An approach to inferring gene regulatory networks via boolean modeling and feature selection

Gene regulatory networks (GRNs) are fundamental to regulating cellular functions and developmental processes. However, inferring GRNs from gene expression data continues to be challenging owing to intricate gene interactions. The traditional STP-based Boolean modeling approach suffers from dimensionality explosion as the network size increases, since the matrix dimensionality grows exponentially with the number of nodes. This limits its scalability and applicability to complex, large-scale biological systems. This paper proposes a novel two-step approach that combines feature selection with the STP method to efficiently infer Boolean network models for GRNs. The XGBoost model improves feature selection through regularization, while Shapley values enhance interpretability by quantifying the contributions of selected features. Our method outperforms STP method in terms of computational efficiency. Furthermore, the proposed method achieves results more accurate than two traditional methods and provides a more comprehensive and nuanced evaluation.