Similarity searching in files of three-dimensional chemical structures: Comparison of fragment-based measures of shape similarity

This paper compares several fragment-based measures that can be used to quantify the degree of similarity between pairs of three-dimensional (3-D) chemical structures. The fragments that are considered contain two, three, or four atoms and encode distance information, angular information or both but do not involve chemical information such as atomic type. The effectiveness of the various measures is compared using eight literature datasets for which biological-activity data and calculated 3-D structures are available, and a set of carbohydrate structures from the Cambridge Structural Database that have been classified into eight distinct groups. Similarity searches on these datasets suggest that the four-atom fragments are the most effective.