DC: Talking through slides of why we want to use the double exp functional form.
The double exp form looks more complicated but is quite simple it has a repulsive term and an attraction term, but has more flexibility and the authors say it has natraul softcore built in but I need to confirm this.
This has been used as a water model before, the authors fit the alpha and beta terms using the tip3p parameters epsilon and sigma and aim to reproduce tip3p.
We have implemented this using smirnoff plugins and we can recreate tip3p experimental values as expected, but we see a different density to that reported but the authors used a different long range correction implemented in amber.
The standard long range correction may not work, we should look into this and maybe try and get it into openmm.
DM: shirts pointed out if you do not have any long range correction the density should depend on the cutoff, which is fine if you always use the same but people tend to tune this for more performance.
Binding free energy results could also be cutoff dependent so we may need to look into this in future.
For now fit the forcefield with a set cutoff and come back to fitting again with the long-range correction.
SB: We could start with a small proof of concept forcefield fit like we have done in the past for sage using esters and alchols to show that it works and see if there is any improvement and then take it from there.
This process should not take very long at all.
DM: We also want to start traning against a water model
DC: We can start by fitting a 3 and 4 point model and then doing the proof of concept and optimising against these new water models.
SM: We could then do some hydration free energies after this to validate the performance.
DC/JH: We have confirmed that the alpha and beta global parameters can be optimised by forcebalance so we should be good to start this.
SB:I put this in a while a good and should work well.
