Protein-RNA interactions are key to post-transcritional regulatory mechanisms, as well as structure and assembly of viruses. Exact identification of mutual recognition sites on the protein and RNA becomes absolutely imperative both for a stable and specific complex formation. In the absence of a known structure of a protein-RNA complex mutual recognition sites cannot be determined and hence methods to predict them are required.
SRCPred is an attempt to predict such mutual recognition sites and hence help in the better understanding of these intricate and vital processes.
SRCPred predicts specific RNA-binding sites between the residue environments of proteins and dinucleotides in RNA. The input is an amino acid sequence in the FASTA format and the output is a set of scores for each amino acid residue, implying the probability of that residue binding to a particular dinucleotide subsequence within an RNA molecules.
The predictions are based on the parameters trained on a non-redundant dataset of 161 protein chains using a five-residue window encoded by PSSM rows of that protein. Apart from contact prediction scores for all 16 possible dinucleotides, the web server also returns sequence alignments used for generating the PSSM based prediction, and allows users to see, which other proteins are similar to the submitted sequence.
Prediction of dinucleotide-specific RNA-binding sites in proteins M Fernandez, Y Kumagai, D Standley, A Sarai, K Mizuguchi, S Ahmad BMC bioinformatics 12 (Suppl 13), S5 (2011)
A similar work for DNA-binding sites resulting in DNA-contact prediction web server sdcpred was published in the following.
Andrabi M, Mizuguchi K, Sarai A, Ahmad S: Prediction of mono- and di- nucleotide specific DNA-binding sites in proteins using neural networks . BMC Structural biology,2009,9:30. PDF