Statistics of Unrelated Sequence Properties to Improve Prediction of B-Cell Based Linear Epitopes

Epitopes play a vital role in the development of peptide based vaccine and diagnosis of diseases including HCV treatment based on immunotherapy and, antigenic drift is well established fact within the epitopic region. However silver-lining in this problem is that the epitopes are distributed among a handful of proteins only and slight changes or mutations within sequence of this group of proteins make alteration of their epitope structures which finally lead to ineffectiveness of any drug or vaccine meant to contain this organism through interacting with these proteins. Since experimental methods for identifying epitope is costly and time-consuming, computational and statistical approaches are quite often considered to accelerate prediction of epitopes. Apart from the fact that more than ninety percent of epitopes are conformational in nature, prediction of b-cell based linear epitope remains a challenge for computational biology. This piece of work dealt with the identification step of B-cell based linear epitopes recruiting Elman backpropagation of neural network inputted by features extracted through statistics of apparently unrelated physicochemical and some other judiciously selected properties of amino acids. As a result, a significantly high accuracy was achieved for B-cell based linear epitope identification.