Clustering Using Flexible 3D Alignment Based Similarity

poster · 6 years ago
by Adrián Kalászi, Gábor Imre, Miklós Szabó (ChemAxon)
Screen JKlustor
There are several approaches to cluster chemical structures. Classical 2D descriptor (for example chemical fingerprints or ECFP) and structure based methods are the most widely spread. Considering 3D information such as conformers, 3D pharmacophore maps or molecular shapes can give more insight thus can facilitate further natural interpretability of the results. ChemAxon’s alignment tool provides automatic 3D shape based flexible alignment of small molecules. The shape similarity score calculated for the best fit can serve as a basis for similarity based clustering. Integration of this approach into JKlustor suite, clustering results and comparison with 2D methods for various small molecule sets are presented. Download poster