Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Adaptive Transition State Refinement with Learned Equilibrium Flows

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Learned equilibrium flow refines transition-state guesses to 0.088 Å.

desk verdict AEFM introduces a genuinely new time-independent flow refinement scheme with real practical gains, but its headline accuracy numbers partly reflect convergence to alternative TSs, so treat the intended-reaction numbers with a grain of salt. read the letter →

arxiv 2507.16521 v1 pith:27TZPBV3 submitted 2025-07-22 physics.chem-ph physics.comp-ph

Samir Darouich , Vinh Tong , Tanja Bien , Johannes Kästner , Mathias Niepert This is my paper · ORCID
classification physics.chem-phphysics.comp-ph
keywords transitionstaterefinementflowmatchingfixed-pointinferenceequivariantneuralnetworksearchreactionbarrierheightsGFN2-xTBDFToptimization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

AEFM is a structure-only refinement method that takes a low-fidelity transition-state geometry, whether from a machine-learning generative model or a tight-binding approximation, and iteratively maps it to a DFT-quality TS structure using only atomic coordinates and no energy or gradient evaluations. The paper claims this learned fixed-point flow reduces the median RMSD of React-OT predictions to 0.088 Å and the median absolute barrier-height error to 0.79 kcal/mol; for GFN2-xTB guesses it cuts the median barrier error by 59% and raises the fraction of chemically valid TS structures from 27% to 68%. If true, transition-state search becomes substantially cheaper, because refined guesses require roughly threefold fewer DFT optimization CPU hours and the refinement itself adds only fractions of a second per structure.

What carries the argument

The load-bearing object is an equilibrium flow field: a neural network $\phi_\theta(x)$ that predicts the high-fidelity TS endpoint directly from a structure, trained with variational flow matching on pairs $(x_1+\sigma\epsilon, x_1)$ where $x_1$ is a reference TS and $\sigma$ is set by the source method's mean RMSD. Inference applies the fixed-point iteration $x_{k+1}=\phi_\theta(x_k)$ with Anderson acceleration until successive iterates differ by less than 0.01 Å RMSD, or until 100 iterations. A physics-based bond loss compares interatomic distances within a 2 Å cutoff against the reference, steering outputs toward realistic bond-length distributions, and the SE(3)-equivariant backbone makes the entire map invariant to rotation, translation, and atom-index permutation.

What would settle it

A paired test set in which low-fidelity TS guesses come from a method with systematically anisotropic errors, for example consistent stretching of the forming and breaking bonds, while AEFM is trained with the isotropic Gaussian prior; if the refiner then maps many inputs to alternative TSs or raises barrier-height error relative to the unrefined guesses, the central assumption is falsified.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a time-independent flow-matching model, trained to predict the clean TS directly from a Gaussian-perturbed version of it and iterated to a fixed point at inference, can refine TS guesses from diverse sources toward the distribution of DFT-level transition states. The adaptive prior sets the perturbation scale from the mean RMSD of the source method, the SE(3)-equivariant backbone respects molecular symmetry, and a bond-length loss keeps local geometries chemically plausible. The result is a general refiner: it improves the median barrier-height error by 27% over React-OT alone and by 59% over GFN2-xTB alone, and it makes downstream DFT saddle-point optimization faster and more likely to converge to a valid TS.

Load-bearing premise

The load-bearing premise is that a low-fidelity TS guess behaves like a reference TS plus isotropic Gaussian noise whose scale is the source method's mean RMSD; if real errors are structured or method-specific, fixed-point refinement can settle on the wrong transition state or fail to converge, as seen in 6 of 1073 React-OT cases and 3 of 945 xTB cases in the paper's tests.

