QsarDL: An end-to-end deep-learning-based model for QSAR prediction


  Research on quantitative structure-activity relationships (QSAR) provides an effective ap proach to accurately determine new hits and promising lead compounds during drug discovery process. In the past decades, various works have gained highlighted performance with the de velopment of machine learning algorithms. The rise of deep learning techniques, along with massive accessible chemical databases, has improved the QSAR predictive performance. This paper proposed a novel end-to-end deep-learning-based model to implement QSAR prediction by the concatenation of encoder-decoder model and convolutional neural network (CNN) ar chitecture. The encoder-decoder model is mainly used to generate fixed-size latent features to represent chemical molecules; while CNN framework is used to train a robust and stable model by these features to predict active chemicals. Two different schemes were investigated to evaluate the validity of our proposed model on the same data sets. Experimental results showed that our proposed method outperforms other state-of-the-art methods, which indicates that our model can successfully identify a chemical molecule whether it is active.
Availability: http://deeplearner.ahu.edu.cn/web/qsarDL.htm



Supplementary files:

drug-target datasets and supplementary files: supplementary (including enzymes, icon channels, GPCRs and nuclear receptors).


ShanShan Hu, Peng Chen, Jun Zhang, Bing Wang, Jinyan Li, An end-to-end deep-learning-based model for QSAR prediction, Submitted.

Copyright @ 2004-2018 by Peng Chen

All Rights Reserved