A New Supervised Learning for Gene Regulatory Network Inference with Novel Filtering Method

摘要

Gene regulatory network (GRN) inference from gene expression data plays an important role in understanding the intricacies of the complex biological regulations for researchers. In this paper, a new hybrid supervised learning method (HSL) is proposed to infer gene regulatory network. In HSL, according to the data imbalance ratio, three different supervised learning methods: direct classification, KNearest Neighbor (KNN) method and complex-valued version of flexible neural tree (CVFNT) model are chosen to classify. A novel filtering method based on integration of mutual information (MI) and maximum information coefficient (MIC) is proposed to eliminate the redundant regulations inferred by HSL. Benchmark data from DREAM 5 are used to test the performance of our approach. The results show that our approach performs better than the popular unsupervised Learning methods and supervised Learning methods.