Prediction of Secondary Structure of Proteins Using Sliding Window and Backpropagation Algorithm

Prediction of protein secondary structure plays a vital role in structural biology. Computational methodology is the initial step in bioinformatics to predict the 3-D secondary structure from a primary sequence and structure homology. This problem lies in the category of NP problem, and thus its time and space complexity is very high. In this paper, in a model for secondary structure prediction of proteins using sliding window and MADALINE, a multilayer feedforward network is proposed. The algorithm starts with encoding of amino acid sequence, which after passing through window is given as input to the neural network. The resultant data is in numeric format and translated back to actual secondary structure. It is observed from the results that the proposed technique provides better prediction with an accuracy more than 75%.