HIGH PERFORMANCE COMPUTING FOR DRUG DISCOVERY

用于药物发现的高性能计算

• 批准号: 7723329
• 负责人: Weifan Zheng
• 金额: $ 0.05万
• 依托单位: CARNEGIE-MELLON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2008
• 资助国家: 美国
• 起止时间: 2008-08-01 至 2009-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7723329
• 关键词: Accounting; Address; Algorithms; Award; Binding Sites; Biological; Biological Process; Biotechnology; Chemical Structure; Chemicals; Collection; Complement; Computer Retrieval of Information on Scientific Projects Database; Computer Vision Systems; Computing Methodologies; Databases; Development; Docking; Education; Elements; Funding; Genomics; Goals; Grant; HIV; High Performance Computing; Historically Black Colleges and Universities; Image; Institution; Investigation; Lead; Libraries; Ligands; Link; Measures; Membrane Proteins; Methods; Molecular; Molecular Structure; North Carolina; Nuclear Hormone Receptors; Pathway interactions; Preclinical Drug Evaluation; Process; Proteins; RNA-Directed DNA Polymerase; Rate; Research; Research Personnel; Resources; Retrieval; Screening procedure; Set protein; Shapes; Side; Source; Structure; Techniques; Trees; United States National Institutes of Health; Universities; Validation; Work; abstracting; base; chemical resource; cheminformatics; computing resources; drug discovery; fitness; improved; inhibitor/antagonist; innovation; interest; neglect; novel; pharmacophore; phosphoric diester hydrolase; programs; protein structure; restraint; scaffold; shape analysis; size; small molecule; structural genomics; tool; virtual

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. ABSTRACT The BRITE Center has been established by North Carolina as the premier Center of Excellence in biotechnology and drug discovery within NC Central University, an HBCU (Historically Black College/University) to advance the education and research in this promising field. As a component of this Center, we are engaged in the research and application of cheminformatics and computational drug discovery tools. Specifically, we aim to conduct the following work using the computing resources of DAC-TeraGrid if we are granted the award. 1. Virtual Computational Drug Screening Background: Current pharmacophore based computational virtual screening methods often ignore the intricate details of the binding site shapes and focus only on the key pharmacophore elements. Thus, they often miss critically important information during the virtual screening process, resulting in many false positives. For example, large molecules with multiple side chains attached to a central scaffold may be selected as false positives, simply because their core structures have the required pharmacophore elements. This situation may be alleviated if binding site excluded volumes are considered. However, the efficiency of such a process is dramatically reduced due to the frequent checks for clashes with the excluded volumes. Thus, there is a great need to increase the efficiency of such method for database searching while taking into account the volume restraints that help reduce the false positive rate. The Computational Method: To address the pitfalls and shortcomings of current pharmacophore methods, we have engaged in the study of a novel structure-based shape pharmacophore method for virtual screening. It takes advantage of a computational geometry algorithm (Delauney tessellation / alpha-shape analysis) to detect the binding site atoms and generate a negative image of the binding site, which complements the binding site shape. This negative image is represented by a set of spheres of different sizes. There are multiple ways to represent the overall shape of this set of spheres. Currently, we use the OEChem Shape library functions to represent it due to the well-known efficiency of the shape matching algorithm. Other computer vision method is also being explored to help improve the accuracy and efficiency of the shape matching process. The innovative aspect of our method comes from the fact that a rigorous computational geometry algorithm has been used to detect the binding site atoms, and a deterministic process to generate the matching image, as well as the representation of this image with OEChem Shape functions. Additional development will include adding more advanced computer vision techniques for shape matching and recognition. The Computing Plan: The above method will be applied to several selected targets: PDE (phosphodiesterases), HIV reverse transcriptase, nuclear hormone receptors and a few other targets. Validation of the method will be performed based on the information on known inhibitor/ligands in the WOMBAT database, which contains over 50,000 molecular structures. The computing intensive *conformational analysis* of these 50,000 molecules as well as *shape matching* with each of the above targets will be conducted in this proposal. The retrieval rate of known active compounds will be compared with ligand-based shape matching (ROCS) as well as FRED docking program (both are computationally intensive as well). 2. Biologically Relevant Molecular Diversity Measure Background: In diversity analysis of compound libraries, most methods look at only the self-dissimilarity among the compound structures, neglecting known information about the biological space revealed by structural genomics projects. This may lead to a hugely diverse set of compounds which may not have any biological effect on most targets. We use the structure-based shape pharmacophore method (see section 1) to evaluate the relevance of a given compounds by comparing its shape with a PANEL of shapes derived from a selected set of protein structures. The RATIONALE behind this is that the shapes of functional pockets on protein surface or binding sites are often the determinant for a molecules biological functions. By using a PANEL of shapes derived from biologically relevant protein pockets, we can evaluate whether a molecule might be biologically active. Such a method is extremely useful especially in the context of the NIH Roadmap initiative where finding chemical probes for biological pathways is the main task. The Computing Plan: a HitMap where the fitness of each molecule in a collection with each of the PANEL shapes will be evaluated. Such a HitMap across a collection of protein structures (>100) will essentially build tentative links between protein structures and small molecule collections. For example, the HitMap for the PubChem molecules will be a useful resource for Chemical Genomics investigations where researchers will have a holistic view of what a molecule might do to other proteins in addition to the target of interest. This grant would greatly enhance our computational effort to obtain such a chemical genomics tree (CGTree) that links chemical structures to their potential biological targets, and ultimately help advance the goal of NIH Roadmap on Chemical Genomics Research.

