4
$\begingroup$

For two organic molecules of essentially unlimited size (up to and including, and sometimes exceeding MW = 1000 g/mol), I would like to generate all possible dimer pairs in vacuum. The pairs may be heterogeneous, so there could be one molecule of MW=1000 g/mol paired with something that is only MW=50 g/mol.

The objective is to find the lowest energy (preferably lowest free energy) dimer pair, in vacuum. The energy minimization I would count as an additional step. Given a set of dimer pairs I can take care of screening energies on my own, but if it will do it, so much the better.

This sort of thing is routinely done for crystal structure prediction, where one generates unit cells, and then crystals. Is it perhaps best to retrofit the more complex operation of crystal structure prediction in something like PyXtal.py, COMPACK, GLEE etc, as examples, to request only dimers rather than all possible crystal structures?

Bonus if the software is open source and MIT licence.

$\endgroup$
2
  • $\begingroup$ Maybe it is helpful if you specify your skill level. Generating the starting geometries with short python scripts would be one possibility. $\endgroup$
    – Greg
    Nov 23 at 18:47
  • $\begingroup$ Skill level is high. Currently generate rigid molecule coordinates randomly (random Center Of Mass and orientation), ensuring radii don't overlap on atoms. This however is inefficient, and, time consuming even when using a molecular mechanics FF. Only do this for a single conformation though, which is another quandary. Searching conformation space is I suppose the real issue $\endgroup$
    – Wesley
    Nov 23 at 18:51
7
$\begingroup$

Generating all possible structures systematically might be difficult. When it comes to the generation and refinement of chemically useful structure guesses, I'd recommend looking into CREST by Stefan Grimme. It will sample your entire structure space, i.e, not only intermolecular docking motifs, but also intramolecular conformations. It uses an MM approach with some additional biasing and crossover techniques to sample the total conformational search space. Beware though that with large enough problems, it will still be difficult to judge how complete your coverage of that search space is.

$\endgroup$

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.