Pith. sign in

REVIEW 3 major objections 5 minor 43 references

TempRe: Template generation for single and direct multi-step retrosynthesis

T0 review · 3 major / 5 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 Writing reaction rules beats writing products in retrosynthesis.

desk verdict TempRe convincingly shows template generation as sequence modeling works for single-step and search-based multi-step planning, and adds a plausible direct multi-step variant; the 'chemically plausible' claim is softer than the evaluation supports. read the letter →

arxiv 2507.21762 v2 pith:4U3YUJDC submitted 2025-07-29 cs.LG

Nguyen Xuan-Vu , Daniel P Armstrong , Zlatko Jončev , Philippe Schwaller This is my paper · ORCID
classification cs.LG
keywords templategenerationretrosynthesisreactiontemplatesSMARTSsequence-to-sequencemulti-stepsynthesisplanningPaRoutesbenchmarkout-of-distributiongeneralization
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

TempRe reframes template-based retrosynthesis as a text-generation task: a Transformer reads a product SMILES and writes the reaction template as a SMARTS string, rather than choosing from a fixed template library or writing reactant SMILES directly. The paper's central claim is that this 'generate the rule, apply the rule' design keeps the chemical validity of template methods while removing their dependence on a fixed library, so the model can scale to hundreds of thousands of templates without growing its parameter count. On the PaRoutes benchmark, TempRe models reached 96% top-80 single-step accuracy, outperforming both a template classifier and a product-to-reactant Transformer, and on molecules far larger than those in training the template generator degraded much less than the reactant generator. In multi-step planning, a strict TempRe variant inside Monte Carlo Tree Search achieved high solve rates and top-k route accuracy, and a direct variant generated entire routes as a sequence of templates in a single pass. A sympathetic reader should care because this positions template generation as a scalable middle path between rigid rule libraries and chemically unreliable free-form generation.

What carries the argument

The central object is the reaction template as a SMARTS string—a substructural pattern describing which atoms react and how—treated as a token sequence for autoregressive generation. The framework factorizes the precursor distribution as $p(R|o)=\sum_{t\in\mathcal{T}} q(R|o,t)\,p_\theta(t|o)$, where $p_\theta$ is a product-to-template Transformer, $\mathcal{T}$ is an external library of valid templates, and $q$ applies the generated template to the product through RDChiral. Two tokenization schemes are explored: plain BPE (P2T) and a frequency-sensitive scheme (P2T-Tok) that encodes popular templates as single tokens; 'strict' variants filter generated templates to those seen in training. For direct multi-step planning, the same generate-then-apply idea is extended by conditioning on the number of reaction steps and emitting a sequence of templates, which is then decoded into a route graph by iteratively applying each template to the current molecular state.

What would settle it

Take a random sample of top-10 predictions from P2T on PaRoutes test products and count how often the generated template fails to apply to the product with RDChiral or yields reactants that fail standard valence and stereochemistry checks; compare that failure rate with the invalid-SMILES rate of the product-to-reactant baseline. If the two rates are comparable, the claimed chemical-validity advantage of template generation evaporates.

Watch

Extended reading notes

Core claim

The discovery the paper argues for is that retrosynthesis can be reformulated as generative modeling over reaction templates. Formally, the precursor distribution is written as $p(R|o)=\sum_{t} q(R|o,t)\,p_\theta(t|o)$, where $p_\theta(t|o)$ is an autoregressive Transformer that generates the template $t$ token by token from the product $o$, and $q(R|o,t)$ deterministically applies the template to the product to obtain reactants. Because the template library is external to the model, the network has 20--22 million parameters even when trained on 235K templates, whereas a classifier over the same library needs 122 million. The paper reports that template-generating models outperform a product-to-reactant Transformer on PaRoutes single-step reactions (96% vs 92% top-80 accuracy) and, crucially, maintain accuracy on out-of-distribution molecules with molecular weight above 500 g/mol, where the reactant generator falls from 70% to 42% top-5 accuracy. It also shows that a template generator restricted to known templates (P2T-Tok-Strict) is the best policy for search-based multi-step planning on PaRoutes, and that Direct TempRe can generate an entire route as a template sequence, outperforming a nested-JSON direct baseline when trained on the same data.

Load-bearing premise

The results rest on the assumption that the template extraction and application pipeline—atom mapping, RDChiral template extraction, and the data filters—faithfully represents real chemical reactivity; if a class of reactions is poorly mapped or generated templates fail to apply cleanly, both the single-step accuracy and the multi-step route quality numbers would overstate the method's true performance.

Editorial extensions

If this is right

  • Single-step retrosynthesis models can be built as template generators with a compact Transformer, avoiding the parameter explosion of classifying over a 235K-template library (22M vs 122M parameters).
  • Template generation bypasses the length-generalization failure of reactant-SMILES generation, so it should be the preferred sequence-based formulation for large, drug-like target molecules.
  • Filtering generated templates to a known library (strict variants) trades a small loss of single-step diversity for substantially higher multi-step route accuracy, giving users a controllable exploration-fidelity dial.
  • Direct multi-step route generation is feasible as a lightweight single-pass alternative to search, with reasonable solve rates (0.68--0.75 on n1) even before any reward-guided search is added.
  • MCTS with a template-generating policy (P2T-Tok-Strict) sets a new reference point on PaRoutes with top-10 route accuracy of 0.62 on n1 and 0.54 on n5.

Reading between the lines

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

  • Editorial inference: because the output-template length is decoupled from input-molecule size, the out-of-distribution robustness should extend to even larger molecules and to other autoregressive architectures, though the paper only demonstrates it up to the tested molecular-weight bins.
  • Editorial inference: the strict-versus-unrestricted gap suggests a tunable 'creativity dial' for synthesis planning, where generated templates could be scored by similarity to known chemistry rather than simply accepted or discarded; this is a testable extension the paper does not pursue.
  • Editorial inference: the observation that the reactant-generating model's high solve rate is driven by chemically implausible shortcuts implies that solve-rate alone is a misleading progress metric for computer-aided synthesis planning; route-level chemical validity auditing should accompany route-accuracy reporting.
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

3 major / 5 minor

Summary. The paper introduces TempRe, a generative framework for retrosynthesis that reformulates template-based approaches as sequence generation: a Transformer takes a product SMILES as input and autoregressively generates a reaction template (SMARTS), which is then applied with RDChiral to obtain precursors. The authors evaluate TempRe variants (P2T, P2T-Tok, and 'strict' versions that filter to training-set templates) on single-step top-k accuracy using the PaRoutes test sets, a newly constructed 'hard' test set enriched for rare templates, and a molecular-weight-based out-of-distribution split. They also integrate TempRe models as policies in an MCTS planner for multi-step retrosynthesis, and train a 'Direct TempRe' model that generates entire synthetic routes as sequences of templates. The central claims are that TempRe models outperform SMILES-based product-to-reactant (P2R) models across the board, that template generation is more robust to large out-of-distribution molecules, that a strict TempRe policy achieves strong route accuracy on PaRoutes, and that direct template-sequence route generation is a viable lightweight alternative to search-based planning.

Significance. If the empirical claims hold, the paper makes a useful contribution to computer-aided synthesis planning by showing that reaction templates can be treated as a generative modeling target rather than as a fixed classification library. The experimental design is careful in several respects: test reactions are removed from training, a hard set and an OOD split are constructed, a data-filtering ablation is reported, and the Direct TempRe-DMS comparison honestly acknowledges reaction-level leakage in the DMS split. The paper also provides a concrete case study illustrating qualitative differences between template-based and template-free route proposals. However, the headline superiority claims rest on single-seed runs without error bars, and the central notion of 'chemically plausible' retrosynthesis is not directly measured by the reported top-k accuracy, since any RDChiral-applicable template that recovers the ground-truth precursor set is counted as correct. These issues are load-bearing for the paper's main conclusions and require additional analysis before the claims can be fully accepted.

