High-Speed Terpene Pathway Analysis via Double-ended Transition State Search Method Combined with ML Potentials Uncovers the Intricate Rearrangement Reaction in Spiroalbatene Biosynthesis

Terpene cyclization reactions frequently proceed through highly concerted rearrangement processes, constructing complex molecular architectures. Identifying transition states (TSs) in such multistep mechanisms has traditionally required substantial computational cost and expert knowledge. In this study, we present a high-speed, high-accuracy, and high-resolution approach for reaction mechanism analysis by combining the Direct MaxFlux (DMF) method with the UMA machine learning potential. Key findings of this work are as follows. (i) Ultrafast TS search: Our method enabled identification of TS structures for a complex, multistep terpene-forming reaction in an average of 4 minutes—more than two orders of magnitude faster than conventional quantum mechanical (QM) scan-based approaches. (ii) DFT-level accuracy: The activation energies and TS geometries obtained by DMF/UMA were in excellent agreement with those from DFT calculations (MAE = 1.65 kcal/mol, average RMSD = 0.24 Å), demonstrating the method’s quantitative reliability. (iii) Mechanistic elucidation of spiroalbatene biosynthesis: Using this approach, we successfully analyzed the rearrangement reaction in spiroalbatene biosynthesis, which involves four concerted chemical events (one σ-bond cleavage and three σ-bond formations). Further DFT analysis revealed that a conformational change is required to reorient orbitals for efficient rearrangement. (iv) Absence of cyclopropylcarbinyl cation: Contrary to previously proposed reaction mechanism, cyclopropylcarbinyl cation does not form due to steric constraints. Specifically, ring strain in the 10-membered intermediate prevents orbital overlap necessary for cyclopropylcarbinyl cation formation. These results demonstrate that the DMF/UMA protocol enables efficient and accurate analysis of complex biosynthetic pathways that were previously difficult to access due to computational or technical barriers.