Editorial extensions

If this is right

  • A refiner trained with the appropriate error scale can be bolted onto various low-fidelity TS guessing methods, requiring only 2–5 model calls per structure and sub-second inference.
  • Refined GFN2-xTB guesses meet the 1.58 kcal/mol chemical-accuracy threshold for 57% of test reactions, up from 25% without refinement.
  • AEFM raises the fraction of structurally valid TSs, defined by exactly one imaginary frequency, from 27% to 68% and raises DFT TS-optimization convergence from 91% to 99%.
  • Fast refinement cuts the median number of DFT optimization steps by 10, a threefold reduction in CPU hours for the tested 100-reaction set.
  • Because refinement consumes no potential-energy-surface evaluations, it can be embedded in high-throughput reaction screening pipelines.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the isotropic Gaussian prior is the main risk; a direct test is to train on paired real low-fidelity and reference TS geometries rather than Gaussian perturbations and compare barrier errors and TS-assignment rates.
  • Editorial inference: because AEFM has no reactant or product context, it can legitimately converge to a different but structurally similar TS; re-ranking refined candidates against the reaction endpoints could recover the intended transition state without sacrificing the structure-only pipeline.
  • Editorial inference: the weak correlation between RMSD and energy change (Pearson 0.17) suggests that future refiners should explicitly optimize local bond, angle, and torsion geometry; adding such terms to the bond loss could further reduce barrier-height errors.
  • Editorial inference: the fixed-point formulation is a learned denoiser, so the same adaptive iteration could be applied to other low-fidelity-to-high-fidelity structure-refinement problems, though the paper only reports chemistry results.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This manuscript introduces Adaptive Equilibrium Flow Matching (AEFM), a structure-only refinement method that aims to improve low-fidelity transition state (TS) guesses. The model is trained to denoise Gaussian perturbations of reference TS geometries from the Transition1x dataset, with the noise scale σ determined per low-fidelity source by matching the mean RMSD of that source. At inference, the time-independent model is applied iteratively (with Anderson acceleration) to convergence, producing a refined structure. The paper reports that AEFM reduces the median RMSD of React-OT TS guesses from 0.092 Å to 0.088 Å, lowers the median absolute barrier-height error from 1.092 to 0.793 kcal/mol, increases the fraction of valid GFN2-xTB TS structures from 27% to 68% on a 100-reaction subset, and reduces the median number of DFT optimization steps by 10 for xTB-initialized searches. The authors also propose a bond-length loss that improves agreement with reference bond-length distributions.

Significance. The paper addresses a practical bottleneck in computational chemistry: the need to refine approximate TS guesses into DFT-quality saddle points. The proposed method is lightweight (a fraction of a second per structure), model-agnostic, and does not require energy or gradient evaluations. The authors are transparent about a key failure mode: AEFM can converge to a chemically valid but unintended TS, and they analyze this in Fig. 3c and the footnote to Table 1. The method is evaluated on a standard benchmark (Transition1x) and compared against established baselines. The main benchmark is not circular because the model is trained on Gaussian perturbations and evaluated on held-out test data. If the validity of the refined structures for the React-OT benchmark can be established, the method would be a useful contribution. However, several load-bearing claims currently lack direct support.

