Re: [Rdkit-discuss] State of the art for shape alignment
On 12/11/2021 01:58, Paolo Tosco wrote: Hi Tim, Open3DAlign is not shape-based, it is atom-based. The score is proportional to the # of matched atoms, weighted by similarity. It will work well for homologous series of compounds with reasonable scaffold similarity, and will in general perform badly with scaffolds that are very dissimilar in size and branching. Shape-it and Align-it worked reasonably well in the past in my hands. A Google or Google Scholar search will return many well-established and more recent methods with different licenses, including but not limited to: * ShaEP * Cresset tools * ROCS * LS-Align * OptiPharm * WEGA I came upon this one the other day on github: Comparison of electrostatic potential and shape https://github.com/hesther/espsim I did not try it, but it could be one more candidate. Regards, F. Cheers, p. On Thu, Nov 11, 2021 at 4:55 PM Tim Dudgeon wrote: I'm looking into the current status of techniques that use RDKit to perform 3D alignments based on shape (e.g. not using AlignMol()) and struggling to find what the best tools are and the status of each. The Open3D align tools are relatively straightforward to use, but in my hands do not seem to give good alignments. e.g with this reference mol: I get this alignment: I also stumbled across shape-it e.g. https://iwatobipen.wordpress.com/2021/03/17/comparison-between-native-implemented-shape-align-method-of-rdkit-and-rdkit-shape-it-chemoinformatics-rdkit-shape-based-align/ But I find this very slow and most alignments seem to fail in my hands. Are there any guides to what tools work well here? Tim ___ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss ___ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss ___ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
Re: [Rdkit-discuss] State of the art for shape alignment
Hi Tim, Open3DAlign is not shape-based, it is atom-based. The score is proportional to the # of matched atoms, weighted by similarity. It will work well for homologous series of compounds with reasonable scaffold similarity, and will in general perform badly with scaffolds that are very dissimilar in size and branching. Shape-it and Align-it worked reasonably well in the past in my hands. A Google or Google Scholar search will return many well-established and more recent methods with different licenses, including but not limited to: * ShaEP * Cresset tools * ROCS * LS-Align * OptiPharm * WEGA Cheers, p. On Thu, Nov 11, 2021 at 4:55 PM Tim Dudgeon wrote: > I'm looking into the current status of techniques that use RDKit to > perform 3D alignments based on shape (e.g. not using AlignMol()) and > struggling to find what the best tools are and the status of each. > > The Open3D align tools are relatively straightforward to use, but in my > hands do not seem to give good alignments. e.g with this reference mol: > [image: image.png] > I get this alignment: > [image: image.png] > > I also stumbled across shape-it e.g. > https://iwatobipen.wordpress.com/2021/03/17/comparison-between-native-implemented-shape-align-method-of-rdkit-and-rdkit-shape-it-chemoinformatics-rdkit-shape-based-align/ > But I find this very slow and most alignments seem to fail in my hands. > > Are there any guides to what tools work well here? > > Tim > > > > ___ > Rdkit-discuss mailing list > Rdkit-discuss@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/rdkit-discuss > ___ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
[Rdkit-discuss] State of the art for shape alignment
I'm looking into the current status of techniques that use RDKit to perform 3D alignments based on shape (e.g. not using AlignMol()) and struggling to find what the best tools are and the status of each. The Open3D align tools are relatively straightforward to use, but in my hands do not seem to give good alignments. e.g with this reference mol: [image: image.png] I get this alignment: [image: image.png] I also stumbled across shape-it e.g. https://iwatobipen.wordpress.com/2021/03/17/comparison-between-native-implemented-shape-align-method-of-rdkit-and-rdkit-shape-it-chemoinformatics-rdkit-shape-based-align/ But I find this very slow and most alignments seem to fail in my hands. Are there any guides to what tools work well here? Tim ___ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss