Deep learning and GPU based approaches to protein secondary structure prediction

Abstract

Characterization of proteins remains a problem of significant importance in analysis of disease progression, drug identification, phylogenetic analysis etc. The popular methods include use of Hidden Markov Models, Support Vector Machines, Neural Networks, hybrid methods and other machine learning methods. However, Deep learning has come out as the most trending approach for a variety of problems including protein characterization. We explore some of the most successful deep learning architectures and the variations there in. Deep learning has also shown promising results in other domains such as speech analysis, natural language processing, bio medical image and signal analysis and human computer interactive systems. GPUs have been extensively used for compute intensive tasks including bioinformatics and deep learning. An effort has been made to highlight the contributions involving deep learning, protein secondary structure prediction and GPU based computing.