Unveiling gene perturbation effects through gene regulatory networks inference from single-cell transcriptomic data

Physiological and pathological processes are governed by networks of genes called gene regulatory networks (GRNs). By reconstructing GRNs, we can accurately model how cells behave in their natural state and predict how genetic changes will affect them. Transcriptomic data of single cells are now available for a wide range of cellular processes in multiple species. Thus, a method building predictive GRNs from single-cell RNA sequencing (scRNA-seq) data, without any additional prior knowledge, could have a great impact on our understanding of biological processes and the genes playing a key role in them. To this aim, we developed IGNITE (Inference of Gene Networks using Inverse kinetic Theory and Experiments), an unsupervised machine learning framework designed to infer directed, weighted, and signed GRNs directly from unperturbed single-cell RNA sequencing data. IGNITE uses the GRNs to generate gene expression data upon single and multiple genetic perturbations. IGNITE is based on the inverse problem for a kinetic Ising model, a model from statistical physics that has been successfully applied to biological networks. We tested IGNITE on two complementary systems of pluripotent stem cells (PSCs): murine PSCs transitioning from the naïve to formative states, and human PSCs differentiating toward definitive endoderm. These datasets differ in species, developmental trajectory, and single-cell technology (10X vs. Fluidigm C1), providing a stringent test of generalizability. Using only unperturbed scRNA-seq data, IGNITE simulated single and multiple gene knockouts (KOs) and produced predictions consistent with independent experimental observations. In mouse PSCs, IGNITE generated wild-type data highly correlated with experiments and accurately predicted the effects of Rbpj, Etv5, and triple KOs, while in human PSCs it correctly predicted differentiation-promoting and differentiation-blocking perturbations, in agreement with published studies. These results demonstrate that IGNITE robustly captures gene interaction logic across species and technologies, enabling robust in silico perturbation analyses directly from scRNA-seq data.Author summary: Physiological and pathological processes rely on complex interactions between genes, organized into gene regulatory networks that control how cells behave and respond to change. Understanding these networks is essential to predict how genetic perturbations influence cell states, but experimental testing of gene function is often time-consuming, costly, or technically limited. Computational approaches can help address this challenge, although many existing methods require additional experimental data or prior biological knowledge. Here, we introduce IGNITE, an unsupervised machine learning framework that infers gene regulatory networks directly from unperturbed single-cell gene expression data. Without relying on prior information, the method reconstructs effective gene interactions and uses them to simulate how cells respond to genetic perturbations, such as gene knockouts. We applied this approach to two distinct biological systems: mouse naive pluripotent stem cells transitioning to the formative state, and human pluripotent stem cells differentiating toward endoderm. Despite differences in species, developmental trajectories, and experimental technologies, the model accurately reproduced observed cell states and predicted the effects of genetic perturbations consistent with independent experiments. Overall, this work introduces a general and scalable machine learning framework for building predictive models of gene regulation directly from single-cell data.