A Single-Cell Network Inference Method Using Mutual Information for scRNA-seq Data Analysis

Lan-Yun Chang (張藍云)1,†, Ting-Yi Hao1,†, Wei-Jie Wang1, and Chun-Yu Lin1,2,3,4,*

1Institute of Bioinformatics and Systems Biology, National Yang Ming Chiao Tung University, Hsinchu 300, Taiwa

2Department of Biological Science and Technology, National Yang Ming Chiao Tung University, Hsinchu 300, Taiwan

3Institute of Data Science and Engineering, National Yang Ming Chiao Tung University, Hsinchu 300, Taiwan

Exploring cell-to-cell heterogeneity is essential for understanding the biological function and cellular responses, particularly in the context of disease. Nowadays, it has become possible to derive inherent biological system information from expression patterns at a single-cell level due to advances in single-cell RNA sequencing (scRNA-seq) technology. The identification of biomarkers and therapeutic targets for diseases has been significantly aided by network modeling, which can estimate the associations between genes to present the complicated and dynamic changes in biological systems. However, it remains challenging to precisely build a single-cell network (SCN) to capture each cell's network architecture and further examine heterogeneity between cells. Here, we propose a Single-cell Network Using Mutual Information (SINUM) approach to estimate SCNs from the scRNA-seq data, which means one network for one cell. SINUM calculates mutual information (MI) between any two genes based on the scRNA-seq data to examine their dependence or independence in the target cell. Using the various scRNA-seq datasets, we validated the accuracy and robustness of SINUM in cell type identification, outperforming the current SCN inference method. For example, the SINUM SCNs display a higher overlap with the human protein-protein interaction (PPI) network and are more likely to fit the scale-free characteristics. SINUM also provides a network-level view of biological systems to identify cell-type marker genes/gene pairs and reveal time-dependent changes in network rewiring during embryo development. We believe SINUM offers a new framework to facilitate network construction at a single-cell resolution and complements the traditional differential expression analysis on scRNA-seq data.