When I attempt to optimize a portion of a molecule for which I don't
have a good homolog, the loop optimization procedure always yields a
rather nondescript coil with no sheet or helix. Today I tried, as an
exercise, to see what "optimization" did to a known crystal structure. I
ran the program on the C-terminal helix of calmodulin, a molecule for
which the structure is very well known. As has been my experience
before, the result was destruction of that nice orderly helix, which was
replaced by a coil going off in no particular direction. Can anyone tell
me why this happens, and how I can use this procedure to get a better
result?
This happens because the loop optimization scoring function is not
always good enough to reproduce native structures, particularly for
large loops. If that were an easy problem to solve, then there'd be no
need for comparative modeling - we could just build everything ab initio.
The good news is that we have improved loop modeling procedures in the
works, but this is not yet a solved problem.