one ring too much

Hi - got a compound that crashes the conformational sampler with a "Clean failed" error:





C[N+]1(C)[C@@H]2C[C@@H](C[C@@H]1[C@@H]3O[C@@H]23)OC(=O)[C@H](CO)c4ccccc4





The tool in question is Odon's Conformers.java, with some changes i don't think they matter - but I join the file anyway.





Cheers!

Hi Dragos,





Sorry, I forgot to inform you that the conformer related API has changed somewhat. One possible way is using the changed API from Clean3D or use the new plugin, which is only in the alpha release yet:


http://www.chemaxon.com/shared/alpha/beansonly/





This was made by Zsolt Mohácsi and he was kind enough to provide a simple code for using it:

Code:

// read the input molecule from a String (molstr) Molecule mol = MolImporter.importMol(molstr); // calculate the conformers // create conformer plugin object ConformerPlugin cplugin = new ConformerPlugin(); // set input molecule (we use the output of the mmsplugin as input) cplugin.setMolecule(mol); // set parameters for calculation cplugin.setMaxNumberOfConformers(400); cplugin.setPrehydrogenize(true); cplugin.setTimelimit(900); // run the calculation cplugin.run(); // get results Molecule[] conformers = cplugin.getConformers(); int conformerCount = cplugin.getConformerCount(); Molecule m; String mstr; for (int i = 0; i < conformerCount; ++i) {     m = conformers[i]; // same as m = cplugin.getConformer(i);     // do something with molecule ...     // or get result as String in "mol" format     // (Note: 3D format required for stereo information)     mstr = m.toFormat("sdf");     System.out.println(mstr); }

If you prefer using the new API in Clean3D, just let me know and I can provide a simple code for it also.





Best wishes,





Ödön

dragos wrote:

Hi - got a compound that crashes the conformational sampler with a "Clean failed" error: C[N+]1(C)[C@@H]2C[C@@H](C[C@@H]1[C@@H]3O[C@@H]23)OC(=O)[C@H](CO)c4ccccc4 The tool in question is Odon's Conformers.java, with some changes i don't think they matter - but I join the file anyway. Cheers!

Hi Dragos,





I might also have a simpler answer. Are you sure about the chirality of atoms #9 and #11? (8&10 if the numbering goes from 0). Could you provide a single valid structure for the ring system with these stereo criteria?





Best wishes,





Ödön

dragos wrote:

Hi - got a compound that crashes the conformational sampler with a "Clean failed" error: C[N+]1(C)[C@@H]2C[C@@H](C[C@@H]1[C@@H]3O[C@@H]23)OC(=O)[C@H](CO)c4ccccc4 The tool in question is Odon's Conformers.java, with some changes i don't think they matter - but I join the file anyway. Cheers!

Hi Dragos,





I double checked the structure and it is indeed a valid structure with some strain. I do not know yet if it is a bug or problem of the Dreiding forcefield itself but the implementation in Marvin always rotates it to invalid chirality even if started from a valid, partially optimized conformer. We need some more work to dig this out, thanks for the structure!





Best wishes,





Ödön

I had a dummy attempt to copy the smiles into Msketch and ask for a "fast" clean3D - and it worked! The "fine", though, would not work until you first do the fast... can we emulate that in with the API, or do you think it ain't a good idea (would the fast mode with other molecules produce such distorted geometries that the fine won't be able to repair - from my previous experience I suppose that the clean3D tool somehow takes input geometry into account if provided?)





God Save the Conformers!

Hi Dragos,





If you do a Clean2D after a "fast" Clean3D for this molecule you will see that one of the eppxi carbons' chirality changes from the original. This is what happen when you use the "fast" method. It does not care much. You can emulate this behavior via adding "[acceptfault]" to the option string. Also have a look to the attached dft optimized partial structure.





Best wishes,





Ödön

dragos wrote:

I had a dummy attempt to copy the smiles into Msketch and ask for a "fast" clean3D - and it worked! The "fine", though, would not work until you first do the fast... can we emulate that in with the API, or do you think it ain't a good idea (would the fast mode with other molecules produce such distorted geometries that the fine won't be able to repair - from my previous experience I suppose that the clean3D tool somehow takes input geometry into account if provided?) God Save the Conformers!