Methods for Pharmacogenomics and Individualized Therapy Trails

这项研究的广泛，长期目标是开发新颖和高影响力的统计和 用于发现与疗效个体间差异相关的遗传变异的计算工具 和毒性，并根据每个患者的遗传特征优化药物治疗 宪法具体目标包括：（1）构建稳健高效的统计方法， 评估SNP基因型和单倍型对具有多种表型的药物应答的影响（例如， 疗效和毒性的二进制和连续测量、右删失生存时间、区间删失时间 tp疾病进展，以及信息性删失的PSA水平和不良事件）;（2） 统计和数据挖掘技术，用于预测药物反应的基础上，高维，高度 相关的基因组数据和复杂的表型;（3）调查统计程序， 为随访提供复杂和高度多变量遗传数据的效应量的低偏倚估计 和确认研究;（4）探索一种新的机器学习形式来识别候选人 临床前研究和临床试验中的个体化治疗。所有这些目标的动机都是 研究人员的应用研究经验，并解决最及时和最重要的问题， 药物基因组学和个体化治疗。拟议的解决方案是建立在健全的统计和 数据挖掘原则。新方法的理论特性将通过以下方式严格建立： 现代经验过程理论和其他先进的数学论证。高效稳定 将设计数值算法来实施新方法。广泛的模拟研究将 进行评估的新的推理和数值程序的操作特点， 现实的设置。申请将提供给大量的癌症研究，其中大部分是进行 在杜克大学和查佩尔山的北卡罗来纳州大学。实用和用户友好 将开发软件并免费向公众分发。我们的研究将改变 药物基因组学研究和个体化治疗试验的设计和分析，这将导致 癌症和其他疾病患者的最佳治疗方法。

The broad, long-term objectives of this research are the development of novel and high-impact statistical and computational tools for discovering genetic variants associated with inter-individual differences in the efficacy and toxicity of cancer medications and for optimizing drug therapy on the basis of each patient's genetic constitution. The specific aims include: (1) construction of robust and efficient statistical methods for assessing the effects of SNP genotypes and haplotypes on drug response with a variety of phenotypes (e.g., binary and continuous measures of efficacy and toxicity, right-censored survival time, interval-censored time tp disease progression, and informatively censored PSA levels and adverse events); (2) development of statistical and data-mining techniques for predicting drug response based on high-dimensional, highly correlated genomic data and complex phenotypes; (3) investigation of statistical procedures for providing low-bias estimation of effect sizes with complex and highly multivariate genetic data for follow-up and confirmation studies; (4) exploration of a new form of machine learning for identifying candidate individualized therapies in both pre-clinical studies and clinical trials. All these aims have been motivated by the investigators' applied research experiences and address the most timely and important issues in pharmacogenomics and individualized therapy. The proposed solutions are built on sound statistical and data-mining principles. The theoretical properties of the new methods will be established rigorously via modern empirical process theory and other advanced mathematical arguments. Efficient and stable numerical algorithms will be devised to implement the new methods. Extensive simulation studies will be conducted to evaluate the operating characteristics of the new inferential and numerical procedures in realistic settings. Applications will be provided to a large number of cancer studies, most of which are carried out at Duke University and the University of North Carolina at Chapel Hill. Practical and user-friendly software will be developed and disseminated freely to the general public. Our research will change the ways pharmacogenomic studies and individualized therapy trials are designed and analyzed, which will lead to optimal treatments for patients in cancer and other diseases.