As far as I know Aggrescan 3d is the only tool developed specifically for the purpose (based on structure). Note that hydrophobicity alone does not necessarily mean the residue is an aggregation hot spot. What is important is its location in the structure i.e. exposure to the solvent. If the hydrophobic residue is in the core of the protein or interacts with other residues stabilizing the structure it is not considered an aggregation hot spot. Only highly exposed residues that are hydrophobic will be regarded as such. 

Aggrescan is based on the the method outlined in the linked publication:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1262731/

At our university, there is a group working on protein and peptide aggregation, and they have a bioinformatics tool to predict aggregation prone sequences in a protein. Tool is called TANGO. See: switchlab.org

see also: https://www.ncbi.nlm.nih.gov/pubmed/15361882

and: https://www.ncbi.nlm.nih.gov/pubmed/27284085

HI Pietro,

Depends on what you want to do: do you want to know the aggregation potential of your protein in the unfolded state (then use short stretch predictors such as TANGO) or do you want to predict the aggregation in native state (Aggrescan3D is definitely good, but you could also look at our Solubis)? Are you looking for ways to reduce the aggregation of your protein? 

There are indeed quite some methods and in many cases they agree, but not always. Most have been benchmarked against experimental data and certainly cover some aspects.

There are good papers describing each algorithm, so best to start there.

Cheers

Joost

Thank you both for replaying. Yes, I am looking for ways to reduce the aggregation of my protein in the native form. I'm trying focusing on the best prediction and then realize it. Given that, I think of integrating information deriving from both tools (aggrescan and solubils) and find if there are some contact points. 

regards 

Hi Pietro,

Nice question anf great answers! 

One thing I find useful is to use multiple predictive methods,  map the consensus results on to the protein structural models (along with structure based predictions such as hydrophobic patches, SAP score and so on) and then choose sites for single point mutations such that they'll improve solubility but not interfere with other properties of the molecule such as thermal stability and activity. I also try to choose mutations that would increase the difference between protein pI and the buffer pH. 

I hope this helps. 

Best regards, 

Sandeep Kumar

But I have another question. How can the interaction of two proteins be best predicted? (Both proteins have the property of aggression)