Scoring when the protein is a different length

[Anonymous]

I would like to truncate a protein (the C terminus). I’m not sure where – the segment is about 70 residues long and is 3 helices with short loop segments of a few residues connecting them. Rather than make 70 truncation mutants in the lab I thought I could narrow it down by comparing the scores of truncated PDBs.

If I relax and score a series of truncation PDBs then the longer proteins nearly always has a higher score. The score drops by 1 to 6 REUs each time an additional residue is included. This makes sense since there are more atoms. I thought about normalizing to the length of the protein… or number of atoms. That makes sense, no?

Also, I need a bit of a more aggressive sampling protocol to come up with a new structure once the protein has been truncated. I suppose I could do some comparative modeling of the whole protein but this seems unnecessary since the final truncated structure wouldn’t be too different from the original. Perhaps I should only do comparative modeling for the domain I truncate off and then dock it back into the main part of the protein.

Has anyone done something like this before. For instance, in loop modeling how does one know how long to make the loop? A longer loop could have a more negative score but not necessarily be more stable.

[Anonymous]

“If I relax and score a series of truncation PDBs then the longer proteins nearly always has a higher score. The score drops by 1 to 6 REUs each time an additional residue is included. This makes sense since there are more atoms. I thought about normalizing to the length of the protein… or number of atoms. That makes sense, no? “

Let’s clarify high versus low. Good structures have -2 or so REU per residue. So your scores should be getting more negative as you add residues, or more positive as you remove them. Normalizing to the number of residues makes sense; number of atoms is implicitly considered by the reference energies, but you could try that too.

“Perhaps I should only do comparative modeling for the domain I truncate off and then dock it back into the main part of the protein. “
This seems reasonable if you want aggressive sampling. I don’t think we have a protocol designed to work with the question of what the remaining half of a truncated helix will do to the folded core it once protected. I think relax alone might work here (do not use the constrain-to-starting-coordinates options). You could try something like FloppyTail in refine-only mode with just the truncated helix flexing.

For loop modeling, generally, “how long to make the loop” is purely empirical – try things and see what looks best to biophysical intuition.

[Author]

Posts