major comments (3)
  1. [Section 3.6 / Appendix A.6.1] The 'chemically plausible' claim in the abstract and conclusion is not supported by the evaluation metric. Top-k accuracy is computed by applying each predicted template with RDChiral and checking whether the resulting precursor set matches the ground truth; a template that is syntactically valid and applicable but chemically nonsensical is counted as correct. Appendix A.6.1 (Figure A.3c) explicitly shows that P2T generates a chemically implausible template for a morpholine ring formation, yet the paper never quantifies how many of the correct top-k predictions rely on such templates, nor how many correct predictions come from novel versus training-set templates. This is especially important because the case study criticizes P2R for proposing chemically questionable 'shortcuts' but does not perform an equivalent audit of TempRe's own top-ranked templates. Please provide a breakdown of correct predictions by template provenance (known vs. novel) and an expert or rule-based plausibility audit on a sample of high-confidence correct predictions, or restrict the chemical-plausibility claim accordingly.
  2. [Section 5.1 / Appendix A.2.1] The main single-step and OOD results are reported without uncertainty estimates. All sequence-to-sequence models are trained once with a fixed random seed (42), and Figures 2a-2c present single runs without error bars or confidence intervals. The headline PaRoutes advantage (96% vs. 92% at Top-80) and the OOD MW-bin differences (57% vs. 42% at the highest MW bin) are therefore not assessable for statistical significance. Given that the paper's central contribution is empirical superiority, please add multiple seeds or bootstrap confidence intervals over the test set, and report the number of test reactions in each MW bin and template-frequency bucket.
  3. [Section 5.3 / Tables 2-3] The claim that the template-representation direct model 'consistently outperforms' DMS's Explorer rests on Direct TempRe-DMS, which is trained on the permissive DMS split that the paper itself identifies as containing reaction-level leakage (Appendix A.8). Although both Direct TempRe-DMS and Explorer use the same split, the comparison only demonstrates an advantage under a leaked training regime; the non-leaky Direct TempRe does not beat Explorer on the n1 solve rate (0.68 vs. 0.74). To support the broader claim that template-based route representation is superior to the nested-JSON representation, please either provide a non-leaky direct comparison or explicitly qualify the advantage as applying only under the DMS data-processing regime.
minor comments (5)
  1. [Section 3.1, Eq. (2)] The symbol T is used both for the generic template library in Eq. (2) and for the training-set template filter used in the strict variants; please use distinct notation or clarify that Eq. (2) applies to strict variants only, since the unrestricted models generate templates outside any fixed library.
  2. [Figure 2] The text in Section 5.1 refers to Figure 2b for template-frequency performance and Figure 2c for OOD performance, but the caption lists subfigures a, b, and c with labels that do not exactly match the in-text ordering; please make the subfigure references consistent.
  3. [Tables 2-3] The row 'MCTS (AZF)(SB) PaRoutes 43K 22M' uses a different training set from all other rows; please separate it visually or add a footnote to make clear that it is an externally trained baseline and not directly comparable to the TempRe-trained models.
  4. [Appendix A.6.1] The statement that 'applying an equivalent filtering process to P2R outputs is less straightforward' would benefit from one or two sentences explaining why RXNMapper-based atom mapping followed by RDChiral template extraction is not a practical filter for P2R outputs in this setting.
  5. [Section 5.2, Figure 3c] The analysis of predicted route lengths reports only average step counts; because the discussion emphasizes that unrestricted models produce shorter routes, please consider also reporting medians or full distributions, which are more robust to the skew discussed in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core results are external-benchmark comparisons, and the few self-citations are tool citations that do not carry the argument.

full rationale

TempRe is an empirical framework paper rather than a derivation. Equation (1) is a standard autoregressive factorization of the template distribution, and Equation (2) defines how template probabilities are marginalized into reactant-set probabilities; neither equation is used to prove a claim that is then fed back as an input. The central single-step and multi-step claims are evaluated on the external PaRoutes benchmark, with baselines trained on the same data, so there is no fitted parameter renamed as a prediction. The only self-citations are operational tool citations (RXNMapper [39] for atom mapping and the Molecular Transformer tokenization [19]); these are published, externally validated tools and are not load-bearing for the conclusions. The direct multi-step model uses an N-step conditioning protocol (Appendix A.8), which is disclosed and does affect route-length statistics; however, the core direct-multi-step comparisons (solve rate, top-k route accuracy) do not reduce to this conditioning input. Appendix A.6.1 also discloses that unrestricted P2T can generate chemically implausible novel templates, which weakens the abstract's 'chemically plausible' wording but is a limitation statement rather than a circular step. Overall, no prediction in the paper reduces by construction to its own inputs.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim depends on empirical training rather than a derivation. The free parameters listed are manual thresholds and hyperparameters that shape the training data and inference, but they do not determine the qualitative conclusion. The axioms reflect reliance on standard cheminformatics tools and public benchmarks.

