I am trying to run an all-atom refinement of candidate conformations with the Relax application. I found the manual page listing all of the relax parameters, however I am looking for guidance as to what these values should be. Can someone point me to a paper that describes the usage of the Relax application?
Specifically I am wondering about the following:
1. what is the relax:min_tolerance parameter used for?
2. how should I prioritize adding cycles to the three stages? That is, if i want to increase the time my refinement runs for should I primarily increase just the stage 3 cycles or should I increase stages 1 & 2 as well?
1) the relax::min_tolerance is used in gradient based minimization - its default value should be fine for almost all cases - i don't think a paper has been published on relax mode from Rosetta3.0
all stages do repacking and minimization and have the following moveset
stage 1 moveset - small/shear - lots of minimization
stage 2 moveset - small/shear - optionally wobblemover
stage 3 moveset - small/shear - lots of minimization
how you increase the number of cycles really depends on your desired effect, if you just want a lower score then increasing only stage3 should be fine but if you want more conformational sampling, I would do stage2 and stage3. In general though, I would recommend increasing nstruct and making more models - instead of doing longer runs, are you familiar with fast relax or with sequence relax, you might find that those provide the functionality you are looking for. If you explain your system a little bit, I will be able to provide a better answer.
thankyou for your reply. I am using a different application (not Rosetta) to generate candidate conformations for refinement. I am using Rosetta just for the all-atom refinement stage (with Relax) since my other application does not do all-atom minimization. At the moment I only have a couple dozen candidate conformations to refine, so i want to squeeze as much as I can out of the minimization step -- potentially running it for a couple hours on each one.
I saw the option for fast relax and sequence relax, but my number of candidate conformations is relatively low right now, so i want to focus on how to squeeze the best minimization out of those conformations.