I am using rosetta3.2 membrane_abinitio2 function to predict some membrane protein with 4 TM helices.
I split the prediction into 10 independent runs and each run resulted in 2000 decoys in a silent file.
When I combined these 10 silent files together, I find that the "score" changes.
Here is the score profile before combination:
SCORE: score vdw Mpair rama Menv Mcbeta Menv_non_helix Menv_termini Menv_tm_proj Mlipo hs_pair ss_pair rsigma sheet co clashes_total clashes_bb time description
SCORE: -199.290 12.289 -48.331 1.597 -50.581 -114.263 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 40.372 0.000 0.000 392.000 S_00000001
The following is score after combination:
SCORE: score vdw Mpair rama Menv Mcbeta Menv_non_helix Menv_termini Menv_tm_proj Mlipo co hs_pair ss_pair rsigma sheet clashes_bb clashes_total silent_score time description
SCORE: -158.917 12.289 -48.331 1.597 -50.581 -114.263 0.000 0.000 0.000 0.000 40.372 0.000 0.000 0.000 0.000 0.000 0.000 -199.290 392.000 S_00000001
Obviously, there is some term called "co" which is positive and change the score term.
I thought there difference is caused by different options during Abinitio run and combine_silent, but it seems not to be the reason.
The options I used in abinitio run and combine_silent are:
Does anyone know what's wrong?
Thanks a lot!
From the examples you've posted, I don't see anything wrong. The columns are reordered, but that's not too surprising as the order of the columns is somewhat arbitrary anyway. (Note the "co" term is present in the original line as well, right after the "sheet" term.) The issue you need to worry about is if the values under a given column name change.
The only difference I see is that the original "score" term is now labeled "silent_score", and a new, different value is used for "score". I can't say for certain, but I'd guess this is because the combine_silent application is rescoring the structures with the standard score function, and isn't appropriately taking into account the membrane adjustments. Just use the "silent_score" column and ignore the "score" column of the combined silent file, and you should be fine.
Thank you, rmoretti!
I did not read the score terms carefully.
I could use the silent_score for energy_rmsd plot,but when I doing clustering, the returned models of the cluster centroids have the changed scores but not the original "silent_score". Do you know how to deal with it?
Besides, I found that this problem also happens to the case that is given in the tutorial package (BRD4.pdb).
I haven't really used the abinitio & cluster applications (membrane or otherwise), but from a quick glance at the code, it looks like the cluster application is written to look specifically for the silent_score column. In the tracer output log (program standard output) there should be a summary section which lists the energies used, and those should be those from the silent_score column. I think it's the case that the clustering is using the correct score values, but for some unknown reason the values aren't getting copied to the output file.
One workaround is to use the scores from the original input file as a lookup table to match the structure tags with the appropriate score values.
I will try!
Thanks for you suggestion!