free parameters (7)
  • template frequency threshold for single-token encoding = 40 occurrences
    Chosen by hand in Appendix A.2.2 to define P2T-Tok tokenizer; affects vocabulary size and performance.
  • MCTS exploration constant C_pucb = 100
    Set in Appendix A.4; controls exploration vs exploitation in search.
  • MCTS policy temperature T = 3.0
    Set in Appendix A.4; flattens policy distribution during search.
  • MCTS expansion size = 10 suggestions
    Chosen in Appendix A.4 and A.7.3; they show 50 decreases AZF performance, so 10 is a manual choice.
  • N-step inference scan range = 2-9 steps
    Chosen in Appendix A.8 for direct multi-step generation; sampling 10 sequences per step count.
  • reaction filtering thresholds = e.g., reactant atoms 10-70, product atoms >=8, etc.
    Hand-selected filters in Appendix A.1.1 to remove noisy reactions; affect the training distribution.
  • beam size for inference = 100 (single-step), 15 (multi-step/direct)
    Set in Appendix A.2.1 to trade off accuracy and speed.
assumptions (5)
  • domain assumption RDChiral correctly extracts and applies reaction templates from atom-mapped reactions.
    All template-based predictions and ground-truth template extraction rely on RDChiral (Section 3.1, Appendix A.1.1).
  • domain assumption RXNMapper provides accurate atom-to-atom mappings.
    Used in data processing to extract templates (Appendix A.1.1).
  • domain assumption PaRoutes ground-truth routes and stock sets are valid benchmarks for retrosynthesis planning.
    All multi-step evaluations are measured against PaRoutes (Section 3.6).
  • domain assumption The USPTO corpus is representative enough for the conclusions.
    All training and test data are derived from USPTO (Section 3.4).
  • domain assumption Applying a template to a product via q(R|o,t) yields chemically valid reactants when the template is in the library.
    Equation 2 in Section 3.1 assumes that the template library limits the search space to valid chemical transformations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TempRe: Template generation for single and direct multi-step retrosynthesis." pith.science (2026). https://pith.science/paper/4U3YUJDC

@misc{pith2026250721762,
  author       = {Pith},
  title        = {Pith review of: TempRe: Template generation for single and direct multi-step retrosynthesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4U3YUJDC}},
  note         = {Machine review of arXiv:2507.21762}
}
read the original abstract

Retrosynthesis planning remains a central challenge in molecular discovery due to the vast and complex chemical reaction space. While traditional template-based methods offer tractability, they suffer from poor scalability and limited generalization, and template-free generative approaches risk generating invalid reactions. In this work, we propose TempRe, a generative framework that reformulates template-based approaches as sequence generation, enabling scalable, flexible, and chemically plausible retrosynthesis. We evaluated TempRe across single-step and multi-step retrosynthesis tasks, demonstrating its superiority over both template classification and SMILES-based generation methods. On the PaRoutes multi-step benchmark, TempRe achieves strong top-k route accuracy. Furthermore, we extend TempRe to direct multi-step synthesis route generation, providing a lightweight and efficient alternative to conventional single-step and search-based approaches. These results highlight the potential of template generative modeling as a powerful paradigm in computer-aided synthesis planning.

Figures

Figures reproduced from arXiv: 2507.21762 by the authors.

Figure 1
Figure 1. Conceptual overview of TempRe. a) Conventional single-step retrosynthesis paradigms based on template classification and template-free reactants generation. b) TempRe models a single￾step transformation as translation-based template generation. c) TempRe as a framework for search￾based and direct multi-step synthesis planning. 2 Related Work Single-Step Retrosynthesis Models. Modern Computer-Assisted Synthesis Plann… view at source ↗
Figure 2
Figure 2. Single-step synthesis planning performance. a) Distribution of training occurrences of reaction templates of the PaRoutes test reactions and the "hard" test reactions, and the corresponding top-k reaction accuracy. The latter contains more rare or unseen reactions during training. b) Single￾step model performance as a function of template training frequency. c) Top-5 accuracy of P2T and P2R in the OOD setting. This … view at source ↗
Figure 3
Figure 3. Multi-step synthesis planning performance on PaRoutes benchmark. a) Solve rate on n1 and n5 sets. b) Top-k route accuracy on n1 and n5 sets. c) Average number of steps in the top-ranked predicted route in comparison. d) Solve rate and top-k route accuracy in correlation with the number of reaction steps of the n5’s ground-truth routes. 5.3 Direct multi-step retrosynthesis In addition to search-based methods to find … view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: (a) Groundtruth synthetic route for dihydrobenzo[a]quinolizine [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 34 canonical work pages

  1. [1]

    Chemistry strategies in early drug discovery: an overview of recent trends.Drug discovery today, 13(15-16):677–684, 2008

    Matteo Colombo and Ilaria Peretto. Chemistry strategies in early drug discovery: an overview of recent trends.Drug discovery today, 13(15-16):677–684, 2008

    work page 2008

  2. [2]

    KC Nicolaou. Organic synthesis: the art and science of replicating the molecules of living nature and creating others like them in the laboratory.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 470(2163):20130690, 2014

    work page 2014

  3. [3]

    Automated synthesis of peptides: Solid-phase peptide synthesis, a simple and rapid synthetic method, has now been automated.Science, 150(3693):178–185, 1965

    RB Merrifield. Automated synthesis of peptides: Solid-phase peptide synthesis, a simple and rapid synthetic method, has now been automated.Science, 150(3693):178–185, 1965

    work page 1965

  4. [4]

