Dear friends of RG.
I have many protein sequences (of a single unstructured domain) and I want to perform an analysis for the prediction of disorder at their specific residues in order to identify regions that become structured upon interaction with partners.
So, I would appreciate knowing the best strategies and algorithms currently used to predict the disorder at the residue level.
The method I am considering using is to combine different results from the best algorithms currently available.
To do so, for each algorithm I was thinking about setting a binary approach to classifying its predicted intrinsic disorder of every amino acid.
I would set 1 for values that were above intrinsic disorder cut-off points of each algorithm and 0 for values below this cut-off. Then I would sum all contributions and make a graph for each residue position to identify possible intrinsic residues or regions. (As shown in the example figure, combining 7 different predictors)
So, I would like to know some state-of-art IDP predictors that could be used and if there is any other approach that might be better than mine.
Check https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5897131/
Also check https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3552272/
Dear Higor, the best way of having good confidence in your IDP/IDR predictions is by using multiple predictors and finally combine the results, as you suggested. Every single predictor has different strategies and is trained by using different databases, so it is very hard to really evaluate which are the tops. Most of the predictors do not work well for short IDRs (around 10 residues), be careful. For general analyses, I prefer to use Pondr, mainly because you can choose multiple predictors and also run other analyses like CDF and Charge-Hydropathy (http://www.pondr.com/). Your idea of combining different IDP predictors can now be quite automatically performed by using any of the several different meta predictors available online. Each meta predictor combines a different amount of single predictors, in different ways, so please refer to their original papers or on the help page. I would suggest:
1) META-Disorder -> http://iimcb.genesilico.pl/metadisorder/
2) MFDp (Multilayered Fusion-based Disorder predictor) -> http://biomine.cs.vcu.edu/servers/MFDp/
But, even better, you could use one of the many IDP databases available. I usually prefer to use the Database of Disordered Protein Prediction (D2 P2 ) - (http://d2p2.pro/). You can search for your protein by using the Uniprot ID. If you are not working with a very unusual protein, it might be there. The interesting part is that D2P2 includes several other predictions linked with IDPs/IDRs, like PTM and prediction of MoRF regions.
A nice reference for you to read before you go through your IDP/IDR prediction analyses would be:Article How disordered is my protein and what is its disorder for? A...
- Badges
- Science topic
- Similar topics
- Chemistry
- Biochemistry