Derek Lowe states something which most of us know well, at least those of us who have spent a lot of time futzing around with Force Fields. He cites a recent paper in Nature MSB which explores the role of micro-evironments and their impact on hydrogen bonds. What he sort of alludes to is the fact that accounting for a lot of the effects in protein structures and their impact of folding and dynamics is something that most current methods do not really do a good job of accounting for. It’s actually a really hard problem, and while statistical techniques do a good job of globally accounting for these interactions, local interactions are very poorly represented. Given that polarization changes as you traverse a trajectory, micro-environments are not handled well by current techniques. I’ve always wondered what it would take to have a system where there is an engine that periodically updates force fields to match local environments. Computationally hard, but not impossible, or so I think. I am not sure we need to move to fully quantum models and QM/MM is at best a kludge. What we can do is come up with a better way of handling dynamic electrostatics in classical models. IMO that’s more likely to work and where we need to spend our efforts.
That’s the point that Derek misses. I think he’s right in that we need to collect more hard data, but having the hard data only helps if we have better formalisms and force field engines in place.
Hydrogen bonds and the weakness of force fields
Derek Lowe states something which most of us know well, at least those of us who have spent a lot of time futzing around with Force Fields. He cites a recent paper in Nature MSB which explores the role of micro-evironments and their impact on hydrogen bonds. What he sort of alludes to is the fact that accounting for a lot of the effects in protein structures and their impact of folding and dynamics is something that most current methods do not really do a good job of accounting for. It’s actually a really hard problem, and while statistical techniques do a good job of globally accounting for these interactions, local interactions are very poorly represented. Given that polarization changes as you traverse a trajectory, micro-environments are not handled well by current techniques. I’ve always wondered what it would take to have a system where there is an engine that periodically updates force fields to match local environments. Computationally hard, but not impossible, or so I think. I am not sure we need to move to fully quantum models and QM/MM is at best a kludge. What we can do is come up with a better way of handling dynamic electrostatics in classical models. IMO that’s more likely to work and where we need to spend our efforts.
That’s the point that Derek misses. I think he’s right in that we need to collect more hard data, but having the hard data only helps if we have better formalisms and force field engines in place.
Related articles by Zemanta