Loader prototype | Histidines getting +1/-1 charges instead of 0/0 might affect parameter assignment Looking into the 2.2.1 FF, #7+1 and #7-1 only appear once, in parameter b40. So having “unusual” N charges that still add up to the correct total charge might be OK. Unsure whether this is critical. The version in components.cif doesn’t have this weird charge assignment, so if we just use that (see point below) then we should avoid this.
JM – Generating protonation variants currently requires a lot of manual patching. JW – Could use general protonation/deprotonation reactions on (amines/carboxylates/imidazoles) to enumerate? JM – Would get hairy and invite false positives/negatives. Might be easier to just do a list of residue and atom names to permute protonation. … (much discussion) … JW – Yes, go ahead and just load the single form of each AA from components.cif, then have a list of atoms to remove and the charge to leave on the neighbor
Use explicit atom lists for this,
Handle disulfides We should get bond stereo from 3D (RDKit docs say that AssignStereochemistryFrom3D only does atom stereo, so we should find a workaround if this doesn’t get bond stereo) JW was supposed to come up with test cases but didn’t, so please start coming up with appropriate test cases for tricky things.
|