Hello Mandar-- > > > >> > >> I have used anneal.inp with NOEs, Dihe, and hydrogen bond > >> restraints to get a good ensemble (~ 0.8 A RMSD for ten lowest > >> energy str.). The run is stored at: > >> > >> https://www.dropbox.com/s/0pse0rk3kl6bi6j/run7.tar.gz?dl=0 > >> > >> Further, I ran the refine.py with above restraints and RDCs using > >> starting structure from the lowest energy structure of above > >> (Cartesian dynamics) calculations. The refined structures are now > >> having an ensemble backbone RMSD of ~ 1.2 A. I played with the > >> some the obvious things but ended with no luck. Is there a way to > >> get lower avg RMSD of the ensemble near to the values given by > >> anneal.inp? > >> > > > > I'm not so concerned about having a less precise bundle of > > structures. Rather, I'm concerned that the quality of the fit of all > > commons terms appears to be worse. Perhaps the addition of the RDCs is > > messing things up. You might try a run7 run without RDCs and see how > > the RMSDs/num violations compares with your initial calculations. > > > > I am seeing issues only when I run python scripts (anneal.py or > refine.py). Same restraints give me a nice bundle with anneal.inp > but if I move to .py then I end up messing RMSD by 0.3-0.4 A. I > already checked .py scripts with and without RDCs and they do not > make much difference. BTW, I also played with weighing RDCs which > did not make any significant difference.
Looking at the run30 .stats file from the run you provide, the all metrics appear to fit better than your run7 calculations. The only problem I see is that the average structure doesn't converge so well due to the large spread in input structures. So the new result seems to be much better: you have a larger spread of structures that appear to fit the data better. The spread might be narrowed by including additional data or empirical terms - HBDB would be useful - see eginput/gb1_rdc/anneal.py. Also, you may wish to refine using the implicit EEFx solvent model - see eginput/eefx/refine.py. I have not done exhaustive analysis of the run7 structures- you might analyze those structures, along with the run30 structures to definitively see how the fits compare yourself. best regards-- Charles _______________________________________________ Xplor-nih mailing list [email protected] https://dcb.cit.nih.gov/mailman/listinfo/xplor-nih