major comments (4)
  1. [Section 4.4 and Table 1] The chemical validity of refined outputs, defined by exactly one imaginary frequency, is reported only for GFN2-xTB-initialized structures on a 100-reaction subset. For the React-OT outputs that produce the headline numbers (median RMSD 0.088 Å, median |ΔE_TS| 0.793 kcal/mol), no analogous frequency analysis is reported. Since the barrier-height errors in Table 1 are computed as single-point energy differences at the refined geometries, and these geometries are not verified to be stationary saddle points, the reported energies may not be meaningful TS energies. Please provide frequency or stationarity checks for a representative subset of React-OT-refined structures, or clearly state that the metrics are computed on non-stationary structures and discuss the implications for the claimed high-fidelity TS geometries.
  2. [Section 4.2.1, Eq. (10)] The training loss is written as an expectation over x0, x1, and t, but φθ has no time input. If t is only used to define the interpolant xt via Eq. (2), this should be stated explicitly, including how t is sampled (e.g., uniform on [0,1]) and whether the expectation is over the resulting mixture distribution. More substantively, the paper claims that removing time conditioning enables the model to implicitly infer the quality of a given input, but a model that sees only the structure cannot distinguish a small isotropic perturbation from a large one. The adaptive behavior of AEFM is currently demonstrated only through the number of fixed-point iterations; no evidence is provided that the iteration count correlates with initial error magnitude. Please add such an analysis (e.g., iterations versus initial RMSD) or revise the adaptivity claims.
  3. [Section 2, Table 1] The mean RMSD increases after AEFM for both React-OT rows (0.183 to 0.188 Å and 0.211 to 0.214 Å), and the mean barrier error for React-OT improves only from 3.405 to 3.341 kcal/mol (approximately 2%). The improvements highlighted in the abstract are median-based, and the distribution appears to be skewed by cases such as the outlier in Fig. 3c, where the barrier error increases from 17.9 to 121.0 kcal/mol. The paper should discuss the cause of the mean degradation and report the fraction of samples that are made worse by refinement, in addition to the median improvements, to support the claim of robust refinement.
  4. [Section 4.2.1, Eqs. (6)-(9)] The training prior models real low-fidelity errors as isotropic Gaussian noise with a source-specific σ fitted to the mean RMSD. This assumption is acknowledged in the paper, but the consequences are not fully quantified. The model must generalize from Gaussian perturbations to the structured, method-specific errors of React-OT and xTB. The paper's own outlier analysis shows that fixed-point refinement can converge to an alternative, chemically valid TS, and the footnote to Table 1 identifies 26/1073 React-OT cases where the alternative TS is at least 30% closer than the intended TS. For a method whose stated purpose is to refine guesses for a specific reaction, the frequency and characteristics of such unintended convergence should be analyzed more directly (e.g., by comparing refined structures against all nearby TSs in the dataset and reporting how often the refined sample matches an alternative TS). Please provide this analysis or explicitly frame the method as a find-any-nearby-TS tool rather than a refine-the-intended-TS tool.
minor comments (6)
  1. [Table 1, footnote (a)] The re-labeling row should be clearly described as an exploratory post hoc analysis, distinct from the main benchmark, to avoid the impression that the main results use re-labeled targets.
  2. [Section 4.5, Eq. (17)] The RMSD definition differs from that used in React-OT (normalization by 3N versus N). The paper notes this in one sentence, but it would help to provide baseline RMSD values recalculated with the same metric for a fair comparison.
  3. [Figure 3c] The caption and the main text disagree on which axis corresponds to the intended TS: the caption says the x-axis is the intended TS, while the text says the y-axis is the intended TS. Please reconcile this inconsistency.
  4. [Section 4.2.1] Equation (6) defines x0 = x1 + σϵ, but two lines below the text writes x0 = x1 − σϵ; the sign is inconsequential, but the inconsistency should be fixed.
  5. [Abstract] The phrase increases the success rate of locating valid TSs by 41% is ambiguous; the main text specifies that this is an absolute percentage-point increase from 27% to 68%, and the abstract should say so explicitly.
  6. [Section 4.3] The bond loss is defined using the ground-truth neighbor list B(x1); the paper should state explicitly that this loss is used only in training and not at inference.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity: the central refinement benchmark is independent of the model's training objective, though the per-source noise-scale calibration is a mild adaptation rather than a circular step.

full rationale

AEFM's derivation chain is self-contained against external benchmarks. The model is trained to denoise Gaussian perturbations of reference Transition1x TS geometries (Eqs. 6, 10) and is then iterated at inference time on actual low-fidelity guesses that were never used to fit the model (Eq. 11). The headline metrics are computed on held-out test reactions against independent DFT reference structures and single-point ωB97x/6-31G(d) energies, and some reported mean RMSDs worsen after refinement, which shows the evaluation is not forced by construction. The per-source noise scale σ is matched to the mean RMSD of each low-fidelity method (Eqs. 7-9, Table S7), so the 'adaptive' behavior is partly a domain-calibration choice, but this does not make the test predictions equivalent to the fitted input: the model must still generalize from Gaussian perturbations to the structured errors of React-OT and GFN2-xTB, and the paper documents failures and alternative-TS convergence. The 26-case re-labeling in Table 1 footnote (a) and the absence of an imaginary-frequency validity check for React-OT samples are evaluation-validity concerns, not circularity. No load-bearing self-citations were found; references to prior generative TS models are external, and no 'uniqueness' argument is smuggled in from the authors' own work.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a learned mapping trained on Gaussian perturbations of reference TSs. The free parameters are the per-source noise scale and the solver hyperparameters. The key axiomatic step is that Gaussian noise adequately represents the errors of real low-fidelity TS sources, which is not derived.

