Abstract—Transcription factors are key elements in the regulation of genetic expressions. Understanding the behavior of the system is the ultimate goal behind modeling biology networks including gene regulatory networks. Prediction of regulation relationship between a transcription factor and a target gene can be viewed as a machine learning problem. Within this perspective, several algorithms have been developed to solve this problem using primordially support vector machine. In this work, we propose a semi-supervised approach to infer gene regulatory networks using both unsupervised and supervised techniques. First a set of reliable negative examples are extracted using clustering’s techniques. Then we use this set for classification using SVM. We have applied the proposed method to simulated data of Escherichia Coli, the experimental results show the competitiveness of the proposed approach as the prediction accuracy of the most tested case is achieved to the best desired value.

Index Terms—Gene regulatory network, machine learning, semi-supervised learning, supervised learning, unsupervised learning.

Meroua Daoudi is with Computer Science Dept, Constantine 2 University, Algeria. She is also with MISC Laboratory, Constantine 2 University, Algeria.
Souham Meshoul is with Computer Science Dept, Constantine 2 University, Algeria.
Fariza Tahi is with IBISC/CNRS Lab Evry, 91000 France.

Cite: Meroua Daoudi, Souham Meshoul, and Fariza Tahi, "A Machine Learning Approach for Gene Regulatory Network Inference," International Journal of Bioscience, Biochemistry and Bioinformatics vol. 9, no. 2, pp. 82-89, 2019.