I ran relax on a structure downloaded from pdb (after I cleaned it using PyRosetta's cleanATOM()), following the instructions here: https://www.rosettacommons.org/docs/latest/preparing-structures.html. The exact command I used is this:
relax.linuxgccrelease \
-in:file:s 2B5I.clean.pdb
-ex1 \
-ex2 \
-use_input_sc \
-flip_HNQ \
-no_optH false \
-relax:constrain_relax_to_start_coords \
-relax:coord_constrain_sidechains \
-relax:ramp_constraints false \
This results in a new .pdb file, which I presume contains the relaxed structure, and a score file, "score.sc". This score file contains an entry "total_score", and I am wondering what it means. Is it the score of the relaxed structure? Or is it the delta-score associated with relaxing (score_relaxed - score_initial)?
To compare, I did a score.linuxgccrelease on the relaxed structure, which results in a new score file, default.sc. However, the score entry in this file is completely different from the "total_score" outputted by relax.
By the way, I am a newbie to rosetta. I appreciate your help.
total_score is the total score of the output structure - it's a sum of most of the other columns (the individual by-term scores). Rosetta does not report a delta-score.
Your secondary scoring reports different values because the PDB format only goes to 3 decimal places, but Rosetta's scorefunction is unfortunately more sensitive than that. The scores ought to be close but not necessarily identical.
I must be doing something wrong. The scores I get are completely different (relax returns -63.202, while score returns -324.035). Can you take a look at the commands I am using? The relax command is above. The score command is:
score.linuxgccrelease \
-in:file:s 2B5I.clean.relaxed.pdb \
where 2B5I.clean.relaxed.pdb is the output of relax.
You did a constrained relax, so the total score that is output by relax is that of the normal scoreterms plus the constraint term penalties. When you use the score application, you only get the normal score terms. You should be able to check this yourself by looking at the individual scoreterm columns. The other columns that are present in both should be more-or-less the same for both scorings, but there will be a few extra columns in the relax output. Most significant of these is going to be the "coordinate_constraint" term, which I'm guessing is going to be somewhere around +260 in the relax output, but either nonexistant or zero in the score output.
Right! That explains it. Thanks!