A Web-Based Automatic Virtual Screening System

摘要虚拟筛选是利用配体和蛋白质结构发现先导分子最实用的方法。不幸的是，大多数研究人员都无法获得基于配体和对接的技术。第一阶段的一个关键成果是锌数据库，它降低了通过公共访问3D放映库对接的进入门槛。相似集成方法（SEA）也是在第一阶段发展起来的，在基于配体的目标识别方面显示出了早期的希望。然而，对于大多数研究者来说，虚拟筛查仍然很难使用。为了进一步降低这些障碍，我们将开发供一般社区使用的数据库和自动化工具，并调查它们在概念验证研究中的有用性。具体目标是：1。开发源自虚拟筛选并启用虚拟筛选的数据库。答：我们将开发一个预先计算的对接点数据库，可以简单地查找和购买大约1000个蛋白质靶点。这将依赖于对接、命中评估和目标间比较的自动化工具（目标2）。我们还将改进第一期开发的虚拟筛选数据库。这些措施包括：B.扩展锌，添加更多的市售化合物和改善其结构。C.提高DUD的稳健性，DUD是虚拟筛选的通用基准集。D.扩展第一期开发的用于蛋白质功能预测的高能中间体（HEI）数据库。2. 为基于配体和基于蛋白质的虚拟筛选创建简单的基于网络的工具。我们将开发和完善两个基于网络的工具，使非专业人员能够为他们的目标发现配体。a .对于基于结构的对接，一个简单的web界面来引导用户对接，选择参数，校准模型，并管理我们集群上的计算。我们将开发自动化工具来评估对接结果的可靠性。B.第二种虚拟筛选工具是基于配体的，用于当靶标结构未知但有许多配体可用时，或者当人们想要探索已知药物或试剂的替代靶标时。我们进一步发展了一种新的化学信息学方法SEA，用于预测靶标关系和脱靶效应。这种方法在识别有趣的多药理学方面已经取得了早期的成功，我们自己也将使用它来预测和测试50到100种FDA药物的脱靶、临床相关效应，并确定约10%的FDA药物的靶标，其靶标未知。

DESCRIPTION (provided by applicant): Abstract Virtual screening is the most practical method to leverage ligand and protein structures for lead discovery. Unfortunately, both ligand-based and docking techniques are inaccessible to most investigators. A key result from the first period, the ZINC database, has lowered the barrier to entry for docking through public access 3D screening libraries. The Similarity Ensemble Approach (SEA), also developed in the first period, has shown early promise for ligand-based target identification. Still, virtual screening remains difficult to use for most investigators. To lower these barriers still further we will develop databases and automated tools for use by the general community, and investigate their usefulness in proof-of-concept studies. The specific aims are: 1. To develop databases that derive from and enable virtual screening. A. We will develop a database of pre-calculated docking hits that can simply be looked up and purchased for about 1,000 protein targets. This will rely on automated tools for docking, hit evaluation, and comparisons among targets (aim 2). We will also improve databases for virtual screening developed in the first period. These include: B. Expanding ZINC, adding more commercially available compounds and improving the structures represented in it. C. Improving the robustness of DUD, a general benchmarking set for virtual screening. D. Expanding the database of high energy intermediates (HEI) developed in the first period for protein function prediction. 2. To create simple web-based tools for ligand-based and protein-based virtual screening. We will develop and refine two web-based tools to enable non-specialists to discover ligands for their targets. A. For structure-based docking, a simple-looking web-interface to docking that guides the user, selects parameters, calibrates the model, and manages the calculation on our cluster. We will develop automated tools to evaluate the reliability of docking results. B. The second virtual screening tool is ligand based, for use when the structure of the target is unknown but many ligands are available, or when one wants to explore alternate targets for a known drug or reagent. We further develop a novel cheminformatic method SEA introduced in the last period to predict target relationships and off-target effects. This approach has had precocious success in identifying interesting polypharmacology, and we will also use it ourselves to predict- and-test off-target, clinically relevant effects of 50 to 100 FDA drugs, and identify the targets of the ~10% of FDA drugs for which a target is unknown. PUBLIC HEALTH RELEVANCE: Virtual screening is widely used to discover new molecular leads for drug discovery and reagents to understand biological processes. Unfortunately, the technique remains difficult to use, and has thus been restricted to a few expert laboratories, limiting its usefulness. In this proposal, we create databases and tools to bring virtual screening to a wide biological audience, much expanding its impact and usefulness.