zkatsim@nimr.mrc.ac.uk wrote: > I'm trying to do some modelling using 6 templates with very low > sequence id (<30), so I wan't to generate as many models as possible. > I was thinking 50 in the first instance (unless someone recommends a > lot more). I have decided to send the job to a cluster (parallel > machines) I have access to in the interest of time, but before doing > so I wanted to clarify some things. > > Am I correct in assuming I need to create 50 diff script files, each > with a different STARTING_MODEL and ENDING_MODEL, and in each script > setting a different random number seed? > > ie. script 1 STARTING_MODEL = 1 ENDING_MODEL = 1 > > script 2 STARTING_MODEL = 2 ENDING_MODEL = 2 .....and > so on
Yes, you could certainly do this.
> Also, if I do this, will all models be intercomparable, because I > thought you were only able to compare models of the same run. Could > someone briefly explain why this is or isn't the case?
You can only compare models built from the same set of restraints, since the energy function is a function of these restraints. But in your case every run will have the same restraints (since your templates and alignment will be the same each time) so the runs will be comparable.
> And also, one last question. Should I be worried about log files > over-writing each other, because I think they will all go to the same > directory in the cluster? Is there maybe some line I can add to each > script do direct the output into its own folder? (As you probably can > tell from that question, my background is not in computing!)
Not only the log files, but the other outputs (schedule, initial model, restraints). The easiest thing to do is to run each job in a separate directory.
Ben Webb, Modeller Caretaker