Multi-algorithm and multi-model based drug target prediction and web server
Abstract
Ying-tao LIU1, #, Yi LI1, #, Zi-fu HUANG1, #, Zhi-jian XU1, Zhuo YANG1, Zhu-xi CHEN1, Kai-xian CHEN1, Ji-ye SHI2, *, Wei-liang ZHU1, *
1Drug Discovery and Design Center, Key Laboratory of Receptor Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, China; 2Informatics Department, UCB Pharma, 216 Bath Road, Slough SL1 4EN, UK
Aim: To develop a reliable computational approach for predicting potential drug targets based merely on protein sequence.
Methods: With drug target and non-target datasets prepared and 3 classification algorithms (Support Vector Machine, Neural Network and Decision Tree), a multi-algorithm and multi-model based strategy was employed for constructing models to predict potential drug targets.
Results: Twenty one prediction models for each of the 3 algorithms were successfully developed. Our evaluation results showed that ~30% of human proteins were potential drug targets, and ~40% of putative targets for the drugs undergoing phase II clinical trials were probably non-targets. A public web server named D3TPredictor (http://www.d3pharma.com/d3tpredictor) was constructed to provide easy access.
Conclusion: Reliable and robust drug target prediction based on protein sequences is achieved using the multi-algorithm and multi-model strategy.
Keywords: drug target; protein sequence; multi-algorithm and multi-model strategy; web server; support vector machine; neural network; decision tree
This work was supported by National Natural Science Foundation of China (81273435 and 21021063), National Science & Technology Projects (2012ZX09301001-004, 2012AA01A305, and 2013ZX09103001-001). Computational resources were provided by supercomputer TianHe-I in Tianjin and the Shanghai Supercomputing Center (SCC). The authors thank the developers of free and/or open source software for academic use, including SignalP-3.0, netOglyc-3.1d, netNglyc-1.0, tmhmm-2.0c and EMBOSS-6.0.1.
# The first three authors contributed equally to this work.
* To whom correspondence should be addressed.
E-mail Jiye.Shi@ucb.com (Ji-ye SHI); wlzhu@mail.shcnc.ac.cn (Wei-liang ZHU)
Received 2013-07-20 Accepted 2013-09-23
Keywords:
1Drug Discovery and Design Center, Key Laboratory of Receptor Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, China; 2Informatics Department, UCB Pharma, 216 Bath Road, Slough SL1 4EN, UK
Aim: To develop a reliable computational approach for predicting potential drug targets based merely on protein sequence.
Methods: With drug target and non-target datasets prepared and 3 classification algorithms (Support Vector Machine, Neural Network and Decision Tree), a multi-algorithm and multi-model based strategy was employed for constructing models to predict potential drug targets.
Results: Twenty one prediction models for each of the 3 algorithms were successfully developed. Our evaluation results showed that ~30% of human proteins were potential drug targets, and ~40% of putative targets for the drugs undergoing phase II clinical trials were probably non-targets. A public web server named D3TPredictor (http://www.d3pharma.com/d3tpredictor) was constructed to provide easy access.
Conclusion: Reliable and robust drug target prediction based on protein sequences is achieved using the multi-algorithm and multi-model strategy.
Keywords: drug target; protein sequence; multi-algorithm and multi-model strategy; web server; support vector machine; neural network; decision tree
This work was supported by National Natural Science Foundation of China (81273435 and 21021063), National Science & Technology Projects (2012ZX09301001-004, 2012AA01A305, and 2013ZX09103001-001). Computational resources were provided by supercomputer TianHe-I in Tianjin and the Shanghai Supercomputing Center (SCC). The authors thank the developers of free and/or open source software for academic use, including SignalP-3.0, netOglyc-3.1d, netNglyc-1.0, tmhmm-2.0c and EMBOSS-6.0.1.
# The first three authors contributed equally to this work.
* To whom correspondence should be addressed.
E-mail Jiye.Shi@ucb.com (Ji-ye SHI); wlzhu@mail.shcnc.ac.cn (Wei-liang ZHU)
Received 2013-07-20 Accepted 2013-09-23