You are here

loop modeling scoring

2 posts / 0 new
Last post
loop modeling scoring
#1

Hi, everyone. Recently I'm using Rosetta to perform the modeling of a loop region. But when I check the score.sc file, I got several values that are extremely high. I don't know how rosetta perform the scoring of these models. Is it due the clash of atoms ? Does it mean that this conformation is bad and should not be taken into consideration ? Besides, I want to convert the rosetta energy into physical energy unit (kcal/mol) ? How can I perform this ? Can anyone give me some tips ?

Thank you very much

Nicky

Category: 
Post Situation: 
Wed, 2015-05-13 20:00
Nicky_Learning_...

The scores you get out are dependant on which exact protocol you're using and which particular score you're looking at. Generally, people tend to look at total score, which is the overall score of the protein, including all regions of a protein.

Internal clashes are very often the source of high scores. The best way to check what's going on is to take a look at the individual score terms. (These are often also listed in the score.sc file.) If clashes are an issue, the fa_rep score will be very positive. It's always positive, and often quite so, but the trick is to compare it to the fa_atr score. Normally these balance out, with the fa_atr score being more negative than the fa_rep is positive. If not, you've got a (rather severe) clash. Sometimes other terms are the cause of the problem, so look to see if any other term is very positive. (e.g. positive enough to account for the multi-hundred REU discrepancy)

What's next is to look at which residues are involved in the clash/positive term. At the bottom of the output PDB is a residue table. Scan through the individual residues until you find one or more which are abnormally positive - these are your residues which are giving you issues. You can also use the per_residue_energies and residue_energy_breakdown (https://www.rosettacommons.org/docs/latest/residue-energy-breakdown.html) applications to examine things in more detail, if you wish (or to generate the table if you have silent file output or are missing it.)

The other thing you probably want to do is score the input PDB - this will tell you if the high energies you're seeing are due to the loop modeling you're doing, or are due to the input PDB having a bad score. (Structures from the PDB often have high Rosetta scores - see Nivon et al http://www.ncbi.nlm.nih.gov/pubmed/23565140 and https://www.rosettacommons.org/docs/latest/preparing-structures.html for more.) If the high scores are pre-existing, then there may be nothing wrong with the loop conformations themselves.

Regarding putting things into kcal/mol, Rosetta energy units are arbitrary, and aren't intended to directly correspond to physical energies. That said, when people have looked at the relationship, more often than not it's linear. However, the exact slope/intercept of the conversion depends on which protocol you use. My suggestion is to take a few examples which are close to the unknown system which you wish to model but which have known energies associated with them. Do the same protocol on those systems and make a REU versus kcal/mol graph. This will tell you how good the correlation is likely to be for your system, and what the conversion factor is.

Mon, 2015-05-25 11:30
rmoretti