KGBN: Augmenting and optimizing logical gene regulatory networks using knowledge graphs

Abstract

Logical gene regulatory network (GRN) models provide interpretable, mechanistic representations of cellular regulation and are widely used in systems biology. However, most existing models remain incomplete, context-specific, and difficult to extend to comprehensive GRNs, limiting their broader applicability to tasks such as drug-response prediction and precision medicine. Here, add some ideas about virtual twins: With AML patients, we can use comprehensive GRNs to predict drug response. We present KGBN (Knowledge Graph-augmented Boolean Network modeling), a computational workflow for systematically augmenting logical GRN models. KGBN incorporates regulatory interactions derived from curated knowledge graphs as alternative logical rules while preserving the validated structure of existing models. Rule probabilities are optimized against experimental data to represent regulatory uncertainty and achieve data-driven calibration. Applying KGBN to acute myeloid leukemia, we show that extending an existing GRN with drug-target pathways and training against ex vivo drug-response data yields mutation-specific models that recapitulate known therapeutic sensitivities and signaling dependencies, demonstrating the utility of KGBN for interpretable, context-aware GRN modeling. (Should be predict)