GAP: Guided Diffusion for A Priori Transition State Sampling

The identification of transition states, first-order saddle points on high-dimensional potential energy surfaces, is a central challenge in the physical and life sciences, as they govern the kinetics and mechanisms of chemical reactions and conformational changes. Existing methods for locating these states often require knowledge about the transition, such as a good initial guess of the transition pathway or reaction coordinates. We introduce GAP, Guided Diffusion for A Priori Transition State Sampling, a new workflow that reframes this search problem as a direct generative task. GAP utilizes a score-based diffusion model trained exclusively on configurations from known metastable states, requiring no prior data from the transition region. During inference, we guide the generative process to sample from the dividing isodensity surface between the stable states with a principled composition of conditional scores. This process is coupled with a Score-Aligned Ascent mechanism that maximizes the energy along the score-based reaction coordinate approximation, effectively collapsing the sampling onto the transition state ensemble. We validate our approach on a series of benchmarks, from 2D potentials to the high-dimensional conformational changes of alanine dipeptide and the folding of the chignolin protein. Our results demonstrate that GAP not only locates transition states with high precision but also discovers competing reaction pathways, a new way of locating transition states in mechanistic studies.