Bayesian hierarchical nonlinear models for pharmacogenomic cytotoxicity studies

描述(申请人提供)：在过去的五年里，人们对“个体化医学”的兴趣与日俱增，因此药物基因组学在癌症研究中也走到了前列。药物遗传学是研究遗传在个体变异中的作用，以应对药物、营养素和其他外源物质。在这个后基因组时代，药物遗传学已经演变为药物基因组学，这一学科被誉为人类基因组科学中已经发生并将继续发生的惊人进步的首批重大临床应用之一。药物基因组学研究也从关注单基因和单SNPs转向基因内单倍型以及基于途径和全基因组的关联研究。基于途径的研究涉及我们已经知道其功能的基因，而全基因组关联研究可以帮助我们识别已知药物转运、药物代谢和药物靶点途径之外的其他基因。因此，这两种方法--基于途径的和全基因组的--是互补的。结合各种数据来源的统计方法可能会提供新的见解，但在药物基因组学研究中是缺乏的。例如，梅奥药物基因组学研究网络(PGRN)小组一直在对抗癌药物对淋巴母细胞系的影响进行多项研究，以开始确定常见基因变异对药物反应表型的影响。这些表型包括多浓度的细胞毒性测量、基础mRNA表达阵列数据、药物处理后的mRNA表达数据、代谢物数据、已知药物途径中基因的重新测序数据，以及整个基因组的单核苷酸多态(SNPs)形式的全基因组遗传信息。目前，对这些细胞系中具有药物浓度终点(细胞毒性)的遗传变异的研究通常是通过分析在单一药物剂量下测量的药物浓度终点或剂量-反应曲线的汇总测量(即，抑制50%细胞生长的剂量，IC50)来完成的。对遗传变异对剂量-反应曲线所有方面的影响进行更全面的分析，可能有助于深入了解特定药物的药物基因组学。一种统计方法可以解释不同的基因组和细胞毒性剂量-反应数据的“层”，该方法基于贝叶斯方法。在过去的几十年里，随着计算机和计算方法的进步，马尔科夫链蒙特卡罗(MCMC)的贝叶斯方法的应用越来越多，特别是在遗传数据的应用上。在目前的赠款申请中，我们打算开发新的药物基因组学模型，用于分析从“人类变异小组”和用吉西他滨治疗的胰腺癌患者细胞系收集的细胞毒性数据。这些新的模型将帮助研究人员产生假说，这些假说将导致更好地理解基因和药物反应之间关系的复杂性质，最终导致为癌症患者开发“个体化治疗”。 公共卫生相关性：在当前的提案中，申请人提议将从“人类变异小组”细胞系和胰腺癌患者收集的许多来源的数据结合到使用贝叶斯方法的统计模型中，这将帮助研究人员产生假设，从而更好地理解基因和药物反应之间关系的复杂性质，最终导致针对癌症患者的“个性化治疗”的发展。这项研究将填补药物基因组学中开发和应用新的遗传统计方法的前沿领域的空白，在药物基因组学中，通过对细胞系、不同药物浓度和各种基因组数据类型的重复测量来进行基因型-表型相关性研究。

DESCRIPTION (provided by applicant): In the past five years, there has been an increased interest in "individualized medicine" and thus pharmacogenomics in cancer research has moved to the forefront as well. Pharmacogenetics is the study of the role of inheritance in individual variation in response to drugs, nutrients and other xenobiotics. In this post-genomic era, pharmacogenetics has evolved into pharmacogenomics, a discipline that has been heralded as one of the first major clinical applications of the striking advances that have occurred and continue to occur in human genomic science. Pharmacogenomic research has also been shifting from a focus on single genes and single SNPs to intragene haplotypes as well as pathway-based and genome-wide association studies. Pathway- based studies involve genes for which we already know function, while genome-wide association studies can help us to identify additional genes outside of known pathways of drug transport, drug metabolism and drug targets. As a result, these two approaches - pathway-based and genome-wide - are complementary. Statistical methods that combine various sources of data would likely provide novel insights, yet are lacking in pharmacogenomic studies. For example, the Mayo Pharmacogenomics Research Network (PGRN) group has been conducting multiple studies of the effect of anti- cancer drugs on lymphoblastoid cell lines to begin to define the effect of common genetic variation on drug response phenotypes. These phenotypes have included cytotoxicity measures at multiple concentrations, basal mRNA expression array data, post-drug treated mRNA expression data, metabolite data, resequencing data for genes in known drug pathways, and genome-wide genetic information in the form of single nucleotide polymorphisms (SNPs) across the genome. Currently, investigation of the genetic variation in these cell lines with drug concentration endpoints (cytotoxicity) is often completed by either analyzing a drug concentration endpoint measured at a single drug dosage or a summary measure of the dose-response curve (i.e., dose that inhibits 50% of cell growth, IC50). A more comprehensive analysis of the impact of genetic variation on all aspects of the dose- response curve may lend insight into the pharmacogenomics of a particular drug. A statistical approach that could account for disparate "layers" of genomic and cytotoxicity dose-response data is based on Bayesian methodology. Over the last few decades, applications of Bayesian methods by Markov Chain Monte Carlo (MCMC) have increased with the advancement of computers and computational methods, particular with application to genetic data. In this current grant application, we intend to develop novel pharmacogenomic models for the analysis of cytotoxicity data collected from the "Human Variation Panel" and pancreatic cancer patient cell lines treated with gemcitabine. These novel models will assist researchers in generating hypotheses that will lead to better understanding of the complex nature of the relationship between genotype and drug response, leading eventually to the development of "individualized therapy" for cancer patients. PUBLIC HEALTH RELEVANCE: In this current proposal, the applicant proposes to combine many sources of data collected from the "Human Variation Panel" cell lines and pancreatic cancer patients into statistical models using Bayesian methods that will help the investigator to generate hypotheses that will lead to better understanding of the complex nature of the relationship between genotype and drug response, leading eventually to the development of "personalized therapy" for cancer patients. This research will fill a void in the cutting-edge field of the development and application of new genetic statistical methods in pharmacogenomics in which genotype - phenotype correlation studies are conducted using repeated measurements on cell-lines, at different drug concentrations, and a variety of genomic data types.