free parameters (5)
  • sigma noise scale per source = 0.19 (xTB), 0.12 (React-OT xTB), 0.11 (React-OT)
    Set via Eq. 9 from the mean RMSD of each low-fidelity source on training data; controls the training perturbation and hence the refinement behavior.
  • beta damping parameter = 1.0
    Chosen by hyperparameter search for Anderson acceleration (Section 4.2.2).
  • m history size = 5
    Chosen by hyperparameter search for Anderson acceleration (Section 4.2.2).
  • rcut bond-loss cutoff = 2.0 Å
    Chosen based on longest equilibrium bond lengths for C, N, O, H with margin (Section 4.3).
  • wb bond loss weight = 1.0
    Fixed hyperparameter weighting the physical consistency loss (Eq. 16).
assumptions (4)
  • ad hoc to paper Low-fidelity TS errors are modeled as isotropic Gaussian noise around the reference TS (Eqs. 6-9).
    This is the central assumption that makes sigma computable from mean RMSD and justifies training with Gaussian perturbations; real errors are structured and method-specific.
  • domain assumption The Transition1x reference TS is the intended target for each reaction.
    Used to define the supervised loss (Eq. 10) and all metrics; the paper later shows alternative TSs exist and can be closer (Figure 3c).
  • ad hoc to paper Fixed-point iteration of phi_theta converges to a chemically valid TS.
    Relying on spectral radius arguments and empirical convergence checks, not on a guarantee; 6 out of 1073 and 3 out of 945 samples fail to converge.
  • domain assumption Single-point DFT energies at predicted TS geometries are valid evaluation metrics.
    Metric in Eq. 18 compares energies at possibly non-stationary points rather than optimized TS energies.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive Transition State Refinement with Learned Equilibrium Flows." pith.science (2026). https://pith.science/paper/27TZPBV3

@misc{pith2026250716521,
  author       = {Pith},
  title        = {Pith review of: Adaptive Transition State Refinement with Learned Equilibrium Flows},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/27TZPBV3}},
  note         = {Machine review of arXiv:2507.16521}
}
abstract

Identifying transition states (TSs), the high-energy configurations that molecules pass through during chemical reactions, is essential for understanding and designing chemical processes. However, accurately and efficiently identifying these states remains one of the most challenging problems in computational chemistry. In this work, we introduce a new generative AI approach that improves the quality of initial guesses for TS structures. Our method can be combined with a variety of existing techniques, including both machine learning models and fast, approximate quantum methods, to refine their predictions and bring them closer to chemically accurate results. Applied to TS guesses from a state-of-the-art machine learning model, our approach reduces the median structural error to just 0.088 $\unicode{x212B}$ and lowers the median absolute error in reaction barrier heights to 0.79 kcal mol$^{-1}$. When starting from a widely used tight-binding approximation, it increases the success rate of locating valid TSs by 41\% and speeds up high-level quantum optimization by a factor of three. By making TS searches more accurate, robust, and efficient, this method could accelerate reaction mechanism discovery and support the development of new materials, catalysts, and pharmaceuticals.

Figures

Figures reproduced from arXiv: 2507.16521 by the authors.

Figure 1
Figure 1. AEFM pipeline for TS structure refinement. a The input consists of low-fidelity TS samples, which may originate from various sources such as ML models or tight-binding approximations. These inputs are iteratively refined to produce high-fidelity, chemically valid TS geometries near the DFT level. b Comparison between actual physical flow and the one learned by AEFM on the Müller–Brown potential energy surface. Integ… view at source ↗
Figure 2
Figure 2. Performance summary of AEFM across diverse low-fidelity sources. a Percentage of test samples showing improvement in energy difference |∆E| relative to the reference TS (irrespective of RMSD), in RMSD (irrespective of energy), and in both RMSD and energy difference (RMSD ∩ |∆E|). b Histogram (colored, left y-axis) and cumulative distribution (grey, right y-axis) of the change in energy difference between the low-fid… view at source ↗
Figure 3
Figure 3. Relationship between energetic and structural changes in AEFM refinements, with a focus on outliers and correlation trends. a Energetic differences of AEFM-refined structures versus initial React-OT predictions on the left y-axis. Points below the diagonal line indicate improved agreement with the reference TS, while points above reflect increased deviation. On the right y-axis, the KDE of improvement weighted by th… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Bond length distributions. Distributions of C–H, C–N, and N–O bond lengths in the Transi￾tion1x dataset compared to those in the React-OT and AEFM-refined structures. sensitivities across different degrees of freedom. To address this, AEFM incorporates an additional bo…
Figure 5
Figure 5. Figure 5: Chemical validation and fixed-point convergence analysis. a Fraction of valid TS struc￾tures, defined by the presence of exactly one imaginary frequency in the Hessian. b Convergence rate of DFT TS optimizations. c Boxplot of DFT optimization steps required to reach a …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Feynman-Kac-Flow: Inference Steering of Conditional Flow Matching to an Energy-Tilted Posterior

    cs.LG 2025-09 conditional novelty 6.0 of 10

    Feynman-Kac particle steering, previously diffusion-only, is derived for conditional flow matching and used to generate chirality-correct chemical transition states.

