Xue L, Godden J W, Bajorath J
New Chemical Entities, Inc., Bothell, Washington 98011, USA.
J Chem Inf Comput Sci. 2000 Sep-Oct;40(5):1227-34. doi: 10.1021/ci000327j.
Combinations of 65 preferred 1D/2D molecular descriptors and 143 single structural keys were evaluated for their performance in compound classification focused on biological activity. The analysis was based on principal component analysis of descriptor combinations and facilitated by use of a genetic algorithm and different scoring functions. In these calculations, several descriptor combinations with greater than 95% prediction accuracy were identified. A set of 40 preferred structural keys was incorporated into a small binary fingerprint designed to search databases for compounds with biological activity similar to query molecules. The performance of mini-fingerprints was tested by systematic similarity search calculations in a database consisting of compounds belonging to seven biological activity classes, which had not been used to select effective descriptors. In these blind test calculations, mini-fingerprints correctly identified approximately 54% of compounds sharing similar biological activity and with 1% false positives. Thus, although the design of mini-fingerprints is conceptually simple, they perform well in activity-oriented similarity searching.
评估了65个优选的一维/二维分子描述符与143个单一结构键的组合在聚焦生物活性的化合物分类中的性能。该分析基于描述符组合的主成分分析，并借助遗传算法和不同的评分函数进行。在这些计算中，识别出了几种预测准确率高于95%的描述符组合。一组40个优选的结构键被纳入一个小型二进制指纹中，该指纹旨在在数据库中搜索与查询分子具有相似生物活性的化合物。通过在一个由属于七个生物活性类别的化合物组成的数据库中进行系统相似性搜索计算，测试了微型指纹的性能，该数据库未用于选择有效的描述符。在这些盲测计算中，微型指纹正确识别了约54%具有相似生物活性的化合物且假阳性率为1%。因此，尽管微型指纹的设计在概念上很简单，但它们在面向活性的相似性搜索中表现良好。
