On Dec 23rd, I successfully install Rosetta 3.1 and CS-Rosetta on my iMac and run the test using GB3 protein which was provided within the CS-Rosetta.
Here are the rough steps I used to get a simulated GB3 protein.
- run “runCSRjob3.com” as instructed in the online manual
- The RunRosetta3.com has to be modified first to perform ab initio simulation by Rosetta3.
(basically, just replace all double colon “::” in the RuRosetta3.com by single colon “:”).
- After ~70 hours simulation using a single CPU (3.06 GHz) on my iMac, I got the simulation done.
- As instructed by CS-Rosetta, one has to generate a “silent output” (the extract_lowscore_decoys.py).
- Before running “runCSrescore3.com”, you have to edit 2 files:
1. go to the CS-Rosetta folder, in the “com” subfolder,edit “extract_pdb3.com”, make sure no double colons are in the script same as step 3.
2. edit runCSrescore3.com,:
at line 28, it was “mkdir output -p” and the new one should be “mkdir -p output”,
at line 101, it was “rm ./pred/……… -f” and the new one should be “rm -f ./pred…….”.
Then you can run “runCSrescore3.com” and no error/warming messages will be shown.
I don’t give the program the defined seg.txt in this test.
The error message at step 5.2 look like the following lines:
==================
userssxx:rosetta>runCSrescore3.com new_silent.out
./output directory generated
mkdir: output: File exists
mkdir: -p: File exists
extracting PDB coodinates…
generate decoys raw score table file : ./output/name.rawscore.txt
decoy with the lowest energy score: S_j001_00000289
calculate rmsd to the lowest raw score decoy ( S_j001_00000289 ) and generate file : ./output/rms2LowRawScore.txt
generate rms_toLowestRawscoreModel verus score table file : ./output/name.rms.rawscore.txt
labadmin:rosetta>runCSrescore3.com new_silent.out gb3.tab
./output directory generated
mkdir: output: File exists
mkdir: -p: File exists
extracting PDB coodinates…
generate decoys raw score table file : ./output/name.rawscore.txt
rescoring decoys using sparta calculated chemical shifts…
(input chemical shift shift gb3.tab )
calculating chemical shift score for decoy S_j001_00000001
rm: -f: No such file or directory
calculating chemical shift score for decoy S_j001_00000002
rm: -f: No such file or directory
calculating chemical shift score for decoy S_j001_00000003
rm: -f: No such file or directory
===============================================================
Now, I am waiting for the back CS-prediction by Sparta and getting the scores. But I check several simulated PDB coordinates of GB3 protein, the structures are quite similar with the published one. Here is a random one.
After the complete of back CS prediction by sparta, the script RunCSrescore3.com is supposed to collect all information and generate few text file which contains the CS_chi2, rms and other information. I don’t know why the script was looking for S_j001_0001000.pdb (because I don’t have it, only 0-999.pdb), so the “pdbrms doesn’t work properly. Therefore name.rms.rescore.txt” does not have the complete information (should have 3 columns).
I did the following steps to get the complete information. All the following commands are extracted from the runCSrescore3.com
- set best_model = `cat name.rescore.txt | sort -g -k4 | head -n1 | awk ‘{print $1}’`
- pdbrms $best_model.pdb S*.pdb | awk ‘{print $2,$1}’ | sort -g > rms2LowReScore.txt
if you have seg.txt, then the commands are:
set seg = `cat ../seg.txt`
pdbrms $best_model.pdb S*.pdb -seg $seg | awk ‘{print $2,$1}’ | sort -g > rms2LowReScore.txt
- paste name.rescore.txt rms2LowReScore.txt | awk ‘{ print $1,$6,$4 }’ > name.rms.rescore.txt
The 1000 simulated GB3 structures seem to have trend of convergence (see the figure below). If more structures as suggested by the CS-rosetta manual (e.g. 10,000 structures), the convergence should be clear.
The 20 lowest Calpha_RMSD structures (left) and 20 lowest energy structures are superimposed by the NMR plugin in pymol which RMSD of backbone atoms (residues 1-56) are 0.6 and 0.68 angstrong, respectively.