DGFE-VG: Dynamic Gene Feature Extraction via Visibility Graph

Abstract

Gene expression time series data plays an important role in the field of bioinformatics and data mining, as the analysis of the expression level and time series could detect specific information like gene function and classifications. In this paper, we propose a dynamic gene feature extraction method via visibility graph (VG). It is carried out in four stages: 1) complex networks are constructed from gene time series data; 2) different features are extracted based on the network structure and the specific characteristics of VG algorithm; 3) different classifiers are adopted to analyze gene time series data compared with different feature extraction methods, while, clustering algorithm are applied based on dynamic feature extraction via VG to achieve better performance; 4) different datasets are used to verify our method including clarifying and clustering according to the feature we extract. Abundant experiment results prove the effectiveness of VG method's in extracting the time varying and specific gene features underlying realistic complex gene expression data from time series.