Dear Rosetta community,
I am working with Rosetta version 2020.08.61146. have run into an issue of massive clashes when refining a part of a ribosome structure using FastRelax. I have previously used Fast Relax extensivly as a preliminary refinement step and its usually works wonders. This has been on individual chains and now I want to try to relax an entire subunit however Fast Relax seems to tear a few nucleotides apart. I attached images pre and post FastRelax as well as my RosettaScripts XML and shell output.
A few general observations.
1) It seems to occur in regions where hydrogenation would introduce a minor clash.
2) It tends to happend between rRNA and protein.
3) It tends to happen in regions of worse density.
4) It does NOT occur when refining isolated chains of the same initial conformation with the same scripts.
5) The minimizer line search repeatedly fails and aborts at low iterations. Seems relevant.
What I have tried:
1) Changing the "elec_dens_fast" density weight, lower is better but broken nucleotides still occurr, I have tried values between 20 and 500.
2) Increasing the FastRelax "repeats", higher is better, very little difference after 5 iterations.
3) Limiting the lineminimizer iterations to 200 by passing "-optimization::default_max_cycles 200"
3) Doing a centroid minimization via MinMover with "lbfgs_armijo_nonmonotone" before calling FastRelax, helps but dues not fully eliminate residues tearing apart.
4) Performing previous steps with dualspace FastRelax, not much difference.
Any insight or help would be appreciated.
Example of basic run script.
<ScoreFunction name="dens" weights="beta_cart">
<Reweight scoretype="elec_dens_fast" weight="%%densweight%%"/>
<LoadDensityMap name="loaddens" mapfile="%mapfile%"/>
<FastRelax name="relaxcart" scorefxn="dens" repeats="3" dualspace="1"/>