From: Axel Kohlmeyer (
Date: Fri May 23 2008 - 18:59:33 CDT

On Fri, 23 May 2008, Christopher Gillespie wrote:

CG> Hello all,

hello chris,

CG> I have been searching through the lists looking for an answer to this. I
CG> want to get the gofr for the ions in my system relative to specific parts of
CG> my protein, which I can do perfectly fine with my small systems. On my
CG> systems approaching 100k atoms the dcd files are on the order of 18-20GB.
CG> So I break these up into 4 smaller dcd files and try to load the dcd but it
CG> is still to big to load directly. I thought about using bigdcd for this but
CG> I am not sure how a gofr calc would get handled since the examples of COM or
CG> RMSD do not require knowledge of the counts from each previous frame.

the same is true for the g(r). it is only the "spatial" correlation
function not a time-correlation. so either using bigdcd and then
"measure gofr" on a single frame as well as operating on chunks of
arbitrary size and then averaging over the results should work fine.


p.s.: please make sure that your "delta r" parameter is small enough.
(e.g. 0.001). i've seen a few (unusual) example systems recently where
the results were not correct with larger delta-r. there has to be a bug
in the construction of the histograms (e.g. a mixup of < and <= or
something similar), but i have not yet been able to identify it.

CG> Is there a way to either use bigdcd with gofr or combine gofrs from smaller
CG> dcd files to make a gofr for the entire run?
CG> So far I break it down to 1ns parts for a 4ns production run, but that is
CG> still too large.
CG> Thanks
CG> Chris

Axel Kohlmeyer
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
If you make something idiot-proof, the universe creates a better idiot.