LEARNING DRUG SPECIFICITY FROM PROTEIN FAMILIES

我们建议开发和应用的方法对接配体片段到家庭的蛋白质，两个特定的目的。其中第一个是开发方法来识别特定家族成员所特有的模式，以避免在设计高度特异性药物时发生交叉反应。第二是开发识别相似性模式的方法，当已知靶蛋白的多个等位基因时，可以使用该方法来帮助避免感染性疾病生物体的耐药菌株的发展。这项工作是我们目前努力的延伸和建议，由单独的第二阶段SBIR资助，以开发比较蛋白质家族内和家族间表面特征的方法。拟议的工作需要开发（a）新的对接方法，可直接使用家族内的特征比较;（B）基于无监督学习方法的新的数据分割工具，以评估从单个家族成员的对接结果中获得的相似性和差异;以及（c）开发基于向量支持机的新分类方法和其他优化分离函数的方法，以正确区分从族中不正确停靠的结构停靠。虽然这项工作是探索性的，但如果成功的话，它可以提供一套强大的药物发现工具，以利用基因组学和结构基因组学研究中出现的越来越丰富的蛋白质序列和结构信息。开发的软件将提供销售，并应用于现有的晶体结构，以开发可立即用于药物发现的专门数据库。拟议的商业应用：赠款的目标是解决当前药物发现中的关键问题，即特异性，交叉反应性和避免传染病生物体对新药的耐药性。 这些方法通过从基因组学和结构基因组学努力中积累知识而得以实现。我们相信，这些和类似的方法将是必要的，以获得这种非凡的知识积累所带来的商业和社会效益。

We propose to develop and apply methods for docking ligand fragments into families of proteins, for two specific purposes. The first of these is to develop methods to identify patterns which are unique to a specific family member, to avoid cross-reactivity in the design of highly specific drugs. The second is to develop methods to identify patterns of similarity, which can be used when multiple alleles of a target protein are known to help avoid the development of resistant strains of infectious disease organisms. The work is an extension of and suggested by our current efforts, funded by a separate Phase II SBIR, to develop methods for comparing surface features of proteins within and across families. The work proposed entails the development of (a) new docking methods which can directly use feature comparisons within families; (b) novel data segmentation tools based on unsupervised learning methods to assess similarity and differences obtained from docking results on individual family members; and (c) the development of new classification methods based on Vector Support Machines and other methods for optimizing separation functions to distinguish correctly docked from incorrectly docked structures within the families. While the work is exploratory, if successful it can provide a focussed set of powerful drug-discovery tools to take advantage of the increasingly rich amount of information on protein sequence and structure emerging from genomics and structural genomics efforts. Software developed will be offered for sale, and applied to available crystal structures to develop specialized databases which can be of immediate use in drug discovery. PROPOSED COMMERCIAL APPLICATIONS: The goals of the grant are to address key issues in current drug discovery, namely, specificity, cross-reactivity, and avoiding resistance in infectious disease organisms to new drugs. The methods are strongly enabled by the accumulation of knowledge from genomics and structural genomics efforts. We believe that these and similar methods will be necessary to reap the commercial and sociological benefit promised by this extraordinary accumulation of knowledge.