Authors: Rajeshkumar S. A, Vishrut Nath Jha
Abstract: The rapid evolution of artificial intelligence in chem- istry has enabled autonomous systems capable of exploring vast chemical spaces and identifying novel compounds with potential pharmacological value. We introduce MoleculAR, an autonomous agentic framework that integrates molecular relationship discov- ery, quantum-level stability analysis, and cheminformatics-based novelty verification. Given a set of input molecules, MoleculAR predicts potential co-functional partners using hybrid structural and functional similarity analysis, followed by energetic and stability evaluation through quantum chemical computations. Subsequently, the system performs compound novelty verification via cross-referencing with the ChEMBL database. Molecules that are predicted to be both chemically stable and absent from ChEMBL are shortlisted for further investigation. By combin- ing agentic reasoning, computational chemistry, and database- driven validation, MoleculAR establishes a closed-loop discovery pipeline that enhances efficiency in de novo compound identifica- tion. Experimental evaluations demonstrate MoleculAR’s ability to autonomously identify stable and novel molecular candidates across diverse chemical classes.
