interpret rosetta score

[rkgasn]

Hello,

I've been working with pyRosetta to create a de novo peptide-ligand covalent complex, which is covalently linked to a Lysine residue at position 20. I used the FastRelax protocol for energy minimization and ended up with a final score of -93.35, down from an initial score of 5462.60.

I'm trying to understand what this negative score implies. Does a negative Rosetta score indicate a more lower energy favorable and stable complex? Also, does the significant decrease in score mean that the structure has been minimized from its starting point? Any insights would be appreciated.

def setup_cartesian_relax(pose, lys_pos=20, cycles=100):
scorefxn = pyrosetta.create_score_function('ref2015_cart')
scorefxn.set_weight(pyrosetta.rosetta.core.scoring.ScoreType.fa_rep, 1.0)
scorefxn.set_weight(pyrosetta.rosetta.core.scoring.ScoreType.fa_atr, 1.5)
scorefxn.set_weight(pyrosetta.rosetta.core.scoring.ScoreType.atom_pair_constraint, 0.01)
scorefxn.set_weight(pyrosetta.rosetta.core.scoring.ScoreType.angle_constraint, 0.01)
scorefxn.set_weight(pyrosetta.rosetta.core.scoring.ScoreType.dihedral_constraint, 0.01)

relax = pyrosetta.rosetta.protocols.relax.FastRelax(scorefxn, cycles)
movemap = pyrosetta.MoveMap()

lig_pos = pose.size()
idxs = pyrosetta.rosetta.utility.vector1_unsigned_long(2)
idxs[1] = lys_pos
idxs[2] = lig_pos

resi_sele = pyrosetta.rosetta.core.select.residue_selector.ResidueIndexSelector(idxs)
movemap.set_bb(allow_bb=resi_sele.apply(pose))

movemap.set_chi(allow_chi=resi_sele.apply(pose))
#movemap.set_bb(False)
movemap.set_jump(True)
relax.set_movemap(movemap)

relax.cartesian(True)
relax.minimize_bond_angles(True)
relax.minimize_bond_lengths(True)

relax = pyrosetta.rosetta.protocols.relax.FastRelax(scorefxn, 100)  

[roccomoretti]

Rosetta scores (in analogy to free energy) are negative-better. So the more negative the Rosetta score, the better Rosetta thinks the structure is. However, the absolute value of the energy doesn't necessarily have a direct meaning. It's often the case that the Rosetta energy calculated from various protocols correlates with experimental results (as in you can plot both the experimental and computational values and get a straight line of best fit -- but not always with a unity slope or zero intercept).

So yes, by using FastRelax to reduce the score from 5462.60 to -93.35, you've made the structure much better (at least in the eyes of Rosetta).

P.S. Your final relax = pyrosetta.rosetta.protocols.relax.FastRelax(scorefxn, 100) basically undoes all the setup you did prior to it. Relax will only optimize the structure once you call the apply() function, so if you reassign the relax variable to a newly created FastRelax mover before you call apply(), it's only that later mover (with the later mover's settings) which will be used when you do call apply().

[rkgasn]

thank you, Rocco, for the suggestion, yes you are correct that the final relax statement undoes the previous cartesian relax settings. but the peptide has a non-cannonical amino acid and the ramachandran scoring for the ncAA seems to be not available in the scoring database. So, I cannot really run the cartesian relax on the nCAA as it gives me the following error:

File: /home/benchmark/rosetta/source/src/core/scoring/RamaPrePro.cc:432
[ ERROR ] UtilityExitException
ERROR: Error in core::scoring::RamaPrePro::get_mainchain_torsions_covered()