On 4/12/16 7:48 PM, Joseph Eschweiler wrote: > I'm using the MC optimizer to optimize protein complexes represented by > spheres. My script loops through various sets of restraints (either > distance or MS-connectivity) and generates an ensemble of structures for > each restraint set. My problem is that the speed of the optimizer > changes drastically each time over the loop, and there is no correlation > between the complexity of the restraint set and the speed of the > optimizer. The speed usually varies between 10 and 30 models per second, > but will sometimes slow to 1 model/10minutes forces me to terminate the > script.
I can't think of anything in IMP that would cause such a drastic speed difference. (There are certain caches which, when invalidated, would have to be rebuilt, which will slow things down a bit - for example, the nonbonded list used by close pairs restraints - but that shouldn't be that noticeable.) Have you tried profiling or attaching a debugger during the slow steps? Possibly you are hitting a corner case where the MS-connectivity restraint gets confused; that restraint isn't used very often outside of native MS work.
Ben