Reference graph

Works this paper leans on

68 extracted references · 67 canonical work pages · cited by 1 Pith paper

  1. [1]

    G., Garrett, B

    Truhlar, D. G., Garrett, B. C. & Klippenstein, S. J. Current Status of Transition-State Theory. J. Phys. Chem. 100, 12771–12800 (1996)

    work page 1996

  2. [2]

    & Paton, R

    Peng, Q., Duarte, F. & Paton, R. S. Computing organic stereoselectivity–from concepts to quan- titative calculations and predictions. Chem. Soc. Rev. 45, 6093–6107 (2016)

    work page 2016

  3. [3]

    L., Argüelles, A

    Dewyer, A. L., Argüelles, A. J. & Zimmerman, P. M. Methods for exploring reaction space in molecular systems. WIREs Comput. Mol. Sci. 8, e1354 (2018)

    work page 2018

  4. [4]

    A., Müller, K.-R

    von Lilienfeld, O. A., Müller, K.-R. & Tkatchenko, A. Exploring chemical compound space with quantum-based machine learning. Nat. Rev. Chem. 4, 347–358 (2020)

    work page 2020

  5. [5]

    Unsleber, J. P. & Reiher, M. The Exploration of Chemical Reaction Networks. Annu. Rev. Phys. Chem. 71, 121–142 (2020)

    work page 2020

  6. [6]

    Nandy, A. et al. Computational Discovery of Transition-metal Complexes: From High- throughput Screening to Machine Learning. Chem. Rev. 121, 9927–10000 (2021)

    work page 2021

  7. [7]

    & Norrby, P.-O

    Jorner, K., Tomberg, A., Bauer, C., Sköld, C. & Norrby, P.-O. Organic reactivity from mech- anism to machine learning. Nat. Rev. Chem. 5, 240–255 (2021)

    work page 2021

  8. [8]

    Taylor, C. J. et al. A brief introduction to chemical reaction optimization. Chem. Rev. 123, 3089–3126 (2023)

    work page 2023

Show all 68 references
  1. [9]

    & Deutschmann, O

    Chacko, R., Gossler, H., Angeli, S. & Deutschmann, O. Interconnected digital solu- tions to accelerate modeling of the reaction kinet- ics in catalysis. ChemCatChem 16, e202301355 (2024)

    2024

  2. [10]

    & Shepard, R

    Banerjee, A., Adams, N., Simons, J. & Shepard, R. Search for stationary points on surfaces. J. Phys. Chem. 89, 52–57 (1985)

    1985

  3. [11]

    An algorithm for the location of transi- tion states

    Baker, J. An algorithm for the location of transi- tion states. J. Comput. Chem. 7, 385–395 (1986)

    1986

  4. [12]

    & Jónsson, H

    Henkelman, G. & Jónsson, H. A dimer method for finding saddle points on high dimensional po- tential surfaces using only first derivatives. J. Chem. Phys. 111, 7010–7022 (1999)

    1999

  5. [13]

    & Jacobsen, K

    Jónsson, H., Mills, G. & Jacobsen, K. W. Nudged elastic band method for finding mini- mum energy paths of transitions. In Classical and quantum dynamics in condensed phase sim- ulations, 385–404 (World Scientific, 1998)

    1998

  6. [14]

    Henkelman, G., Uberuaga, B. P. & Jónsson, H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys. 113, 9901–9904 (2000)

    2000

  7. [15]

    Peters, B., Heyden, A., Bell, A. T. & Chakraborty, A. A growing string method for determining transition states: Comparison to the nudged elastic band and string methods. J. Chem. Phys. 120, 7877–7886 (2004)

    2004

  8. [16]

