You are here

Working through tutorials: expected output scorces differ from calculated results

3 posts / 0 new
Last post
Working through tutorials: expected output scorces differ from calculated results
#1

Hi,

I'm new to Rosetta and currently working through the tutorials.

I often recognized that  the the score values produced by the recent version "Rosetta 2019.47" on my computer differ from the ones cited in the tutorials and given expected_output.

e.g.:

in the Scoring Tutorial  (subchapter "Changing the Score Function") the command

ROSETTA3/bin/score_jd2.linuxgccrelease @flag_docking

with flags

-in:file:l input_files/pdblist

-score:weights docking

-out:file:scorefile output_files/score_docking.sc

delivers the following score.sc-files:

SEQUENCE: 
SCORE: total_score       score dslf_ca_dih dslf_cs_ang dslf_ss_dih dslf_ss_dst      fa_atr      fa_dun     fa_elec     fa_pair      fa_rep      fa_sol hbond_bb_sc hbond_lr_bb    hbond_sc hbond_sr_bb linear_chainbreak overlap_chainbreak               time description 
SCORE:    -111.103    -111.103       0.000       0.000       0.000       0.000    -169.606       5.373      -3.798      -2.237       5.683      73.720      -0.822      -8.609      -2.680      -8.128             0.000              0.000              0.000 1qys_0001
SCORE:     -83.850     -83.850       0.000       0.000       0.000       0.000    -131.980       4.479      -3.295      -2.059       3.845      58.363      -1.663      -5.786      -1.092      -4.661             0.000              0.000              0.000 1ubq_0001

while expected output file says:

SEQUENCE: 
SCORE: total_score       score dslf_ca_dih dslf_cs_ang dslf_ss_dih dslf_ss_dst      fa_atr      fa_dun     fa_elec     fa_pair      fa_rep      fa_sol hbond_bb_sc hbond_lr_bb    hbond_sc hbond_sr_bb linear_chainbreak overlap_chainbreak               time description 
SCORE:     -89.596     -89.596       0.000       0.000       0.000       0.000    -143.190       5.640      -1.371      -2.577       3.929      62.290      -0.824      -5.653      -2.502      -5.338             0.000              0.000              0.000 1qys_0001
SCORE:     -55.214     -55.214       0.000       0.000       0.000       0.000    -119.403       4.578      -1.412      -2.044      10.233      66.103      -3.048      -3.799      -3.360      -3.061             0.000              0.000              0.000 1ubq_0001

As the tutorials say that the locally calculated scores should be the same as in the tutorial, I am now uncertain if these deviations occure due to a wrong software usage of mine or weather Rosetta's scoring functions could have changed since the tutorial was written.

It would be great if someone could help me with this question.

Kind regards,

Jan

 

(btw, if it's interesting for anyone: Rosetta seems not to run automatically on the latest Ubuntu 19.10 distribution. The terminal says something about that the GCC is no longer supported. First I (also new to Linux) was insecure because somewhere  I read that on Ubuntu it's not necessary to install additional compilers. Anyway I could build the software using a MX Linux 19.0 distribution)

Category: 
Post Situation: 
Sat, 2020-01-18 13:16
j.schmue

You're correct that the scorefunction behavior has changed since the tutorial has been written. The current default scorefunction environment for recent versions of Rosetta is REF2015. Now while there's a weights file with this name, the changes which made REF2015 are more than just the weights file. There's additional underlying behavior which was changed when moving to REF2015, and this is also reflected in the scorefunctions which aren't REF2015.

I think you're running into that here. The tutorial was written before we switched to REF2015, and as such the example outputs are using a slightly different scorefunction. (docking weights in a talaris environment versus docking weights in a REF2015 environment) I wouldn't be too concerned about a difference in scoring here.

 

Mon, 2020-01-20 08:42
rmoretti

Thanks for your answer.  Since the most tutorials were written around 2016 and as REF2015 is also mentioned in the scoring tutorial, I assuemd that REF2015 scorefunction was used in all the tutorials. But as far as I understand there were also changes in the software offside the REF2015-function that contribute can contribute to slightly different resultas.

Anyway I'm now more concernd about differences in pdb files originated from tutorial's tasks.

For example in the Fold-tree-tutorial the given command

$> <path_to_Rosetta_directory>/main/source/bin/rosetta_scripts.default.linuxgccrelease -in:file:s inputs/capsid.pdb -parser:protocol inputs/caps_relax1.xml -out:prefix test1_

delivers a pdb-structure that has big differences to the example outputs shown in the tutorial. On top of that, the  tutorial's next command

$> ../../../main/source/bin/rosetta_scripts.default.linuxgccrelease -in:file:s inputs/capsid.pdb -parser:protocol inputs/caps_relax2.xml -out:prefix test2_

which runs the same relax protocol, but from the opposite direction in terms of C- and N-termina, produces nearly the same structure as the first command.

To show this I will share the two pdb-files originated from these two commands:

 

Tue, 2020-01-28 13:31
j.schmue