a neural network based MHC Class I Binding Peptide Prediction Server

IMTECH Home

nHLAPred
A neural network based MHC Class-I Binding Peptide Prediction Server


Binding Motifs:The binding of a peptide to an allele is examined on the basis of occurrence of specific residues at specific position. These residues are known as anchor residues and positions are known as anchor positions. According to motif based prediction, the presence of motifs will determine whether a peptide will bind to specific allele or not. The methods based on motifs have 60-70% accuracy of prediction and being extensively used for identifying MHC binding peptides (Rammensee et al., 1995). The presence of motif does not ensure the binding of peptide to MHC because only 30% of MHC binding peptides have motifs (Buus, S. 1999). The usefulness of the motifs is further diminished due to the presence of the secondary anchor residues at the non-conserved positions (Ruppert et al., 1993).


Quantitative Matrices:These quantitative based methods consider the contribution of each residue at each position in peptide instead of anchor positions/residues.Quantitative matrices provide a linear model with easy to implement capabilities. Another advantage of using the matrix appoarch is that it covers a wider range of peptides with binding potential and it gives a quantitative score to each peptide. Their predictive accuracies are considerable.The quantitaive matrices used for prediction are available here


Artificial Neural Networks(ANNs): The ANNs are self training systems that are able to extract and retain the patterns present in submitted data and subsequently recognise them in previously unseen input.The ANNs are able to classify the data of MHC binders and non binders accuractely as compared to other. The ANNs is able to generalise the data very well. The major constriants of neural based prediction is that it require large dat for training.


Structure based method : Structure based methods are logically very sound but computationally complex. These methods calculate binding energy of peptide-MHC complex and the energetically favorable peptides are predicted as binders. These methods are in stages of development due to less amount of three demsional data about the MHC and peptide interactions.The structure base prediction require time and enormous computational power.so they are not yet fully developed