AI-driven early-stage assessment strategy for rational design and synthesis of chitinase inhibitors

Lepidopteran pests cause severe agricultural losses worldwide, while traditional insecticides face increasing resistance and toxicity to nontarget organisms. There is an urgent need for novel insecticide discovery with high efficacy and low toxicity. However, the conventional "Design-Make-Test" cycle in pesticide discovery has long times and high failure rates. In this study, an AI-driven early-stage assessment strategy was introduced into chitinase inhibitor discovery and an efficient "Design-AI assessment-Make-Test" workflow was established. Based on the 4,5,6,7-tetrahydrobenzo[b]thiophene-3-carboxylate scaffold, pocket-based rational design was conducted. In silico evaluation was performed using three AI tools (APPi model, APPi rule, and BeeMSAI) and a total of 45 candidate compounds were selected and synthesized. Biological evaluation showed that compound 40 exhibited potent OfChtI inhibitory activity (K_i = 0.55 μM) and compound 1 showed insecticidal activity against Plutella xylostella comparable to the positive control tebufenozide (LC₅₀ = 25.36 mg/L). Importantly, 89% of the synthesized compounds showed insecticidal activity, supporting the utility of the AI early-stage assessment strategy. This study provides a representative case for AI-driven rational pesticide design and offers new insights for accelerating the discovery of novel environmentally friendly and efficient insecticides.