXcan).

Purpose Cisplatin is one of the most commonly used chemotherapy drugs worldwide and one of the most ototoxic. We sought to identify genetic variants that modulate cisplatin-associated ototoxicity (CAO). Experimental Design We performed a genome-wide association study (GWAS) of CAO using quantitative audiometry (4–12 kHz) in 511 testicular cancer survivors of European genetic ancestry. We performed polygenic modeling and functional analyses using a variety of publicly available databases. We used an electronic health record cohort to replicate our top mechanistic finding. Results One SNP, rs62283056, in the first intron of Mendelian deafness gene WFS1 (wolframin ER transmembrane glycoprotein) and an expression quantitative trait locus (eQTL) for WFS1 met genome-wide significance for association with CAO (P=1.4×10−8). A significant interaction between cumulative cisplatin dose and rs62283056 genotype was evident, indicating that higher cisplatin doses exacerbate hearing loss in patients with the minor allele (P=0.035). The association between decreased WFS1 expression and hearing loss was replicated in an independent BioVU cohort (n=18,620 patients, Bonferroni adjusted P<0.05). Beyond this top signal, we show CAO is a polygenic trait and that SNPs in and near 84 known Mendelian deafness genes are significantly enriched for low P-values in the GWAS (P=0.048). Conclusions We show for the first time the role of WFS1 in CAO and document a statistically significant interaction between increasing cumulative cisplatin dose and rs62283056 genotype. Our clinical translational results demonstrate that pre-therapy patient genotyping to minimize ototoxicity could be useful when deciding between cisplatin-based chemotherapy regimens of comparable efficacy with different cumulative doses.