1029-D-SPIN: Generative Regulatory Network Modeling

The paper introduces D-SPIN, a sophisticated computational framework designed to build mechanistically interpretable and generative models of gene regulatory networks using single-cell RNA sequencing data. By analyzing how cells respond to thousands of distinct perturbations, such as drug treatments or gene knockdowns, the framework can simulate and predict diverse cellular states and population structures. Unlike many black-box AI methods, this tool explicitly models the internal logic of biological pathways, allowing researchers to identify key regulators of cell fate and understand how drug combinations interact. Performance tests demonstrate that D-SPIN reconstructs these networks with significantly higher accuracy than existing methods, especially when identifying hidden interactions that are typically masked in normal conditions. Ultimately, the system serves as a powerful cell-state simulator that bridges the gap between large-scale genomic data and functional biological understanding.

References:

• Jiang J, Chen S, Tsou T, et al. D-SPIN constructs regulatory network models from scRNA-seq that reveal organizing principles of perturbation response[J]. Cell, 2026.