Participants
PB: Updates on the sterics and fits comparison.
DLM:
Here, the difference in QM and MM is coming from the steric clash between oxygen and a proton. There are many more down the list that have similar steric clashes.
DLM: First of all we want to have a filter in our analysis to weed out the in-ring torsions. Reg. sterics, the difference in peaks of MM and QM are the result of strong sterics and we want to avoid them in future fitting cases. Change the threshold value from current value of 8 kcal/mol, go down to 6 or 4 and check whether the molecules have strong steric effect or not.
We see improvement in objective value from non-interpolated (fit3) to interpolated parameters (fit0) but this solely cannot be a metric to say we have a better FF.
PB: For the MM comparisons, something similar to release-1 BM (suggested by Jeff and Simon in core-devs meeting) where we can make a table of absolute QM vs MM relative energies differences averaged over conformers and check the torsion profiles.
JM: Which datasets did you use to check for steric effects?
PB: The 14 datasets currently being used for fits, listed here
JM: Did you have trouble making targets for different datasets?
PB: Yeah, I have the same trouble as you with missing datapoints Ben mentioned. I am currently using TG’s targets from the fits. You can use those or place a try, except statement where this error is occurring so that you can skip the points that give errors and continue looping over other points.
JM: Why are we fitting to in-ring torsions?
DLM: Ideally we should not, but there are cases where we want to do that if there is some buckling we want to account for, which are usually handled by impropers. …….
JM: I am currently making progress on the k-val Vs wbo plots, will look into adding more data. So this analysis goes towards wbo interpolation?
DLM: Basic question here is whether the trends (or slopes) we get with the by-molecule kval-wbo plots are the same as QM barriers Vs wbo plots of PB or at least some similarity with those. Our analysis is now based on these QM plot trends since this is the only data we have now, and these may not hold true and we have to look back and check whether our current splitting of parameters and chemical series makes sense. So this is a validation on whether we are going in the right direction. Since there are four torsions involved we may have a scaling factor of 4, or JM can plot both the data on the same plot with a primary and secondary y-axis having different scales, so that we don’t have to worry about the scaling.
0 Comments