Time-Varying Gene Regulatory Networks Inference Using KL Divergence from Single Cell Data

Correct reconstruction of dynamic gene regulatory networks from time-series single-cell RNA sequencing (scRNA-seq) data is essential for understanding biological processes, but remains challenging due to high-dimensionality, sparsity, and temporal heterogeneity. We propose a novel framework that integrates Kullback-Leibler divergence (KL) divergence-based temporal variation measurement with an autoregressive model and different regularization methods to infer time-varying regulatory networks from time-series scRNA-seq data. Partial correlation analysis further refines the sign of interactions (activation or inhibition). Simulation studies on a 10-gene synthetic dataset and a THP-1 monocyte differentiation dataset demonstrate that our approach accurately recovers dynamic network structures and maintains temporal consistency.