Has reversed engineering been done on known protein to find where the lowest movement would produce the greater change in energy?
We could find the strongest ligand and allow them less movement and allow more movement to weaker ligands.

---

I've never had a biology class in a post-secondary level (beyond my recent Biophotonics) so I can't comment reliably on the science involved. However, my impression has been that the energy gradient (measure of how fast the energy would move as related to the position of the links) primarily depends on position, not on the actual material (amino acid).

If that's the case, then this method would be worthless since what we are doing now is trying to optimize the program for work on UNKNOWN proteins. Without prior knowledge of the best configuration, there would be no way to tell which ligands to weight in which direction.

Since the position of the amino acids is so important, even if there is a significant effect from the amino acids themselves, it would likey be shadowed by the position of the final protein. My (admittedly poor) intuition tells me that the final position of each link would have much more effect on the energy gradient.

Again, it would be a great idea if they were trying to optimize the program to be able to find the structure of known proteins for whatever reason. However, since it ultimately won't be used with a "native" or known protein, there would be no way to find the final energy gradient.

Another approach might be to find the energy gradient at each step and simply move each link towards the lowest energy. This is impractical though, since the resulting optimization function would have so many degrees of freedom, that there would be an astronomical number of local minimums -- so many that you wouldn't get anywhere CLOSE to the lowest possible energy state. In order to jerk the system out of each local minimum, you'd have to use the random jumping that's done now. The whole process would take much more computation time for little or no improvement in performance!

Anyway, some of that is just my intuition talking, but I DO have a lot of experience in optimizing systems (primarily in lens design) and enough math study to fill my head many times over. It's quite possible that I'm vitally misunderstanding the processes involved in the makeup of proteins, and I'd welcome correction!

---

Quick reply since it's christmas, and I just have a short break while my nieces are watching "Frosty the Snowman".

I'm not sure I understand the question, so I'll just throw out a few things, in no particular order, in hopes that they might be what you are looking for.

- We use an approximate, empirical energy function when evaluating protein conformation. When we refine known protein structures with this energy function, an ensemble of structures with slightly different conformations and energies is produced, but generally, these are all still in a fairly deep global minimum, which is deeper than false local minima that incorrect decoys find.

- The proteins being folded on Rosetta@home right now are ones without ligands or cofactors. These are easier to start with. By ligands I mean separate molecules that bind non-covalently to the protein. I have a feeling your definition of ligands is different than mine, but I'll let you clarify before attempting to answer that part of your question.

- Both the position and the actual amino acid is important. When predicting a structure, first we optimize the configuration of the protein backbone using a simplified energy function, with a simplified representation of amino acids. This is the ab initio, centroid level stage

- With the simplified amino acids, we ask: is it polar or not? are they fairly close together in 3D space or too spread out (extended)? We try to optimize these simple properties

- Then we use a very detailed, atomic resolution energy function in the second stage (full atom refinement). In this stage the actual physical shapes of the amino acids, including where each carbon, hydrogen, nitrogen and oxygen (and sometimes sulfur) is, is considered, and we optimize the structure for a very fine level of atomic detail.

- By link do you mean the links (bonds) in the amino acid chain (protein)? Yes, you are right, we cannot test each configuration of each link (residue), the search space is too big. What we are trying to do is develop a searching strategy to guarantee that we alwyas sample the native structure, because if we sample it our energy function is usually good enough to recognize it.

I've started a thread a while ago on protein biochemistry and plan to fill that in soon. Hopefully that will help too.

---