--
AndrewLeaverFay - 25 Apr 2005
In the rotamer substitution step of simulated annealing, Rosetta calculates the energy difference a rotamer substitution would create. Rosetta looks up the pair interaction energies for the current rotamer assigned to a particular residue and then looks up the pair interaction energies for the alternative rotamer being considered on that residue. Proper bookkeeping would let us avoid the first set of lookups: after all, the energies of the current state assignment were computed previously when the current state was the alternate being considered. Excluding the bookkeeping costs, the savings would yield a factor of 2 speedup.
I have already implemented this bookkeeping procedure in my C++ energy_graph structure. However, the energy_graph seems slower than Rosetta's existing energy2b structure, and takes ~66% more memory. I want to prepare a version that integrates within pack.cc and release it to the rosetta community right away. I'm making this my immediate priority as I can do it rapidly.
to top