2022-01-17 BespokeFit meeting notes

15 mins

Preliminary protein-ligand binding data with openff-1.0.0 (and maybe bespoke-fit)

@Mateusz Bieniek @Joshua Horton

• MB showing the results of recreating the openff-1.0.0 using the PMX scripts and setup from Marry only for the TYK2 system.

• Some results have failed and gave nans, it seems like david also had some issues there is a csv files called failed_simulations but it has not got any TYK2 cases.

• DM the workflow is not perfect as there is issues with mapping these will be edges that always fail, we also had issues that were random failures and will work if run again. Check with david H to see what his random faliure rate was and see if it is the same.

• MB plots comparing local runs with David H data, we seem to have around 0.3-0.4 spead in results between runs.

• DM – This RMS difference is roughly what I expected. I’d be shocked if it were as small as 0.1 or 0.2.

• DC if we have around a 0.3 kcal/mol error on each calculation then this should explain the difference between the DH result and ours.

• DC one worry with the bespoke parameters is that we might not seem much inprovment as this is quite a well-behavied system.

• DC – Do we have results with bespoke parameters yet?

  • MB – Trying to use OpenEye with the charge fit but that breaks PMX. But AmberTools charges work OK.

  • JH – One complication is that pmx uses the old OFF toolkit (like, 0.5) and there was an OE API change since then, so the versions are coupled.

  • DM – It’d be good to update to a more recent version of OFFTK - It’s hard to, for example, say that sage is being applied consistently if details of the charge method has changed.

  • JW – Please loop me in on this update if there’s any problems - Hopefully it will just be a namespace/import statement change.

• DC – MB, have you checked whether structures look reasonable/nothing has gone particularly sideways in the simulations?

  • MB – Not yet, but I will check this.

•  

• DM – DC, the ad board would be interested in hearing about alternate functional form experiments. Is there something that you’d be able to present?

  • DC – That’s fine in theory, do we have data, JH?

  • JH – We have been struggling with hydration free energy calcs to verify the fits. We’d like to figure out how to scale the functional form to get good overlap between the states - We’re especially having trouble covering the end states. We’re planning on assembling folks (DM, MShirts, maybe others) to help us with this.

  • DM – Could give an update on condensed phase fitting in a month, or a more extensive update in two months. This isn’t your job, but this could certainly be interesting.

  • DC – Yeah, I’d be up to present what we have, tentatively next month.

  •