Classifying Proteins As Intrinsically Disordered

Question

[Daniel Standage]

I've worked a little bit with classifying protein folds and a very tiny bit with homology modeling. However, I've heard a lot more recently about intrinsically disordered proteins--proteins that don't really have a regular 3-dimensional structure. Such disordered proteins have been implicated in a variety of functions, so it appears that a regular structure is not necessary for a protein to be biochemically functional.

If I remember homology modeling correctly, most approaches will try their hardest to fit a novel protein to the structure of the closest related homolog, even if it's not a great fit. Is this the case? Say, for example, I have the peptide sequence of an intrinsically disordered protein and I try to predict its structure with homology modeling--I'm going to get some kind of structure in the end, even if it's completely bogus, right?

So my real question is: are there any methods or tools for predicting whether a protein is intrinsically disordered? Such a method/tool could be helpful when predicting protein structure so that you don't force a protein to have a structure it doesn't really have.

Answer 1

[Khader Shameer]

If I remember homology modeling correctly, most approaches will try their hardest to fit a novel protein to the structure of the closest related homolog, even if it's not a great fit. Is this the case?

Yes, depend on your query sequence and the availability of similar structure / template in your target database (PDB) you will get a model. Rosetta suite have an approach that is aware of disorder regions in proteins.

Say, for example, I have the peptide sequence of an intrinsically disordered protein and I try to predict its structure with homology modeling--I'm going to get some kind of structure in the end, even if it's completely bogus, right?

Yes, AFAIK current homology modeling or fold-recognition algorithms do not expect a protein sequence to be ordered or disordered. You can first run a Disopred/FoldIndex before pursuing a modeling and see if regions predicted to be disorder is retained as long, unstructured loops in the modelled structure.

Are there any methods or tools for predicting whether a protein is intrinsically disordered?

Yes, you can use Disopred or FoldIndex.

Intrinsically disordered proteins may or may not remain completely unstructured. IDPs have shown to have adopt structured regions upon binding to an interacting partner or another structured protein (See: manuscript ; F1000 report).

Also a protein may not be completely disordered, looking at your prediction results or checking for your protein of interest at DisProt / DisEMBL could help you to understand these aspects.

I have some interesting manuscripts/reviews on IDPs at my citeulike.

[Leonor Palmeira]

Very nice answer. I would add IUPred to the prediction tools for unstructured regions: http://iupred.enzim.hu/

[Khader Shameer]

Nice tool Leonor, thanks for sharing.