A Priori Sampling of Transition States with Guided Diffusion

arXiv cs.LG / 3/30/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces ASTRA, a method to locate chemical and conformational transition states without relying on heuristic assumptions about reaction pathways or coordinates.
ASTRA reframes transition-state search as an inference-time scaling problem for a score-based diffusion generative model trained on configurations from known metastable states.
During inference, it guides trajectories toward the isodensity surface separating metastable basins using a principled composition of conditional scores, then applies a Score-Aligned Ascent (SAA) process for reaction-coordinate approximation.
By combining conditioned-score differences with physical forces, ASTRA converges onto first-order saddle points and is evaluated on benchmarks ranging from 2D potentials to biomolecular conformational changes and chemical reactions.
The results claim high-precision transition-state discovery and the ability to uncover multiple reaction pathways, supporting mechanistic studies of complex molecular systems.

Abstract

Transition states, the first-order saddle points on the potential energy surfaces, govern the kinetics and mechanisms of chemical reactions and conformational changes. Locating them is challenging because transition pathways are topologically complex and can proceed via an ensemble of diverse routes. Existing methods address these challenges by introducing heuristic assumptions about the pathway or reaction coordinates, which limits their applicability when a good initial guess is unavailable or when the guess precludes alternative, potentially relevant pathways. We propose to bypass such heuristic limitations by introducing ASTRA, A Priori Sampling of TRAnsition States with Guided Diffusion, which reframes the transition state search as an inference-time scaling problem for generative models. ASTRA trains a score-based diffusion model on configurations from known metastable states. Then, ASTRA guides inference toward the isodensity surface separating the basins of metastable states via a principled composition of conditional scores. A Score-Aligned Ascent (SAA) process then approximates a reaction coordinate from the difference between conditioned scores and combines it with physical forces to drive convergence onto first-order transition states. Validated on benchmarks from 2D potentials to biomolecular conformational changes and chemical reaction, ASTRA locates transition states with high precision and discovers multiple reaction pathways, enabling mechanistic studies of complex molecular systems.