    & Head-Gordon, M

    Mardirossian, N. & Head-Gordon, M. Thirty years of density functional theory in computa- tional chemistry: an overview and extensive as- sessment of 200 density functionals. Mol. Phys. 115, 2315–2372 (2017). 13

    2017

  9. [17]

    Pozun, Z. D. et al. Optimizing transition states via kernel-based machine learning. J. Chem. Phys. 136, 174101 (2012)

    2012

  10. [18]

    & Jóns- son, H

    Koistinen, O.-P., Maras, E., Vehtari, A. & Jóns- son, H. Minimum energy path calculations with Gaussian process regression. Nanosyst.: Phys. Chem. Math. 925–935 (2016)

    2016

  11. [19]

    B., Ásgeirs- son, V., Vehtari, A

    Koistinen, O.-P., Dagbjartsdóttir, F. B., Ásgeirs- son, V., Vehtari, A. & Jónsson, H. Nudged elas- tic band calculations accelerated with Gaussian process regression. J. Chem. Phys. 147, 152720 (2017)

    2017

  12. [20]

    & Kästner, J

    Denzel, A. & Kästner, J. Gaussian Process Re- gression for Transition State Search. J. Chem. Theory Comput. 14, 5777–5786 (2018)

    2018

  13. [21]

    & Kästner, J

    Denzel, A., Haasdonk, B. & Kästner, J. Gaussian Process Regression for Minimum Energy Path Optimization and Transition State Search. J. Phys. Chem. A 123, 9600–9611 (2019)

    2019

  14. [22]

    A., Jennings, P

    Garrido Torres, J. A., Jennings, P. C., Hansen, M. H., Boes, J. R. & Bligaard, T. Low-Scaling Algorithm for Nudged Elastic Band Calculations Using a Surrogate Machine Learning Model. Phys. Rev. Lett. 122, 156001 (2019)

    2019

  15. [23]

    Heinen, S., von Rudorff, G. F. & von Lilienfeld, O. A. Transition state search and geometry re- laxation throughout chemical compound space with quantum machine learning. J. Chem. Phys. 157, 221102 (2022)

    2022

  16. [24]

    Peterson, A. A. Acceleration of saddle-point searches with machine learning. J. Chem. Phys. 145, 074106 (2016)

    2016

  17. [25]

    Schreiner, M., Bhowmik, A., Vegge, T., Jør- gensen, P. B. & Winther, O. NeuralNEB—neural networks can find reaction paths fast. Mach. Learn.: Sci. Technol. 3, 045022 (2022)

    2022

  18. [26]

    Zhang, S. et al. Exploring the frontiers of condensed-phase chemistry with a general reac- tive machine learning potential. Nat. Chem. 16, 727–734 (2024)

    2024

  19. [27]

    R., Ulissi, Z

    Wander, B., Shuaibi, M., Kitchin, J. R., Ulissi, Z. W. & Zitnick, C. L. Cattsunami: Accelerat- ing transition state energy calculations with pre- trained graph neural networks. ACS Catal. 15, 5283–5294 (2025)

    2025

  20. [28]

    Yuan, E. C.-Y. et al. Analytical ab initio hessian from a deep learning potential for transition state optimization. Nat. Commun. 15, 8865 (2024)

    2024

  21. [29]

    Zhao, Q. et al. Harnessing machine learning to enhance transition state search with inter- atomic potentials and generative models. Adv. Sci., e06240 (2025)

    2025

  22. [30]

    B., Grambow, C

    Pattanaik, L., Ingraham, J. B., Grambow, C. A. & Green, W. H. Generating transition states of isomerization reactions with deep learning. Phys. Chem. Chem. Phys. 22, 23618–23626 (2020)

    2020

  23. [31]

    & Pearson, J

    Jackson, R., Zhang, W. & Pearson, J. TSNet: predicting transition state structures with tensor field networks and transfer learning. Chem. Sci. 12, 10022–10040 (2021)

    2021

  24. [32]

    Zhang, J. et al. Deep reinforcement learning of transition states. Phys. Chem. Chem. Phys. 23, 6888–6895 (2021)

    2021

  25. [33]

    Prediction of transition state struc- tures of gas-phase chemical reactions via machine learning

    Choi, S. Prediction of transition state struc- tures of gas-phase chemical reactions via machine learning. Nat. Commun. 14, 1168 (2023)

    2023

  26. [34]

    Z., Verma, N., Larson, E

    Makoś, M. Z., Verma, N., Larson, E. C., Frein- dorf, M. & Kraka, E. Generative adversarial net- works for transition state geometry prediction. J. Chem. Phys. 155, 024116 (2021)

    2021

  27. [35]

    & Kulik, H

    Duan, C., Du, Y., Jia, H. & Kulik, H. J. Accurate transition state generation with an object-aware equivariant elementary reaction diffusion model. Nat. Comput. Sci. 3, 1045–1055 (2023)

    2023

  28. [36]

    & Kim, W

    Kim, S., Woo, J. & Kim, W. Y. Diffusion-based generative AI for exploring transition states from 2D molecular graphs. Nat. Commun. 15, 341 (2024)

    2024

  29. [37]

    Galustian, L., Mark, K., Karwounopoulos, J., Kovar, M. P.-P. & Heid, E. GoFlow: Efficient Transition State Geometry Prediction with Flow Matching and E(3)-Equivariant Neural Networks. Preprint at https://chemrxiv. org/engage/chemrxiv/article-details/ 680f4ca9e561f77ed42b9163

  30. [38]

    Duan, C. et al. Optimal transport for generat- ing transition states in chemical reactions. Nat. Mach. Intell. 7, 615–626 (2025)

    2025

  31. [39]

    & Okanohara, D

    Hayashi, A., Takamoto, S., Li, J., Tsuboi, Y. & Okanohara, D. Generative Model for Construct- ing Reaction Path from Initial to Final States. J. Chem. Theory Comput. 21, 1292–1305 (2025)

    2025

  32. [40]

    Lipman, Y., Chen, R. T. Q., Ben-Hamu, H., Nickel, M. & Le, M. Flow Matching for Gen- erative Modeling. In The Eleventh International Conference on Learning Representations (2022). 14

    2022

  33. [41]

    & Liu, Q

    Liu, X., Gong, C. & Liu, Q. Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow. In The Eleventh Interna- tional Conference on Learning Representations (2022)

    2022

  34. [42]

    Peng, X., Guan, J., Liu, Q. & Ma, J. MolDiff: Addressing the Atom-Bond Inconsistency Prob- lem in 3D Molecule Diffusion Generation. In Pro- ceedings of the 40th International Conference on Machine Learning, 27611–27629 (PMLR, 2023)

    2023

  35. [43]

    Williams, D. C. & Inala, N. Physics-Informed Generative Model for Drug-like Molecule Con- formers. J. Chem. Inf. Model. 64, 2988–3007 (2024)

    2024

  36. [44]

    & Deane, C

    Vost, L., Chenthamarakshan, V., Das, P. & Deane, C. M. Improving structural plausi- bility in diffusion-based 3D molecule genera- tion via property-conditioned training with dis- torted molecules. Digital Discovery 4, 1092–1099 (2025)

    2025

  37. [45]

    Wohlwend, J. et al. Boltz-1 Democratizing Biomolecular Interaction Modeling. Preprint at https://www.biorxiv.org/content/10.1101/ 2024.11.19.624167v4

    2024

  38. [46]

    Rasmussen, M. H. & Jensen, J. H. Fast and au- tomatic estimation of transition state structures using tight binding quantum chemical calcula- tions. PeerJ Phys. Chem. 2, e15 (2020)

    2020

  39. [47]

    Amini, A., De Bortoli, V., Maddison, C. J. & Gal, Y. Variational flow matching: A unifying perspective on score-based generative modeling. In Advances in Neural Information Processing Systems (2024)

    2024

  40. [48]

    & Winther, O

    Schreiner, M., Bhowmik, A., Vegge, T., Busk, J. & Winther, O. Transition1x - a dataset for building generalizable reactive machine learning potentials. Sci. Data 9, 779 (2022)

    2022

  41. [49]

    & Grimme, S

    Bannwarth, C., Ehlert, S. & Grimme, S. Gfn2- xtb—an accurate and broadly parametrized self-consistent tight-binding quantum chemi- cal method with multipole electrostatics and density-dependent dispersion contributions. J. Chem. Theory Comput. 15, 1652–1671 (2019)

    2019

  42. [50]

    Du, Y. et al. A new perspective on building effi- cient and expressive 3d equivariant graph neural networks. Adv. Neural Inf. Process. Syst. 36, 66647–66674 (2023)

