On 04/05/2011 03:24 PM, Geoffrey Hutchison wrote: > Moreover, I'd be happy to donate bond perception code to RDKit, but > these things only really work well when you have 3D coordinates. > (I've tried many implementations.)
Yes, I fully agree. We are basically hoping we can get away with much less sophisticated algorithms for the bulk of our compounds. The 20% of the effort that gives us 80% of the results. Even if only as a start. > But I can't quite figure out how you'd want to "perceive" bond types > without any obvious information. For something trivial like acetic acid, the connectivity suffices for a chemist to infer the bond types based on a small set of very trivial rules. But it is true that we will probably very quickly get to the point where 3D coordinates are required for successful bond-type perception. Regarding donating code: that is up to you and Greg to agree on, of course. I cannot promise I can help, but in principle I am proficient in C++ and I thoroughly studied the strengths and weaknesses of the bond-type perception algorithms that were in use at the Cambridge Crystallographic Data Centre when I worked there. Best wishes, -- Dr Jacco van de Streek Senior Scientist Avant-garde Materials Simulation Freiburg im Breisgau, Germany ------------------------------------------------------------------------------ Xperia(TM) PLAY It's a major breakthrough. An authentic gaming smartphone on the nation's most reliable network. And it wants your games. http://p.sf.net/sfu/verizon-sfdev _______________________________________________ Rdkit-discuss mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
