AI-enabled engineering of hesperidin/ursodeoxycholic acid nanomedicine for synergistic treatment of drug-induced liver injury.

Drug-induced liver injury (DILI) is a complex and intractable disease because existing anti-oxidate therapies in clinic fail to modulate multiple pathological pathways concurrently. Here, we present a direction-aware framework that integrates disease-network analysis, AI-guided molecular screening, and self-assembled nanomedicine design for precise protection of DILI. Time-resolved transcriptomic profiling of DILI identifies two complementary repair axes: the suppression of cytokine-cytokine receptor signaling for inflammation control together with the activation of glutathione biosynthesis for antioxidant defense. Guided by these DILI-driven mechanisms, we develop a dual-constraint deep-learning model that jointly evaluates the interaction between therapeutic molecules and disease targets, enabling the identification of candidate molecules whose biological effects match the desired intervention. Through independent screening from FDA-approved active pharmaceutical ingredients pool, we explore hesperidin (HES) and ursodeoxycholic acid (UDCA) as combination molecules capable of self-assembling into uniform nanomedicines (HUNMs) with predicted biological activities. Flash nanocomplexation-based engineering of HES and UDCA produces stable carrier-free nanocrystals with improved aqueous dispersibility. In an acetaminophen-challenged DILI mice, HUNMs alleviate hepatic injury, suppress inflammatory responses, restore glutathione homeostasis, and accelerate liver recovery. Together, our insights highlight an AI-native strategy that harnesses smart molecules to develop a precise and translatable nanomedicine for efficient management of DILI and other complex diseases.