Cell-Integral-Diversity Criterion: A Proposal for Minimizing Cluster Artifact in Cell-Based Selections

详细信息    查看全文

• 作者：Obdulia Rabal ; Rosalia Pascual ; José ; I. Borrell ; Jordi Teixidó
• 刊名：Journal of Chemical Information and Modeling
• 出版年：2007
• 出版时间：September 2007
• 年：2007
• 卷：47
• 期：5
• 页码：1886 - 1896
• 全文大小：632K
• 年卷期：v.47,no.5(September 2007)
• ISSN：1549-960X

文摘

Cell-based methods and the diversity integral criterion (a distance-based technique) are commonly usedapproaches for assessing the diversity of collections of compounds in terms of space coverage. The maindeficiency with cell-based methods is the arbitrariness of cell boundaries which leads to edge effects orcluster artifacts, i.e., situations in which similar molecules separated by a cell boundary yield a higher diversityscore than molecules falling within the same cell but which are less similar to each other. We describe astraightforward diversity metric based on quantifying the distance to the center of the bins resulting frompartitioning the descriptor space which aims at bypassing these artifacts. The mentioned criteria are comparedfor the diversity assessment of a set of selections carried out on three combinatorial libraries of differentcardinalities. For each method, the influence of its parameters (reference partition and number of points) ontheir efficacy is examined. Furthermore, the proposed diversity metric is also applied to designing diverselibraries for three test cases. We show that full arrays selected by minimizing the sum of distances to thecenter of the cells are formed by compounds spaced further apart than selections obtained by maximizingthe degree of cell occupancy.