Assigning secondary structure is pretty ingrained, but if you _really_ want to do it, you can open the attached "noksdssp.py" file before your models and it will prevent ksdssp from being called. Just remember that it's a HACK :-) and may have side effects if you do anything else that might depend on secondary structure type. The size problem, on the other hand, needs a little more analysis. Would it be possible for you to send me a large dataset so that I can try out a few things? Thanks. Conrad On 6/28/12 6:23 AM, Hurt, Darrell (NIH/NIAID) [E] wrote:
Hi Conrad,
Thanks again for your hard work. This works very well for me. I really like the clustering algorithm in Chimera because it adjusts the clustering threshold automatically and does its thing very quickly.
Speed is very important for my purposes, so it was necessary to comment out the MatchMaker part of the script (thanks for including that; I'm glad it seems like that wasn't the majority of your work!). Another thing slowing down the clustering was the invocation of "ksdssp" upon opening of the individual PDBs. Is there an option to skip that?
Finally, I want to run this on at least 1000 PDBs. After about 50, Python had gobbled up more than 1 GB of memory and my laptop slowed to a crawl. I was able to kill it, but this may be a concern for production work. We intend to run this on a high-memory node of a cluster, but is there some way to reduce the memory used?
Thanks for everything!
Darrell