[phenixbb] Out of Memory Errors

[Randy Read]

Hi,

The current defaults in Phaser are not kind to memory consumption for large cases.  As it happens, we're working on changing those defaults sensibly so that the memory usage will depend much less on the size of the problem, and hopefully also so it recovers more gracefully when an attempt to allocate memory fails.

In the meantime, you'll have to switch to Phaser-MR to avoid all memory issues, because AutoMR lacks some of the fine control.  The main thing you can do is to reduce the resolution in proportion to the mean radius of the assembly you're looking for.  (In practice this means that the resolution should be proportional to the cube root of the molecular weight -- if you would use 2.5A data for a 30kDa model, then use 5A data for a 240kDa model.  On that basis, even 4A is generous for your 12x48kDa search model.)  This will get you through everything but the final refinement step, where by default Phaser carries out rigid-body refinement against all data.  In the Phaser-MR GUI, there's a box labelled "High resolution for refinement", and you'll have to set that at something less than the full resolution.  Apart from resolution, you can look at our FAQ on memory consumption, which you can find at http://www.phaser.cimr.cam.ac.uk/index.php/FAQ, and that discusses some of the other expert parameters you could set to reduce memory needs.

However, in your case it's not clear to me why you would want to run molecular replacement with a model containing four trimers.  Is the structure really a dodecamer?  If not, then I wouldn't want to assume that the relationship among trimers is conserved from one crystal to another.

Best wishes,

Randy Read

On 20 Apr 2011, at 21:08, Heather Condurso wrote:

> Hello all,
> 
> I am currently running the Phenix GUI version 1.7-650 on an iMac with an 3.2Ghz Intel Core i3 processor, 8GB of memory operating under OS 10.6.  During many of my runs (mostly AutoMR runs but sometimes refine) I get out of memory errors.  I have my system booted in 64 bit mode and I watch the activity monitor and find that phenix doesn't use much more than 2GB of memory.  My search model is very large.  My unit cell contains 4 trimers (each monomer is 48kDa).  I hadn't encountered this issue until I began to increase my search model from one trimer to four.  Do you have any suggestions as to how I could avoid this error without minimizing my search model? Often my AutoMR runs get the out of memory error at the start of refinement, but just now the memory was exhausted during the fast rotation function.  During this run I used all default parameters with a max res of 4.0 and RMSD of 1.0.
> 
> Any advice you can give would be greatly appreciated!
> Heather
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb

------
Randy J. Read
Department of Haematology, University of Cambridge
Cambridge Institute for Medical Research      Tel: + 44 1223 336500
Wellcome Trust/MRC Building                   Fax: + 44 1223 336827
Hills Road                                    E-mail: rjr27 at cam.ac.uk
Cambridge CB2 0XY, U.K.                       www-structmed.cimr.cam.ac.uk