Using spectral components for predicting treatment effects on time series microarray gene expression profiles
Analyzing time series gene expression profiles is an increasingly popular method for understanding the behavior of a wide range of biological systems. One could study the status of a disease by analyzing the induction or repression activity and effects from a number or a specific group of genes. In such a scenario, it is often natural for biological researchers to pose out the question of whether one could predict the treatment effects by using such time series microarray gene expression profiles. However, such problem is a big challenge considering their specific nature: usually such time series gene expression profiles are short and the sampling rates are not uniform. Our experiments with a real-world dataset show that traditional machine learning methods such as support vector machine will not perform well in such a case. In this paper, we decompose a time series gene expression profile into frequency components and apply machine learning algorithms to help improve the prediction accuracy. Experimental results show that our algorithm is both accurate and effective.
Qian Xu Hong Xue Qiang Yang
Bioengineering Program HKUST Clearwater Bay, Kowloon, Hong Kong Dept.of Biochemistry HKUSTClearwater Bay, Kowloon, Hong Kong Dept.of Computer Science and Engineering HKUST Clearwater Bay, Kowloon, Hong Kong
国际会议
成都
英文
1-5
2010-06-18（万方平台首次上网日期，不代表论文的发表时间）