Assessment of the performance of cluster analysis grouping using pharmacophores as molecular descriptors

The present work describes the results of a comparative study designed to assess the performance of using four-point pharmacophores as molecular descriptors coupled with cluster analysis as grouping technique for molecular diversity analysis. For this purpose 31 globulin binding steroids were considered as test set. Each of the molecules was investigated for the number of pharmacophores capable to fulfill, considering or not molecular flexibility, respectively. The cosine coefficient was used as similarity measure and cluster analysis as grouping technique. Specifically, two hierarchical clustering methods were used: on the one hand, the group average method and on the other, the Ward's method. The results obtained were compared with the activity of the molecules, as well as with the performance of previous diversity studies published in the literature using the same data set.