| SB – Just want to make sure that my previous work is doscoverable/you don’t have to reinvent it. I’d send a DM with some starting points for integrating sqm with geomeTRIC CD – Currently I’m looking at compiling a database of these failing cases. Also found three approaches to prevent proton migration (only do one optimization cycle, use SHAKE to restrain protons, something else) SB – Talk more about your current approaches? CD – SHAKE seems promising (prevents proton migration), but the molecule ends up adopting a really weird conformation. SB – My understanding is that OE keeps dihedrals fixed, but lets bonds and angles move. So there may be a non-geomeTRIC way to do that. But what I found is that the geomeTRIC-sqm combination is SUPER slow. Is there some way to apply implement MM potentials in sqm? CD – I saw something promising from the QM/MM chapter in the AmberTools manual. SB – I had seen/tried that. But the interface is a little messy – This requires special particles that overlap between the QM and MM potentials. It may be more complicated than it’s worth. OM – That sounds like a bad path. SB – Agree
SB – Where are you at now? CD – Currently I’m gathering examples of this problem. So right now I’d like 20-30 examples where we match w/ openeye. SB – Instead of matching to OpenEye, we could look at “how well do charges reproduce some more detailed QM-based electric fields?”. This way, you could have reference data that DOESN’T have proton migrations, as well as reference data that DOES, and you could see how much the restraint scheme affects previously-OK results. SB – Dimer energies as well. SB – Please reach out to me if you’d like to try matching to QM electrostatics or dimer energies.
OM – How concerned should we be with getting consistent output between OE and AmberTools? SB - For more molecules, we could send them through MSKCC lilac. SB – I’ll do this and send you the molecules that get flagged. Please send me the script you’re using. JW – In the long run, we should get CD set up on a Colorado workstation/supercomputer. This will be helpful when we start iterating on solutions.
JW – At what magnitude of charge difference between OE and AT should we contact you, SB? SB – There’s no absolute cutoff that I could say. OM – Also there’s the possibility that we make better charges (according to RESP) but make the output worse. SB – Could also ask MShirts, since he’ll have some experience with gauging these charge differences.
|