[AMBER-Developers] ifcap and GB - potential problem?

From: Ben Roberts <roberts.qtp.ufl.edu>
Date: Wed, 24 Mar 2010 18:41:36 -0400

Hi all,

I've discovered what appears to be a fairly significant trap for young players in the combination of a solvent cap and generalised Born solvent. Insofar as it bit me, I think a patch in CVS (or, if necessary, after the release) would be helpful.

The situation is this. A day or so ago, subscribers to the main Amber list may recall that I sent an email highlighting extreme energies and atom collapses when I try to do GB simulations with a solvent cap. The full text of my email can be found here: http://archive.ambermd.org/201003/0522.html.

Not having received a response, I dug into the code this afternoon, and found that, in src/sander/rdparm.f, unless hybridgb is greater than zero, the sections RADIUS_SET, RADII and SCREEN will not be read with ifcap = 1. This seems to cause GB calculations (and perhaps PB, though I haven't tested those since my nickel ions are missing PB radii at present) to go horribly wrong. In any case, if I change the conditional to ask the sections to be read in if ifcap=1, the GB energy drops greatly. I'm running a minimisation at the moment, to see whether the problem with atoms collapsing into each other persists.

However, even if I get output that's not obviously wrong, I'm not sure whether my "solution" is sensible?

If it is, I'm happy to submit a patch. If not, though, I think it would be good for sander to exit with an informative error message, instead of giving garbage results. If the informative error is more appropriate, can someone advise on the most appropriate way and place to invoke one?

AMBER-Developers mailing list
Received on Wed Mar 24 2010 - 16:00:03 PDT
Custom Search