Can anyone recommend a protocol for building the 3D structure of a small molecule de novo, given only a 2D rendering? So far my idea has been to build the molecule using the fragment builder in PyMol, then make conformers using Frog2. But I have no idea how realistic the starting structure is from PyMol and how exhaustively the conformers are sampled in Frog2.

The downstream application of this would be to dock the ligand into a protein binding pocket. If anyone has a more rigorous protocol for making sure all the conformational space is fully sampled, it would be greatly appreciated!


Most universities have a ChemDraw site lincence, so you may want to check that for drawing a non–3D-embedded molecule. Alternatively there are many other programs. PyMOL is not really a good choice. Also there are online drawing tools everywhere, where you can download a SMILES (see below).

Chemical formula is the wrong way to go, unless you are "enumerating". The chemical formula of glucose is C6H12O6. This is the same for all hexoses. What you want is a SMILES. This string gives the connectivity of the atoms. For D-glucose it's OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O, the @ symbols are chiral centres. In Cheminformatics making a list of all possible compound with n-carbons is a big deal, this is called enumeration and you likely don't have to reinvent wheel. Furthermore, there are datasets that are filtered by the Lipinski rule.

  • $\begingroup$ Thank you for the advice. I should clarify, the chemical formula is not what I am starting with, I do have the 2D structures rendered in ChemDraw with the absolute stereochemistry. My problem is really, how to best sample torsional angles at all rotatable bonds to get a reasonable 3D structure? $\endgroup$ – Sevy Jan 28 at 17:26
  • 1
    $\begingroup$ Rendered as in exported as a raster image (JPG) or 2D draw (as opposed to 3D embedded)? If the latter, in ChemDraw, right clicking on the molecules allows one to export as a SMILES string, which is can go into one's notes or whatever. For downstream applications, exporting as a mol file is the best bet. To generate conformers, open-babel is an excellent option and has some ways to generate good diversity of conformers —including ones with unhappy torsions if one wanted. If the ligand is a natural metabolite (e.g. ATP) the biggest caveat docking is getting the protonation right. $\endgroup$ – Matteo Ferla Jan 28 at 17:35
  • 1
    $\begingroup$ I should say that compchemists generally used the Python module RDKit —but that requires knowledge of Python. $\endgroup$ – Matteo Ferla Jan 28 at 17:38
  • $\begingroup$ I have them as a JPG image, that's why I used PyMol to build them and get the stereochemistry correct (it's a bit more intuitive for me to see them in 3D rather than to write the SMILES string by hand). I then exported them as a mol file. I have tried both OpenBabel and RDKit, both seem good, I'm just not sure how extensive the conformational search is $\endgroup$ – Sevy Jan 28 at 18:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.