PROBselect: accurate prediction of protein-binding residues from proteins sequences via dynamic predictor selection

Help | Materials | References | Acknowledgments

PROBselect webserver

PROBselect suggests a predictor that is likely to provide the best prediction of protein-binding residues the for the input proteins. PROBselect uses predictions generated by SCRIBER and estimated AUC of SSWRF and CRFPPI to make the recommendation.

Please follow the three steps below to make predictions:

1. Upload a file with protein sequences, or paste them into text area

Server accepts up to 5 (FASTA formated) protein sequences with minimum length of 21 residues and maximum length of 5000 residues. Either upload a file or enter each protein in a new line in the following text field (see Help for details):

2. Provide your email address (required)

Please provide your email address to be notified when results are ready.

3. Predict:

Click button to launch prediction.

Help

PROBselect accepts either single or multiple protein sequences and the input is limited to 5 protein sequences at a time. The user should submit the protein sequence(s) in FASTA format.

The format of the input file is as follows (Here is an example of input file.)

  • Line1: >protein ID
  • Line2: protein sequence (1-letter amino acid encoding)

Each protein requires two lines and multiple proteins should be placed in consecutive lines.

Materials

    Datasets used to design and evalaute PROBselect:
  • Training dataset: Click here to download
  • Test dataset: Click here to download
  • The format of the above files is explained here

Acknowledgments

We acknowledge with thanks the following software used as a part of this server:

  • ANCHOR - Prediction of protein binding regions in disordered proteins
  • ASAquick - Fast neural network-based predictor of solvent accessibility
  • HHBlits - Algorithm for homology detection by iterative HMM-HMM comparison
  • PSIPRED - Prediction of secondary structure