The Application of Deep Learning in the Prediction of HIV-1 Protease Cleavage Site

Abstract

HIV-1 protease cleavage site is critical for the design of HIV-1 protease inhibitors. Classification algorithms based on traditional machine learning are often used to deal with the prediction of HIV-1 protease cleavage sites. Unlike the classification algorithms of machine learning, the classification algorithms based on deep learning can extract the characteristics of the data well and get better performance. In this paper, HIV-1 protease cleavage site data is innovatively converted to One-hot data, and then two better classification models are proposed based on RNN and LSTM. At last, the experimental results are compared with the support vector machine algorithm and the random forest algorithm in traditional machine learning algorithm. The results show that the network structure based on deep learning designed in this paper can achieve higher accuracy than traditional algorithms after the HIV-1 protease cleavage site data is One-hot encoded, and the effects of RNN and LSTM are outstanding. Furthermore, the RNN-based classifier and LSTM-based classifier in this paper have much better Recall rate and F1-Measure than CNN and have high generalization ability.