Reaction Enumeration Based on NBO-Informed Molecular Graphs

Enumeration of reaction products and pathways is essential for understanding chemical reactivity and for the rational design of new chemical technologies. Traditional graph-based enumeration methods represent molecules through atomic connectivity and simple valency rules, with bond breaking and formation encoded as edge modifications. However, this representation becomes ambiguous for many important classes of chemical systems, including transition-metal complexes, hypervalent compounds, and molecules with extensive resonance or delocalization. Here, we introduce a molecular graph representation constructed at the level of valence orbitals rather than atomic connectivity alone. When combined with quantum-mechanical natural bond orbital (NBO) analysis, the method automatically detects delocalized bonding motifs and adaptively increases reaction-step complexity when chemically warranted. This integration enables systematic and computationally efficient identification of chemically meaningful reaction pathways that lie beyond the practical scope of conventional graph-based approaches. We demonstrate the robustness of the method across challenging bonding environments, including organometallic catalytic cycles, pericyclic reactions in highly delocalized systems, main-group hypervalent chemistry, and multicomponent reaction networks.