一种基于三维结构的分子相似性新定义

相似性模型在化学诸多领域的广泛应用激发了许多研究者努力探讨分子相似性更中肯的计算方法。我们根据语义模型定义了一种新的分子相似性指数。这种方法首先将分子结构安某种方式拆分，得到可以表征分子结构特征的结构碎片（子结构），它们构成了描述分子结构相似性的度量空间，再根据这些碎片的语义关系定义了分子相似。通过１６个含５个碳的醇之间的分子相似性的计算举例说明这个新方法。

A NEW DEFINITION OF MOLECULAR SIMILARITY BASED ON SEMANTIC RELATION OF 3D STRUCTURES

The usefulness of similarity models in many fields of chemistry stimulated a number of the researchers to make great efforts on finding relevant definitions of molecular similarity. A new definition of molecular similarity is proposed, based on the semantic relation of 3D structures of molecules. First, we divide the 3D structures into fragments (3D substructures) which can characterize the structural features of molecule. These fragments constitute a metric space in which the molecular similarity can be measured. Then we define the molecular similarity based on semantic relation between these fragments in the space. This novel method is illustrated by calculation of the molecular similarities of 16 five\|carbon alcohols.