Predict potential drug targets from the ion channel proteins based on SVM
详细信息 查看全文
- 作者：Chen Huang ; Ruijie Zhang ; Zhiqiang Chen ; Yongshuai Jiang ; Zhenwei Shang ; Peng Sun ; Xuehong Zhang ; Xia Li
- 关键词：Two-stage prediction ; Secondary structure ; Subcellular localization ; Feature selection ; Cross-validation
- 刊名：Journal of Theoretical Biology
- 出版年：2010
- 出版时间：21 February 2010
- 年：2010
- 卷：262
- 期：4
- 页码：750-756
- 全文大小：177 K
文摘
The identification of molecular targets is a critical step in the drug discovery and development process. Ion channel proteins represent highly attractive drug targets implicated in a diverse range of disorders, in particular in the cardiovascular and central nervous systems. Due to the limits of experimental technique and low-throughput nature of patch-clamp electrophysiology, they remain a target class waiting to be exploited. In our study, we combined three types of protein features, primary sequence, secondary structure and subcellular localization to predict potential drug targets from ion channel proteins applying classical support vector machine (SVM) method. In addition, our prediction comprised two stages. In stage 1, we predicted ion channel target proteins based on whole-genome target protein characteristics. Firstly, we performed feature selection by Mann-Whitney U test, then made predictions to identify potential ion channel targets by SVM and designed a new evaluating indicator Q to prioritize results. In stage 2, we made a prediction based on known ion channel target protein characteristics. Genetic algorithm was used to select features and SVM was used to predict ion channel targets. Then, we integrated results of two stages, and found that five ion channel proteins appeared in both prediction results including CGMP-gated cation channel beta subunit and Gamma-aminobutyric acid receptor subunit alpha-5, etc., and four of which were relative to some nerve diseases. It suggests that these five proteins are potential targets for drug discovery and our prediction strategies are effective.