CBS-Pred Help

CBS-Pred prediction module consists of webservers to predict carbohydrate-binding sites using single sequences (CBS-SS)
or server generated evolutionary information in the form of PSSM (CBS-PSSM). CBS-Pred is an implementation of our earlier algorithm [Malik & Ahmad, 2007].

CBS-SS works in the following steps:

1. Representation of each residue by 20 bit binary vector (all units are zero except one which represents its identity).
2. Appending four C-terminal and four N-terminal neighbor vectors of each residue to its own representation.
3. Compute amino-acid composition of protein and its length (21 units).
4. Append composition vector to each residue representation.
5. Use the previously trained neural network for making predictions.
Prediction accuracy:
Sensitivity=51.9% & Specificity = 69.90%

CBS-PSSM works in the following steps:

1. Creation of Position Specific Scoring Matrix (PSSM) using PSI_BLAST against an NCBI database of protein sequences, formatted by BLAST.
2. Representation of each residue by rows of PSSM (profile).
3. Appending four C-terminal and four N-terminal neighbor PSSM rows of each residue to its profile.
4. Compute amino-acid composition of protein and its length (21 units).
5. Append composition vector to each residue profile.
6. Use the previously trained neural network for making predictions.
Prediction accuracy:
Sensitivity=48.2% & Specificity = 88.0%


BACK