DimitryASuplatov wrote:
In 9v2 manual (as in all previous) it is insistently recomended to use .assess_dope() for single chains only. What exactly does it mean? 1/ Could I possibly apply dope to multichain proteins? 2/ How should I use dope on multichains? Should I split each chain to single pdb and apply dope for each one sepparately? Would this be correct? Should`t dope include chain-chain interactions?
See http://salilab.org/archives/modeller_usage/2007/msg00333.html and http://salilab.org/archives/modeller_usage/2007/msg00329.html for the most recent discussions on this topic. But to summarize again: DOPE is a statistical potential, so it is dangerous to apply it to systems not covered in its training set (which included only single chains). But you can certainly apply it to multichain systems (and the procedure is no different than for single chains) and it will include chain-chain interactions. (We just make no "guarantees" about the statistical reliability of the scores, although in practice they seem to perform rather well.) You should certainly not split your PDB into chains and score each one separately, as that would explicitly exclude the chain-chain interactions, which is clearly wrong.
Ben Webb, Modeller Caretaker
--
http://www.salilab.org/modeller/
Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage