Adaptive prioritized expansion for cost-conscious retrosynthetic route planning

Retrosynthetic planning plays a significant role in the synthesis of drug molecules, constituting a pivotal challenge at the intersection of computational chemistry and bioinformatics, as it demands algorithmic strategies that can deconstruct complex molecular architectures into experimentally accessible precursors. Conventional learning-driven search frameworks have advanced pathway discovery, yet their reliance on static node expansion impedes the exploration of economically and chemically optimal solutions. To overcome these limitations, we propose Adaptive Prioritized Expansion A* (APE-A*), a learning-driven search framework that integrates stochastic candidate selection with cost-aware heuristic evaluation. APE-A* dynamically modulates node prioritization according to both predicted synthetic feasibility and precursor cost through a molecular price prediction ensemble, enabling economic constraints to be incorporated directly into the search process. By sampling from a probabilistically ranked subset of candidate nodes, APE-A* balances exploration and exploitation in deep synthetic trees, mitigating combinatorial explosion while improving route quality. Experimental evaluation on 190 challenging target molecules from the USPTO benchmark demonstrates that APE-A* achieves a success rate of 96.84% and identifies lower-cost synthetic routes for 35.91% of the evaluated targets, yielding total cost savings of 10.6-17.4% relative to competing methods. These results indicate that incorporating molecular cost information into heuristic search can improve the practicality and economic efficiency of retrosynthetic planning.