该子项目是利用该技术的众多研究子项目之一 资源由 NIH/NCRR 资助的中心拨款提供。子项目和 研究者 (PI) 可能已从 NIH 的另一个来源获得主要资金， 因此可以在其他 CRISP 条目中表示。列出的机构是 对于中心来说，它不一定是研究者的机构。 摘要 BRITE 中心由北卡罗来纳州建立，作为北卡罗莱纳州中央大学（HBCU（历史上的黑人学院/大学））内生物技术和药物发现方面首屈一指的卓越中心，旨在推进这一前景广阔的领域的教育和研究。作为该中心的组成部分，我们从事化学信息学和计算药物发现工具的研究和应用。具体来说，如果获奖，我们的目标是利用DAC-TeraGrid的计算资源开展以下工作。 1. 虚拟计算药物筛选背景：当前基于药效团的计算虚拟筛选方法常常忽略结合位点形状的复杂细节，而仅关注关键药效团元素。因此，他们经常在虚拟筛选过程中错过至关重要的信息，从而导致许多误报。例如，具有连接到中心支架的多个侧链的大分子可能被选择作为假阳性，仅仅因为它们的核心结构具有所需的药效团元素。如果考虑结合位点排除体积，这种情况可能会得到缓解。然而，由于频繁检查与排除体积的冲突，这种过程的效率大大降低。因此，非常需要提高这种数据库搜索方法的效率，同时考虑到有助于降低误报率的数量限制。计算方法：为了解决当前药效团方法的缺陷和缺点，我们研究了一种用于虚拟筛选的新型基于结构的形状药效团方法。它利用计算几何算法（德劳尼曲面细分/α形状分析）来检测结合位点原子并生成结合位点的负像，以补充结合位点形状。该负像由一组不同大小的球体表示。有多种方法可以表示这组球体的整体形状。目前，由于形状匹配算法的效率众所周知，我们使用 OEChem Shape 库函数来表示它。其他计算机视觉方法也正在探索中，以帮助提高形状匹配过程的准确性和效率。我们方法的创新之处在于，使用严格的计算几何算法来检测结合位点原子，并使用确定性过程来生成匹配图像，以及使用 OEChem Shape 函数表示该图像。额外的开发将包括添加更先进的计算机视觉技术以进行形状匹配和识别。计算计划：上述方法将应用于几个选定的目标：PDE（磷酸二酯酶）、HIV逆转录酶、核激素受体和一些其他目标。该方法的验证将根据 WOMBAT 数据库中已知抑制剂/配体的信息进行，该数据库包含超过 50,000 个分子结构。该提案将对这 50,000 个分子进行计算密集型“构象分析”以及与上述每个目标的“形状匹配”。已知活性化合物的检索率将与基于配体的形状匹配 (ROCS) 以及 FRED 对接程序（两者的计算量也很大）进行比较。 2. 生物学相关的分子多样性测量背景：在化合物库的多样性分析中，大多数方法只考虑化合物结构之间的自相似性，忽略了结构基因组学项目揭示的有关生物空间的已知信息。这可能会产生极其多样化的化合物，这些化合物可能对大多数目标没有任何生物效应。我们使用基于结构的形状药效团方法（参见第 1 节），通过将给定化合物的形状与源自一组选定的蛋白质结构的形状面板进行比较来评估给定化合物的相关性。这背后的基本原理是蛋白质表面或结合位点上的功能袋的形状通常是分子生物功能的决定因素。通过使用源自生物相关蛋白质袋的形状面板，我们可以评估分子是否具有生物活性。这种方法非常有用，特别是在 NIH 路线图倡议的背景下，其中寻找生物途径的化学探针是主要任务。计算计划：一个 HitMap，其中将评估集合中每个分子与每个 PANEL 形状的适合度。这样一个跨越蛋白质结构集合（>100）的 HitMap 本质上将在蛋白质结构和小分子集合之间建立暂时的联系。例如，PubChem 分子的 HitMap 将成为化学基因组学研究的有用资源，研究人员将全面了解除了感兴趣的靶标之外，分子可能对其他蛋白质产生的作用。这笔资助将极大地增强我们的计算工作量，以获得将化学结构与其潜在生物靶标联系起来的化学基因组树（CGTree），并最终帮助推进 NIH 化学基因组学研究路线图的目标。