Title: Deep recurrent neural networks in HIV-1 protease cleavage classification

Authors: Nima Shayanfar; Vali Derhami; Mehdi Rezaeian

Addresses: Computer Engineering Department, Faculty of Engineering, Yazd University, Yazd, Iran ' Computer Engineering Department, Faculty of Engineering, Yazd University, Yazd, Iran ' Computer Engineering Department, Faculty of Engineering, Yazd University, Yazd, Iran

Abstract: Many of the drugs that are used by HIV infected humans can be catalogued as protease inhibitors. These drugs mainly restrict the protease activity and therefore reduce the formation of mature proteins. By studying the protease and predicting its cleavage sites, one can hope to achieve better drugs. Predicting HIV-1 protease cleavage problem has been addressed by many machine learning approaches but its classification false positive and the lack of an approach with good generalisation remains a challenge. Furthermore, deep neural networks have shown many promising results in machine learning community. Therefore, we present a new deep learning approach to predict protease cleavage sites based on bidirectional Gated Recurrent Units and feed-forward networks. We have tested this approach on several datasets and achieved 89.04% accuracy and 92.45% AUC on average.

Keywords: deep neural network; gated recurrent unit; bidirectional recurrent networks; HIV protease; prediction.

DOI: 10.1504/IJDMB.2017.091364

International Journal of Data Mining and Bioinformatics, 2017 Vol.19 No.4, pp.298 - 311

Received: 19 Sep 2017
Accepted: 12 Feb 2018

Published online: 27 Apr 2018 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article