Question about constraints during FastRelax for mutation studies

[brycejoh16]

Hi Rocco (or Rosetta Team),

We’re running the following pipeline:

1. Initial preparation: Clean the input PDB using clean_pdb.py, then run FastRelax with 10 repeats, selecting the lowest-energy structure from 10 nstructs.

2. Mutation step: Introduce a side-chain mutation and run FastRelax again (we’re currently doing ablation tests to determine the optimal number of repeats and other hyperparameters for this step).

Our goal is to reduce computation of (2) while still obtaining comparable total_score. One proposed idea is to do minimize and/or repack only within a defined Å-distance around the mutated residue in step (2).

One approach we’re considering is to run step (1) exhaustively (many repeats and nstructs) so that residues far from the mutation are already in favorable positions and don’t need to be redone each time in step (2).

Rosetta’s current recommendations for the preparation step are:

-relax:constrain_relax_to_start_coords
-relax:coord_constrain_sidechains
-relax:ramp_constraints false

Our concern is that these constraints may inhibit the very flexibility we’re trying to allow in step (1) to make step (2) less computationally expensive—since they limit how far atoms can move from their starting positions.

Would you agree that it’s reasonable to remove these constraints during the preparation step to potentially reduce computation in step (2)? Or do you think removing them could introduce undesirable behavior, and there’s an important reason they’re recommended?

As always, thanks very much for your help!

Best regards,
Bryce

Edit: This thread may answer my question that it is in fact necessary to avoid going to deep into an energy well... #368