A Translational Bioinformatics Approach in the Drug Interaction Research

描述（由申请人提供）：药物-药物相互作用（ddi）对公众健康构成越来越大的威胁，估计每年造成195,000例住院和74,000例急诊室就诊。目前的DDI研究调查了药物相互作用的不同方面，包括计算和实验。虽然这些方法是相辅相成的，但它们通常是独立进行的，没有协调。体外药理学实验使用完整细胞、微粒体蛋白组分或重组系统来研究药物相互作用机制。药物流行病学（在人群中）使用基于人群的方法和大型电子病历（EMR）数据库来调查DDI对药物疗效和药物不良反应（adr）的贡献。在这项拨款提案中，将开发新的生物信息学数据挖掘方法来从EMR中挖掘ddi，并将在体外进一步验证。以下是具体目标。在Aim 1中，提出了一种动态嵌套病例对照设计，以检测单一药物或DDI对ADR的影响。开发了一种新的经验贝叶斯方法来测试药物和ddi诱导的adr，并将估计错误发现率。在Aim 2中，提出了一种新的广义倾向评分方法来分析高维药物数据。与传统的倾向评分法相比，该方法在识别高度相关药物的不良反应效应方面具有更强的能力。目标3，使用目标1和目标2中开发的单变量和多变量数据挖掘方法，我们将使用EMR数据库和高通量酶筛选法检测增加一种明确的不良反应（肌病）风险的新药和ddi。在我们的初步工作中，使用我们提出的方法和220万条记录EMR数据库，确定了6个肌病风险ddi (p < 5 - 10-6)，包括新发现的喹硫平和氯喹之间的相互作用。综上所述，由于喹硫平通过CYP3A4途径抑制氯喹代谢并阻断OATP1B1/1B3跨膜转运蛋白，它们增加的肌病风险比单独增加的风险高2.17倍。因此，这项工作的成功实施将展示一个完整的转化范围，从基于emr的DDI发现开始，到通过药理学实验阐明分子DDI机制结束。总之，这些初步数据表明，我们的转化方法是临床DDI研究的一种高度可行和极其强大的方法，可能对这一新兴的公共卫生危害产生有价值的见解。

DESCRIPTION (provided by applicant): Drug-drug interactions (DDIs) represent an increasing threat to public health, causing an estimated 195,000 annual hospitalizations and 74,000 emergency room visits. Current DDI research investigates different aspects of drug interactions, both computationally and experimentally. Although these approaches are complementary, they are usually conducted independently and without coordination. In vitro pharmacology experiments use intact cells, microsomal protein fractions, or recombinant systems to investigate drug interaction mechanisms. Pharmaco-epidemiology (in populo) uses a population-based approach and large electronic medical record (EMR) databases to investigate the contribution of a DDI to drug efficacy and adverse drug reactions (ADRs). In this grant proposal, novel bioinformatics data mining approaches will be developed to mine DDIs from EMR, and they will be further validated in vitro. The following are specific aims. In Aim 1, a nove dynamic nested case-control design is proposed to detect of either single drug or DDI effects on the ADR. A new empirical Bayes method is developed to test the drug and DDI-induced ADRs, and it will estimate false discovery rates. In Aim 2, a novel generalized propensity score method is proposed to analyze high dimensional medication data. This method possesses more power in identifying ADR effects from highly correlated drugs, than the conventional propensity score method. Aim 3, using the univariate and multivariate data mining methods developed in aims 1 and 2, we will detect novel drugs and DDIs that increase the risk of one well-defined ADR, myopathy, using the EMR database and high-throughput enzymatic screening assays. In our preliminary work, using our proposed methodology and a 2.2 million record EMR database, six myopathy risk DDIs were identified (p < 5�10-6), including a newly discovered interaction between quetiapine and chloroquine. If taken together, they increase myopathy risk 2.17-fold higher than their added individual risks due to quetiapine inhibition of chloroquine metabolism by the CYP3A4 pathway and blockage of the OATP1B1/1B3 transmembrane transporter. Thus, the successful execution of this work will demonstrate a complete translational scope, starting with EMR-based DDI discovery, and ending with the elucidation of molecular DDI mechanisms through pharmacological experiments. Together, these preliminary data demonstrate that our translational approach is a highly feasible and extremely powerful method for clinical DDI research, likely to yield valuable insight into this emerging public health peril.