Calculating target bias in small molecules for library design

描述(由申请人提供)：目标：目标是开发一种计算系统来预测小分子偏向的目标。该工具将优化偏向文库，针对特定目标进行测试的优先顺序，在表型屏幕中识别目标，并允许用户轻松地从化合物转移到目标。意义：该工具将使研究人员能够轻松识别调节目标的小分子。它还将允许他们构建偏向于此类目标的库。该工具的用户包括希望为目标快速访问生物分子的研究人员，对建立偏向于此类目标的库感兴趣的公司，以及希望扩大他们销售的库的有用性的供应商。动机是一种可以迅速将化学物质带给生物学家的工具。理论/背景：我们和其他人已经证明，可以预测以前未知的药物的“脱靶”。我们的相似性集成方法(SEA)使用化学结构来预测任何分子的目标，方法是将该分子与蛋白质目标注释的配体中发现的相似模式相关联。虽然许多商业上可用的分子没有靶标连接，但平均而言，对于任何给定的供应商，我们预测大约有T个这样的分子。目标1.开发一种将目标分配给大型图书馆中的小分子的服务。我们将建立一个系统，允许用户查询小分子文库，提出两个问题中的一个。A.对于一个给定的靶点，这个文库中的哪些小分子有望对其进行调制？B.对于一个给定的文库，它的分子最有可能作用于什么靶点？里程碑：该系统的基本特征已经存在，并在使用SEA预测超过25种药物的新靶点方面证明了原理。在这里，我们扩大了方法，使非专家能够在大型复合库上使用。有两个务实的里程碑。I.通过实验测试预测的化合物-目标关联，以证明其可行性。二、开发可与供应商目录集成的网络界面。目标2.开发一种可以针对目标覆盖范围优化图书馆的服务。A.给定特定的文库大小，我们能否对其进行优化以涵盖最大数量的药理相关靶点？B.相应地，我们能为最大的靶标偏差优化文库吗？里程碑：此方法基本上已在手中，但尚未经过测试。我们将与我们的商业合作伙伴合作，为目标覆盖优化他们的文库，通过预测和实验测试目标活性的分子来证明这一点(六个月)。目的3.预测活性化合物在表型和动物实验中的靶标。制药研究和化学生物学的一个最新进展是重新发现了表型，甚至整个生物体筛选化合物活性，经常使用携带已知疾病相关基因的改良细胞或生物体。一个核心挑战是确定与观察到的表型有关的分子靶标。事实上，我们在制药领域的合作伙伴表示，他们可以在细胞、组织、器官或动物模型中拥有数百种具有活性的相关化合物，但不知道实际的分子靶标，从而限制了对机理的理解和优化。我们将使用SEA来预测这些活性分子的目标。里程碑：与制药合作伙伴合作，我们将预测具有动物模型活性的化合物系列的靶标，但其靶标仍未知。最初，我们预计将在受体结合试验中测试10种这样的化合物。尽管这些目标雄心勃勃，但广泛的初步结果表明，它们是可行的。 与公共健康相关：研究人员经常测试大量化合物，以发现新药的化学起点。在这项提案中，我们的重点是提供一种工具来预测哪些商业上可获得的化合物最有可能在治疗靶点上发挥作用。这节省了时间和金钱，因为减少了研究人员在早期药物发现过程中需要测试的化合物数量。

DESCRIPTION (provided by applicant): Goals: The goal is to develop a computational system to predict the targets toward which small molecules are biased. The tool will optimize biased libraries, prioritize them for testing against particular targets, identify targets in phenotypic screens, and allow users to easily move from compounds to targets. Significance: The tool would allow investigators to easily identify small molecules to modulate a target. It also would allow them to build libraries that are biased toward such targets. Users of the tool include investigators looking to quickly access biological molecules for their targets, companies interested in building libraries biased toward such targets, and vendors looking to expand the usefulness of libraries that they sell. A motivation is a tool that can rapidly bring chemical matter to biologists. Theory/Background: We and others have shown that it is possible to predict previously unknown "off- targets" for drugs. Our Similarity Ensemble Approach (SEA) uses chemical structure to predict targets for any molecule by associating that molecule with similar patterns found in ligands annotated to protein targets. Whereas many commercially available molecules have no target linkage, on average for any given vendor we predict that about T of them do. Aim 1. To develop a service to assign targets to small molecules in large libraries. We will build a system that will allow users to query a library of small molecules, asking one of two questions. a. For a given target, what small molecules in this library would be expected to modulate it? b. For a given library, on what targets are its molecules most likely to work? Milestone: The essential features of this system exist, and proof of principle has been demonstrated in predicting new targets for over 25 drugs using SEA. Here we broaden the method to enable use by non- experts on large compound libraries. There are two pragmatic milestones. i. Experimentally testing predicted compound-target associations to demonstrate feasibility. ii. Development of a web-interface that can be integrated with vendor catalogs. Aim 2. To develop a service that can optimize a library for target coverage. a. Given a particular library size, can we optimize it to cover the maximum number of pharmacologically relevant targets? b. Correspondingly, can we optimize a library for the maximum target bias? Milestone: This method is essentially in hand, but has not been tested. We will work with our commercial partners to optimize their library for target coverage, demonstrating this by predicting and experimentally testing molecules for target activity (six months). Aim 3. To predict the targets for compounds active in phenotypic and animal assays. A recent development in pharmaceutical research and chemical biology has been the rediscovery of phenotypic, even whole organism screens for compound activity, frequently using modified cells or organisms that carry a known disease-associated genotype. A core challenge is identifying the molecular targets responsible for the observed phenotype. Indeed, our partners in pharma indicate that they can have hundreds of related compounds with activity in cells, tissues, organs or animal models, but do not know the actual molecular targets, limiting mechanistic understanding and optimization. We will use SEA to predict targets for these active molecules. Milestone: In collaboration with a pharmaceutical partner, we will predict targets for a compound series with animal model activity, but for which targets remain unknown. Initially we anticipate testing ten such compounds in receptor-binding assays. Whereas these goals are ambitious, extensive preliminary results suggest that they are feasible. PUBLIC HEALTH RELEVANCE: Investigators often test large numbers of compounds to discover a chemical starting point for a new drug. In this proposal we focus on providing a tool to predict which commercially-available compounds are most likely to be active at therapeutic targets. This saves time and money by reducing the number of compounds investigators need to test during early-stage drug discovery.