MG – What fraction of the benchmark set does this affect?
CC – For small peptides (2-5 residues), there’s a set of 2-mers that are all capped so they don’t have this problem, but some 3-mers that aren’t capped, as well as 4- and 5-mers. So 1/3-1/2 of the dataset.
CC – For something like an entire protein, it’s not a big deal if we mistreat the termini. But for small peptides it’s more difficult.
MG – Has anyone run these benchmarks with AMBER?
CC – Yes, other studies run with charged C termini at pH 7.
CC – One exception is that CSimmerling published on ala tetramer using ff19, where there ARE neutral C terminus charges.
MG –
CC – Could just try running with different protonation states and see how that affects results.
MG – Worried that this will multiplicatively increase the number of calcs we need to do.
CC – I think the right way to do it is to match the pH of the experiment regardless.
JW – Could make OH its own residue with net 0 charge? It’s a bad answer but it may be the least bad?
CC – Want to avoid easily-attackable ideas, avoid gapsys paper-style criticism.
DM – But if there’s no “correct” solution, we can just write that in the papers and do something else.
MG –
CC – So, decision on whether to use parameters that CSimmerling sends me?
MG + CC – …
CC – some manual work to copy over the parameters that CSimmerling used for ff19sb paper.
MG – CS may not have charges for all the termini that we need. So then we’d need to make stuff up, right?
CC – Yeah, but we could take the values from ALA, and use the same approach for GLY and VAL.
MG – Maybe Nerenberg params then?
CC – There’s not too different except for charges on amide nitrogen
MG – Do those look similar to mainchain/nonterminal amide nitrogens?
CC – Difference between restraints from Paul and Carlos is that Carlos made it explicity like a charged terminal residue, whereas Paul did a mainchain residue.
MG – So the Carlos solution is like a cap, whereas Paul let changes propagate deeper into the resiude?
CC – Yes
MG – Paul published this?
CC – Yes.
JW – PN charges sound better - It keeps us from needing to run “a method like what CS did”, because we can plug-and-play the exact values from PN’s work.
DM – Agree
MG – Agree
CC – So we can move ahead with Paul’s, but if we hear from CS that he has charges for all 3, does that take precedence over Paul's?
MG – Given that, even if we get them from CSimmerling, they’re still not published, it may be good to let Carlos know that this is the plan.
CC – Agree, that’s a good plan.
JW – I think the incentives work best if we use Paul’s
MG – Also PN’s parameters are published. So I think the best plan is to inform CSimmerling and see if he strongly objects.
PB (chat) – is this Paul's work that's being referred here, https://pubs.acs.org/doi/full/10.1021/ct2000183 -- Optimizing Protein−Solvent Force Fields to Reproduce Intrinsic Conformational Preferences of Model Peptides --