Department of Chemistry, AstraZeneca R&D Charnwood, Bakewell Road, Loughborough, Leicestershire LE11 5RH, United Kingdom.
J Chem Inf Model. 2010 Aug 23;50(8):1350-7. doi: 10.1021/ci100084s.
An algorithm to automatically identify and extract matched molecular pairs from a collection of compounds has been developed, allowing the learning associated with each molecular transformation to be readily exploited in drug discovery projects. Here, we present the application to an example data set of 11 histone deacetylase inhibitors. The matched pairs were identified, and corresponding differences in activity and lipophilicity were recorded. These property differences were associated with the chemical transformations encoded in the SMIRKS reaction notation. The transformations identified a subseries with the optimal balance of these two parameters. Enumeration of a virtual library of compounds using the extracted transformations identified two additional compounds initially excluded from the analysis with an accurate estimation of their biological activity. We describe how the WizePairZ system can be used to archive and apply medicinal chemistry knowledge from one drug discovery project to another as well as identify common bioisosteres.
已经开发出一种算法，可以自动识别和提取化合物集合中的匹配分子对，从而可以轻松利用与每个分子转化相关的学习成果来开展药物发现项目。在这里，我们将该算法应用于一个包含 11 种组蛋白去乙酰化酶抑制剂的示例数据集。确定了匹配对，并记录了相应的活性和脂溶性差异。这些性质差异与 SMIRKS 反应符号中编码的化学转化相关联。确定的转化识别出具有这两个参数最佳平衡的亚系列。使用提取的转化对虚拟化合物库进行枚举，确定了另外两个最初被排除在分析之外的化合物，它们的生物活性得到了准确估计。我们描述了如何使用 WizePairZ 系统将一个药物发现项目中的药物化学知识存档并应用于另一个项目，以及识别常见的生物等排体。
