Hi,
I want to run FARFAR2 MPI protocol to build 3D model of 70 nt long RNA with command:
mpiexec -np 8 ${rna_denovo_mpi} -fasta seq.fasta -secstruct_file rna.secstruct -minimize_rna true -cycles 20000 -nstruct 5000 -fragment_homology_rmsd 1.2 -exclusion_match_type MATCH_YR -out:file:silent $outfile_name > out.log
I am looking for suggestion about two parts:
1. How to decide optimal value for nstruct? I understand it could depend on RNA length, is there any thumb rule to decide -nstruct option?
2. I also want to include "base pair steps" sampling option. Is it related to options "-fragment_homology_rmsd 1.2 -exclusion_match_type MATCH_YR", as they seem to originate from FARNA protocol. However, I could not find more information on these options.
Thanks,
Mandar
Category:
Post Situation:
> How to decide optimal value for nstruct? I understand it could depend on RNA length, is there any thumb rule to decide -nstruct option?
You'll never really know for sure, but one thing I used to like is to stop when I saw multiple occupancy clusters at a particular RMSD radius. If I am sampling the same clusters multiple times and have relatively few single-occupancy clusters, I've probably explored the conformational landscape pretty well at that level of granularity. I would suggest using more than 8 processors!
> I also want to include "base pair steps" sampling option. Is it related to options "-fragment_homology_rmsd 1.2 -exclusion_match_type MATCH_YR", as they seem to originate from FARNA protocol. However, I could not find more information on these options.
These are unrelated. Those two flags you mention are important only for the benchmarking we did (to ensure that we did not borrow information from PDB structures homologous to our benchmark cases) and won't have any effect without a native provided -- and if all you want is to produce the best possible model, can be omitted entirely.
Thanks for your suggestions! This clarifies a confusion for me. I am planning to run these calculations on 128 processors.
If I sampled 5000 structures and now need to extend the rna_denovo run to sample more, could you please suggest how to restart the run correctly with the extended—nstruct value? I could not see any checkpoint file in my working folder.
Thanks,
Mandar
In test runs, I realized that keeping the .out file in the working folder extends the run automatically.