Statistical Genetics of Dose Response Traits

National Institute Of Environmental Health Sciences

Investigators

Linked publications & trials

Abstract

One of the goals of the project is to evaluate the impact of synergistic effects of drug combination. My research group participated in a group competition to compare statistical methods for quantifying and predicting synergy. The effectiveness of most cancer targeted therapies is short-lived. Tumors often develop resistance that might be overcome with drug combinations. However, the number of possible combinations is vast, necessitating data-driven approaches to find optimal patient-specific treatments. Here we report AstraZeneca's large drug combination dataset, consisting of 11,576 experiments from 910 combinations across 85 molecularly characterized cancer cell lines, and results of a DREAM Challenge to evaluate computational strategies for predicting synergistic drug pairs and biomarkers. 160 teams participated to provide a comprehensive methodological development and benchmarking. Winning methods incorporate prior knowledge of drug-target interactions. Synergy is predicted with an accuracy matching biological replicates for >60% of combinations. However, 20% of drug combinations are poorly predicted by all methods. Genomic rationale for synergy predictions are identified, including ADAM17 inhibitor antagonism when combined with PIK3CB/D inhibition contrasting to synergy when combined with other PI3K-pathway inhibitors in PIK3CA mutant cells. With my student Farida Akhtari, we evaluated the impact of very fine measures of ethnicity on overall dose response. Various studies have shown that people of Eurasian origin contain traces of DNA inherited from interbreeding with Neanderthals. Recent studies have demonstrated that these Neanderthal variants influence a range of clinically important traits and diseases. Thus, understanding the genetic factors responsible for the variability in individual response to drug or chemical exposure is a key goal of pharmacogenomics and toxicogenomics, as dose responses are clinically and epidemiologically important traits. It is well established that ethnic and racial differences are important in dose response traits, but to our knowledge the influence of Neanderthal ancestry on response to xenobiotics is unknown. Towards this aim, we examined if Neanderthal ancestry plays a role in cytotoxic response to anti-cancer drugs and toxic environmental chemicals. We identified common Neanderthal variants in lymphoblastoid cell lines (LCLs) derived from the globally diverse 1000 Genomes Project and Caucasian cell lines from the Children's Hospital of Oakland Research Institute. We analyzed the effects of these Neanderthal alleles on cytotoxic response to 29 anti-cancer drugs and 179 environmental chemicals at varying concentrations using genome-wide data. We identified and replicated single nucleotide polymorphisms (SNPs) from these association results, including a SNP in the SNORD-113 cluster. Our results also show that the Neanderthal alleles cumulatively lead to increased sensitivity to both the anti-cancer drugs and the environmental chemicals. Our results demonstrate the influence of Neanderthal ancestry-informative markers on cytotoxic response. These results could be important in identifying biomarkers for personalized medicine or in dissecting the underlying etiology of dose response traits. Also, with Farida, we have evaluated potential new confounders in the LCL model. Lymphoblastoid cell lines (LCLs) are a widely used model system in pharmacogenomics and toxicogenomics studies due to their scalability, efficiency and cost-effectiveness. Since LCLs are cultured from individuals from a wide range of demographic populations and environmental exposures, we sought to identify the confounders for drug response in LCL assays. LCLs were cultured from 93 breast cancer patients from the University of North Carolina Lineberger Comprehensive Cancer Center Breast Cancer Database, undergoing paclitaxel chemotherapy. Each LCL was assayed at 10 different concentrations of paclitaxel to measure cell viability. The patient data included treatment regimens, cancer status, demographic and environmental variables and clinical outcomes. We used the multivariate analysis of variance (MANOVA) method to identify the in vivo variables associated with in vitro dose response. We also analyzed relationships between in vitro dose response and in vivo clinical variables using various statistical methods. In a novel data set that includes both in vivo and in vitro data from breast cancer patients, race (p-value = 0.0049) and smoking status (p-value = 0.0050) were found to be significantly associated with in vitro dose response in LCLs. The smoking status of the donor individuals, from whom the LCLs are created, is usually unknown and hence not controlled for in dose response analyses in LCLs. Our results indicate that in vivo smoking status could be a confounder for in vitro dose response assays in LCLs and hence should be recorded and controlled for in the statistical analyses of dose response assays in LCLs. Further research is required to understand the mechanism by which exposure to smoking in vivo affects in vitro dose response in LCLs. We also recently completed a high throughput screen of 44 anti-cancer drugs in this model. Cancer patients exhibit a broad range of inter-individual variability in response and toxicity to several widely used anticancer drugs. Genetic association mapping can be used to understand the genetic etiology of cancer drug response by identifying genes related to differential response. To identify novel genes that influence the response of 44 FDA-approved anticancer drugs widely used to treat various different types of cancer, we screened 680 lymphoblastoid cell lines (LCLs) from the racially and ethnically diverse 1000 Genomes Project with these drugs. Our genome-wide association mapping identified several novel genetic variants associated with the response of a broad range of anticancer drugs. We conducted further analyses and functional validation for one of the genes from our association mapping results, NAD(P)H quinone dehydrogenase 1 (NQO1), to identify the mechanism of action by which it influences drug response. Our results show that the expression levels of NQO1, an oxidative stress gene, are positively correlated with cell viability in LCLs exposed to multiple anticancer drugs. Additionally, the compendium of high-throughput dose response data along with the systematic genome-wide analyses reported in this study provides an invaluable resource for future pharmacogenomic studies aiming to optimize cancer therapeutics. Within these assays, we have extended the use of the model to evaluate synergism/antagonism in combination treatment/exposure. Combination therapy is quite common in modern chemotherapy treatment since drugs often work synergistically, and it is an important progression in the use of the LCL model to expand work for drug combinations. In the present work, we demonstrate that synergy occurs and can be quantified in LCLs across a range of clinically important drug combinations. LCLs have been commonly employed in association mapping in cancer pharmacogenomics, but it is so far untested as to whether synergistic effects have a genetic etiology. Here we use cell lines from extended pedigrees to demonstrate that there is a substantial heritable component to synergistic drug response. Additionally, we perform linkage mapping in these pedigrees to identify putative regions linked to this important phenotype. This demonstration supports the premise of expanding the use of LCL model to perform association mapping for combination therapies. There are also a number of methodological challenges related to quantifying dose response curves and synergism/antagonism. With my PhD student Jun Ma, we have been working an evolutionary algorithm method for quantifying dose response. Nonli