MolDeNovo

MolDeNovo is an evidence-grounded medicinal chemistry copilot that combines multimodal AI with computational chemistry to make molecular optimization more accessible and reliable. The platform transforms scientific literature into actionable molecular optimization workflows. Researchers can upload scientific figures containing chemical structures, define optimization goals, protect important molecular groups, and run optimization campaigns. MolDeNovo uses SAM 3.1 to identify molecular structures, Muse Wonder to interpret them, and Muse Spark 1.3 to plan evidence-informed molecular modifications using matched molecular pair statistics. RDKit executes the actual transformations and enforces chemical constraints, rejecting invalid modifications and recording the reasons for rejection. Researchers can examine molecular candidates, property changes and supporting evidence, then save projects or export PDF reports. The current implementation includes a matched molecular pair evidence base of 3,806 EGFR compounds and achieved 65% accuracy in predicting the direction of potency change on 245 held-out EGFR pairs, compared with a 57% majority-class baseline. MolDeNovo aims to democratize early-stage molecular optimization by giving researchers, particularly those in resource-constrained environments, access to an intelligent and computationally verifiable drug-design workflow.