PB-kpred

We have shown that protein 3D structure information can be abstracted advantageously into a simplified 1D representation in terms of a protein blocks sequence (PB-assignment). Indeed, it was shown previously by us (Tyagi et al, 2006, Tyagi et al, 2008) that the knowledge of the PB-sequence for a protein may help to compare it with other structures, identify its fold and function.

For many proteins, the structures have not been determined yet and consequently, the fold and function cannot be assigned. Because the knowledge of the protein blocks sequence (PB sequence) can help towards this end, we developed PB-kPRED which is a server for predicting PB sequence from amino acid sequence. It is built upon the PB-PENTAdb database. The approach used for prediction is hence a knowledge-based one. Below is the overall scheme for the prediction. The methodology is detailed here.

PB-kPRED submission form

The user can input below one or several protein sequences in FASTA format (with no blank spaces or carriage returns in sequence name). The program will use a sliding window of 5 residues to parse each amino acid sequence. For each pentapeptide, it will interrogate PB-PENTAdb for its presence and get a list of all possible protein blocks. In the majority rule method, it reports the most frequently occurring PB. In the hybrid method, the most probable PB is reported by also taking into consideration the information of local structures of adjacent residues.