基于特征挖掘的离子通道功能类型预测与跨膜域识别

项目名称： 基于特征挖掘的离子通道功能类型预测与跨膜域识别

项目编号： No.61202256

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 林昊

作者单位： 电子科技大学

项目金额： 25万元

中文摘要： 离子通道是一种跨膜成孔蛋白，能够选择性控制离子进出细胞，对维持正常生命过程至关重要，其结构与功能的异常会引起多种疾病。尽管人们已对离子通道的结构和功能有一定了解，但由于解析膜蛋白结构非常困难，因此，缺乏对离子通道分子机制的完整详细的描述。借助计算的方法系统研究离子通道势在必行。本项目是申请者在蛋白质信息学研究方法的基础上，建立一套基于特征提取的离子通道预测的理论模型，构建高精度的离子通道功能类型识别和跨膜域预测工具。首先，利用信息熵解决序列中氨基酸间长短程的关联问题，并用于特征提取；其次，利用统计检验和流形学习方法，进行特征优化、筛选；进而，提出信息距离测度，结合机器学习方法对离子通道功能类型进行预测；最后，利用改进的结构重现图方法,结合判别分析方法对离子通道跨膜域进行预测。本课题的研究对揭示离子通道的生理功能、合理选择药物靶标都有重要的科学意义。

中文关键词： 离子通道；数据库；特征筛选；物化性质；机器学习

英文摘要： Ion channels are a diverse group of proteins that poke through the lipid membrane of cells and form the channel pores. They can selectively control movement of ions into or out of cells, which is the key extremely important for maintaining the activity of life. The abnormity of structure and function of ion channels can result in multi-disease. Some aspect of structure and function of ion channel have been studied and understood, however, because it is difficult to determine the structure of membrane proteins, there is still lack of exhaustive knowledge for their mechanisms. It is necessarily to study ion channels systematically by using computational approaches. In this project, based on the protein informatics appraoches, we will build a set of theoretical models for studying ion channels by use of feature mining techniques. Firstly, we will develop shannon entropy-based method to study the long and short correlation of amino acids, which can be regarded as standard of feature extraction. Secondly, we will develop various statistical methods and manifold learning methods to optimize feature sets. Furthermore, we will construct novel information-based distance and combine it with machine learning approach for predicting the function types of ion channels. Last, we will develop and mend structural recurrence gra

英文关键词： ion channel；database；feature selection；physicochemical property；machine learning