Trust but verify.

The question of how much to trust computational methods is brought up here at Chemiotics II. My answer is that it depends on what one is looking for in the first place.

If one is looking for some sort of completely accurate and precise way to have all biological phenomenona fall out of "first principles," well, I wouldn't hold my breath. Of course, I don't think anyone is really waiting for that. At least I hope not. I believe my feelings on these sorts of issues are best described by personal experiences I've had with computational methods.

In grad school, I had an interest in this one mid-sized protein (somewhere between 40 to 60 kDa) that was known to bind this particular ligand. There was a crystal structure of the protein with and without ligand, although of course it was hardly the entire story (which is why it was the subject of my research attentions). In any case, collaborators did some MD simulations, and it was consistent with what we had found and was known. In their next bit of work, they mentioned that they found something new regarding the mechanism of ligand binding. This was going on the same time as I was doing some work, and as it turned out, my data did not rule it out. And so new research was inspired for those who took up the project after I left.

Currently, I am embroiled in a sordid and complex tale of transmembrane signaling involving the receptor and varying amounts of soluble cytoplasmic proteins that propagate that signal. There was a fairly recent paper detailing MD studies of the signaling process. Well, part of it, I suppose - huge chunks on either end of the transmembrane receptor were not included, and none of the cytoplasmic proteins that bind and are modified by the receptor were included in the study. Certainly a daring attempt, but it's hard to get too worked up over it when it doesn't resemble anything that I actually work with on a daily basis.

In short....I think properly used, it can be a useful way to bridge what is measured experimentally with the metaphors we use to describe processes. (For example - people love using descriptions involving simple machines, but what is actually measured are thermodynamic or spectroscopic quantities. Of course, "force spectroscopy" looks to change this, but when you yank apart a protein, you are no longer just gently playing around at kT or sub-kT conditions to see what kinds of deformations you get naturally or as a response to some stimulus. Anyway....) Certainly, for small enough systems, I am inclined to give them a proper reading, and in cases where the system might be larger but is somewhat well characterized, the same applies. In giant systems where they toss out a number of critical components or oversimplify to the point of absurdity, I am generally far more skeptical.

Merry Christmas to those who celebrate, Happy Hanukkah to those who celebrate, and a delightful winter holiday season to the rest.