Archives

Bioinformatics has witnessed considerable progression in recent years; the prediction of antigen sequence in big data environment still remains challenging. A novel approach is proposed here to generate and evaluate tri-peptide markers, where a combination of high frequency tri-peptides can signify a characteristic of target antigen sequence. A dataset of Plasmodium falciparum antigen sequences is extracted from benchmark uniref100 protein sequence database; Training and test set are generated from extracted P. falciparum dataset. Genetic Algorithm (GA) is used here to identify an optimal set of tri-peptide markers from training set. Through different generations of GA, markers are evaluated using approximate selection function. A total 100 tri-peptides are identified using GA and the rest 150 are extracted by examining fitness function using iterative convergence algorithm. A back propagation neural network is trained to predict target antigen sequences using selected tri-peptide markers. The algorithm is tested on a test set which is non-inclusive in training set and the prediction result obtained shows 93% accuracy. This algorithm can also be useful to synthesis new sequence as possible drug antigen for given target protein.
Keywords: Plasmodium Falciparum; Tri-peptide Residue; Occurrence Frequency; Population Ratio; Genetic Algorithm; Iterative Convergence Algorithm; Back-propagation Neural Network.