EnGRNT: Inference of gene regulatory networks using ensemble methods and topological feature extraction

Abstract

The development of computational methods to predict gene regulatory networks (GRNs) from gene expression data is a challenging task. Many machine learning methods have been developed, including supervised, unsupervised, and semi-supervised to infer gene regulatory networks. Most of these methods ignore the class imbalance problem which can lead to decreasing the accuracy of predicting regulatory interactions in the network. Therefore, developing an effective method considering imbalanced data is a critical task. In this paper, we propose EnGRNT (Ensemble methods for Gene Regulatory Networks using Topological features) approach to infer GRNs with high accuracy that uses ensemble-based methods. We compare the performance of state-of-the-art inference methods on simulated networks under different scaling conditions. Experimental results demonstrate that the appropriateness of the inference method relies on the size and type of expression profiles in microarray data. Except for multifactorial experimental conditions, the proposed approach outperforms unsupervised methods. Consequently, the proposed method can provide satisfactory performance to infer GRNs for networks with (<150) nodes in experimental conditions (knockout, knockdown, and multifactorial). For large networks, it is vital to consider biological conditions for selecting an appropriate algorithm. The aim of this study is to explore the effective method to infer GRN, which helps biologists and medical specialists in drug design.