    2023

  43. [51]

    Ditchfield, R., Hehre, W. J. & Pople, J. A. Self- consistent molecular-orbital methods. ix. an ex- tended gaussian-type basis for molecular-orbital studies of organic molecules. J. Chem. Phys. 54, 724–728 (1971)

    1971

  44. [52]

    & Head-Gordon, M

    Chai, J.-D. & Head-Gordon, M. Systematic op- timization of long-range corrected hybrid density functionals. J. Chem. Phys. 128 (2008)

    2008

  45. [53]

    A., Pattanaik, L

    Grambow, C. A., Pattanaik, L. & Green, W. H. Reactants, products, and transition states of el- ementary chemical reactions based on quantum chemistry. Sci. Data 7, 137 (2020)

    2020

  46. [54]

    J., Sancho-García, J

    Brémond, É., Li, H., Pérez-Jiménez, Á. J., Sancho-García, J. C. & Adamo, C. Tackling an accurate description of molecular reactivity with double-hybrid density functionals. J. Chem. Phys. 156 (2022)

    2022

  47. [55]

    & Cai, W

    Fu, H., Zhou, Y., Jing, X., Shao, X. & Cai, W. Meta-analysis reveals that absolute binding free- energy calculations approach chemical accuracy. J. Med. Chem. 65, 12970–12978 (2022)

    2022

  48. [56]

    M., Isayev, O

    Zhao, Q., Anstine, D. M., Isayev, O. & Savoie, B. M. δ 2 machine learning for reaction property prediction. Chem. Sci. 14, 13392–13401 (2023)

    2023

  49. [57]

    Falck, F. et al. A fourier space perspective on dif- fusion models. arXiv preprint arXiv:2505.11278 (2025)

    2025 arXiv

  50. [58]

    Lam, Y.-h. et al. Applications of quantum chem- istry in pharmaceutical process development: Current state and opportunities. Org. Process Res. Dev. 24, 1496–1507 (2020)

    2020

  51. [59]

    & Kolter, Z

    Bai, S., Koltun, V. & Kolter, Z. Stabilizing Equi- librium Models by Jacobian Regularization. In Proceedings of the 38th International Conference on Machine Learning , 554–565 (PMLR, 2021)

    2021

  52. [60]

    Broyden, C. G. A class of methods for solving nonlinear simultaneous equations. Math. Com- put. 19, 577–593 (1965)

    1965

  53. [61]

    Anderson, D. G. Iterative procedures for non- linear integral equations. J. ACM 12, 547–560 (1965)

    1965

  54. [62]

    Tong, A. et al. Improving and generalizing flow- based generative models with minibatch optimal transport. Trans. Mach. Learn. Res. (2024)

    2024

  55. [63]

    & Jaakkola, T

    Stark, H., Jing, B., Barzilay, R. & Jaakkola, T. Harmonic Self-Conditioned Flow Matching 15 for joint Multi-Ligand Docking and Binding Site Design. In Proceedings of the 41st International Conference on Machine Learning , 46468–46494 (PMLR, 2024)

    2024

  56. [64]

    Bai, S., Kolter, J. Z. & Koltun, V. Deep Equilib- rium Models. In Advances in Neural Information Processing Systems, vol. 32 (Curran Associates, Inc., 2019)

    2019

  57. [65]

    A new C–C bond formation model based on the quantum chemical topology of electron density

    R.Domingo, L. A new C–C bond formation model based on the quantum chemical topology of electron density. RSC Adv. 4, 32415–32428 (2014)

    2014

  58. [66]

    Software update: the orca program system, version 5.0

    Neese, F. Software update: the orca program system, version 5.0. WIREs Comput. Mol. Sci. 12, e1606 (2022)

    2022

  59. [67]

    Larsen, A. H. et al. The atomic simulation en- vironment—a python library for working with atoms. J. Phys.: Condens. Matter 29, 273002 (2017)

    2017

  60. [68]

    Adaptive Transition State Refinement with Learned Equilibrium Flows

    Hermes, E. D., Sargsyan, K., Najm, H. N. & Zá- dor, J. Sella, an open-source automation-friendly molecular saddle point optimizer. J. Chem. The- ory Comput. 18, 6974–6988 (2022). 16 Supplementary Information for "Adaptive Transition State Refinement with Learned Equilibrium Flo...

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.