Molecular design operates through fragment-based reasoning, yet many widely used molecular encodings rely on atom-level representations. This misalignment makes fragment-by-fragment construction and scaffold-centered designs hard to realize, especially for fused-ring systems. We present MOLCANO, a molecular language that represents molecules as chemically interpretable fragments linked by explicit junction tags, supporting single-bond, face-to-face, and even substructure-level connections for versatile scaffold assembly. It facilitates the construction of chemically complex structures while maintaining interpretability throughout the generation process. We validate MOLCANO across three open datasets spanning small-molecule drugs and organic electronic materials, covering de novo generation, superstructure growth/decoration, and medicinal-chemistry scaffolding. This also integrates with reinforcement learning for goal-oriented optimization of properties and with large language models for interactive, text-guided editing. By aligning computational representations with chemists’ modular reasoning, MOLCANO broadens the accessible chemical design space and accelerates molecular innovation in drug discovery, materials science, and beyond.
Na et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: