Reverse Engineering Transcriptional Regulatory Network for Rice from Gene Expression Data

Abstract

Transcriptional Regulatory Networks (TRN) expresses the underlying gene-protein regulatory mechanisms in living biological organisms. Computational inference models are introduced to construct these interaction networks in organisms to overcome the problem of time and cost taken for conducting huge amount of wet-lab experiments to gure out TRNs. Still computational inference approaches are built only for very few and simple model organisms. These models need decent amount of biological experiment data to conduct the inference and have di culty to infer TRNs when it comes to complex organisms which lack experimental data. We present a mathematical model for inferring transcriptional regulatory network (TRN) for complex biological organisms, when prior knowledge of underlying interactions is unavailable. The model incorporates linear regularization into network component analysis (NCA) model to bring out e cient solution. Convex Optimization is used to model the regularization equations for the decomposition framework (NCA) in Matlab environment. A major contribution of our research is constructing TRN for rice (Oryza Sativa) with our model by only using high-dimensional data generated by microarray experiment. We also experiment the use of novel partition algorithm RAC in clustering high-dimensional rice microarray data to reduce computational complexity of convex optimization model. The hypothetical hemoglobin data experiment with the inference model brought out most of the interaction intensity values correctly even without using proper prior knowledge. The outcome TRN network predicts linear interactions between previously identi ed rice transcription factors (TF) and genes that were reported as to have high regulation in experimented owering condition of the input data.