Robust computational framework for predictive ADME-Tox modeling

描述(由申请人提供)： 这项建议寻求建立一个普遍适用和稳健的预测ADME-TOX模型框架，该框架基于严格的定量结构、活性/性质关系(QSAR/QSPR)建模。在我们多年来在QSPR方法学开发和对实验数据集的应用方面的研究过程中，该框架得到了完善，导致了新的分析方法、描述符、模型验证方案、整体QSPR工作流程设计和多个端点研究。这项建议侧重于设计优化的QSPR协议，以开发至关重要的ADME-Tox性质的可靠预测值。ADME的特性将包括但不限于水溶性、膜渗透性、P450代谢抑制和诱导、代谢稳定性、人体肠道吸收、生物利用度、转运体和PK数据；将探索对人类健康至关重要的各种毒理学终点；它们可从最近开发和标准化毒性数据的倡议中获得，如美国FDA、NIEHS和EPA DSS-TOX和其他数据库项目。该项目的最终目标是通过基于网络的预测性ADME-TOX门户与研究社区共享建模软件和专门的预测值。项目目标将通过同时开发QSPR方法(具体目标1)、建立高预测性、健壮的已知ADME-TOX属性的QSPR模型(具体目标2)以及通过专门的门户网站部署建模软件和单个预测者(具体目标3)来实现。为了实现这个项目的目标，专注于开发和交付专门的工具和严格的预报器，我们组建了一个研究团队，主要由高级调查人员组成，他们在计算药物发现、实验毒理学、统计建模以及软件开发和集成领域具有互补的技能和成就记录；团队中的两名成员在过渡到学术界之前最近有过工业经验。据我们所知，这项提议的结果将导致第一个公开可用的Silico ADME-Tox建模框架和预报器，可供研究界用于分析任何一组化学物质(即虚拟和真实化合物组)。该框架将对化合物的优先顺序、化学库设计以及临床前和临床开发的候选人选择产生重大影响。

DESCRIPTION (provided by applicant): This proposal seeks to establish a universally applicable and robust predictive ADME-Tox modeling framework based on rigorous Quantitative Structure Activity/Property Relationships (QSAR/QSPR) modeling. The framework has been refined in the course of many years of our research in the areas of QSPR methodology development and application to experimental datasets that led to novel analytical approaches, descriptors, model validation schemes, overall QSPR workflow design, and multiple end-point studies. This proposal focuses on the design of optimized QSPR protocols for the development of reliable predictors of critically important ADME-Tox properties. The ADME properties will include, but not limited to, water solubility, membrane permeability, P450 metabolism inhibition and induction, metabolic stability, human intestinal absorption, bioavailability, transporters and PK data; a variety of toxicological end-points vital to human health will be explored; they are available from recent initiatives on development and standardization of toxicity data, such as the US FDA, NIEHS, and EPA DSS-Tox and other database projects. The ultimate goal of this project is sharing both modeling software and specialized predictors with the research community via a web-based Predictive ADME-Tox Portal. The project objectives will be achieved via concurrent development of QSPR methodology (Specific Aim 1), building highly predictive, robust QSPR models of known ADME-Tox properties (Specific Aim 2), and the deployment of both modeling software and individual predictors via a specialized web-portal (Specific Aim 3). To achieve the goals of this project focusing on the development and delivery of specialized tools and rigorous predictors, we have assembled a research team of mostly senior investigators with complimentary skills and track records of accomplishment in the areas of computational drug discovery, experimental toxicology, statistical modeling, and software development and integration; two of the team members have had recent industrial experience before transitioning to academia. To the best of our knowledge, the results of this proposal will lead to the first publicly available in silico ADME-Tox modeling framework and predictors that can be used by the research community to analyze any set of chemicals (i.e., virtual and real compound sets). The framework will have a significant impact on compound prioritization, chemical library design, and candidate selection for preclinical and clinical development.