Uncoupling Type I Interferon Benefits From Inflammatory Toxicity: Transformer-Prioritized Precision Agonists for Potent and Safer Cancer Immunotherapy.

Paclitaxel (PTX) chemotherapy is constrained by an "immunomodulatory paradox," where antitumor Type I Interferon (IFN-I) activation is coupled with detrimental pro-inflammatory cascades. To address this challenge, we developed Deep Learning for Innate Immunity Modulatory Potential (DLINP), a Transformer-based framework designed to identify precision immunomodulators that uncouple IFN-I induction from deleterious inflammatory signaling. Screening 123 million entities identified Co68-an organometallic PNP-pincer complex-as a dual-functional agent with superior potency to conventional taxanes. In pancreatic ductal adenocarcinoma (PDAC) models, Co68 elicited robust antitumor responses that exceeded those of the gold-standard STING agonist DMXAA. Single-cell and spatial transcriptomics revealed that Co68 selectively re-engineered the myeloid compartment, reprogramming tumor-associated macrophages toward an interferon-stimulated gene (ISG)-high phenotype while quenching the pro-inflammatory IL1β-PGE2 feedback loop. This reconfiguration converted "cold" tumor microenvironments into "hot" landscapes, enhancing NK and CD8+ T cell recruitment and synergy with anti-PD-1 therapy. Mechanistically, Co68 engages the TLR4-MD2 complex via a non-canonical binding mode, bifurcating innate signaling: triggering the TLR4-TRIF-IFN-I axis while attenuating NF-κB-driven inflammation through an early, IFNAR-independent TLR4-SYK-STAT1 pathway. Collectively, Co68 represents a taxane-inspired precision therapeutic that uncouples beneficial antiviral-like immunity from pathogenic inflammation, offering a transformative strategy for refractory solid tumors.