    Elias James Corey and W Todd Wipke. Computer-assisted design of complex organic syntheses: Pathways for molecular synthesis can be devised with a computer and equipment for graphical communication.Science, 166(3902):178–192, 1969

    work page 1969

  5. [5]

    Planning chemical syntheses with deep neural networks and symbolic ai.Nature, 555(7698):604–610, 2018

    Marwin HS Segler, Mike Preuss, and Mark P Waller. Planning chemical syntheses with deep neural networks and symbolic ai.Nature, 555(7698):604–610, 2018

    2018

  6. [6]

    Computer-assisted retrosynthesis based on molecular similarity.ACS Cent

    Connor W Coley, Luke Rogers, William H Green, and Klavs F Jensen. Computer-assisted retrosynthesis based on molecular similarity.ACS Cent. Sci., 3:1237–1245, 2017

    work page 2017

  7. [7]

    Concerning one system of classification and codification of organic reactions

    GE Vleduts. Concerning one system of classification and codification of organic reactions. Information Storage and Retrieval, 1(2-3):117–146, 1963

    work page 1963

  8. [8]

    Predicting retrosyn- thetic pathways using transformer-based models and a hyper-graph exploration strategy.Chem

    Philippe Schwaller, Riccardo Petraglia, Valerio Zullo, Vishnu H Nair, Rico Andreas Haeusel- mann, Riccardo Pisoni, Costas Bekas, Anna Iuliano, and Teodoro Laino. Predicting retrosyn- thetic pathways using transformer-based models and a hyper-graph exploration strategy.Chem. Sci., 11:3316–3325, 2020

    work page 2020

Show all 43 references
  1. [9]

    SMILES, a chemical language and information system

    David Weininger. SMILES, a chemical language and information system. 1. introduction to methodology and encoding rules.J. Chem. Inf. Comput. Sci., 28:31–36, 1988

    1988

  2. [10]

    Daylight Chemical Information Systems

    Inc. Daylight Chemical Information Systems. Smarts-a language for describing molecular patterns. 2019

    2019

  3. [11]

    The chemical space project.Accounts of chemical research, 48(3): 722–730, 2015

    Jean-Louis Reymond. The chemical space project.Accounts of chemical research, 48(3): 722–730, 2015

    2015

  4. [12]

    Computer-assisted synthetic planning: the end of the beginning.Angewandte Chemie International Edition, 55(20):5904–5937, 2016

    Sara Szymku ´c, Ewa P Gajewska, Tomasz Klucznik, Karol Molga, Piotr Dittwald, Michał Startek, Michał Bajczyk, and Bartosz A Grzybowski. Computer-assisted synthetic planning: the end of the beginning.Angewandte Chemie International Edition, 55(20):5904–5937, 2016

    2016

  5. [13]

    Automatic retrosynthetic route planning using template-free models.Chemical science, 11(12):3355–3364, 2020

    Kangjie Lin, Youjun Xu, Jianfeng Pei, and Luhua Lai. Automatic retrosynthetic route planning using template-free models.Chemical science, 11(12):3355–3364, 2020

    2020

  6. [14]

    Retrosynthetic re- action prediction using neural sequence-to-sequence models.ACS central science, 3(10): 1103–1113, 2017

    Bowen Liu, Bharath Ramsundar, Prasad Kawthekar, Jade Shi, Joseph Gomes, Quang Luu Nguyen, Stephen Ho, Jack Sloane, Paul Wender, and Vijay Pande. Retrosynthetic re- action prediction using neural sequence-to-sequence models.ACS central science, 3(10): 1103–1113, 2017

    2017

  7. [15]

    Root-aligned smiles: a tight representation for chemical reaction prediction.Chemical Science, 13(31):9023–9034, 2022

    Zipeng Zhong, Jie Song, Zunlei Feng, Tiantao Liu, Lingxiang Jia, Shaolun Yao, Min Wu, Tingjun Hou, and Mingli Song. Root-aligned smiles: a tight representation for chemical reaction prediction.Chemical Science, 13(31):9023–9034, 2022

    2022

  8. [16]

    Permutation invariant graph-to-sequence model for template- free retrosynthesis and reaction prediction.Journal of chemical information and modeling, 62 (15):3503–3513, 2022

    Zhengkai Tu and Connor W Coley. Permutation invariant graph-to-sequence model for template- free retrosynthesis and reaction prediction.Journal of chemical information and modeling, 62 (15):3503–3513, 2022

    2022

  9. [17]

    Neural template extraction-learning the language of smirks

    Amol Thakkar and Teodoro Laino. Neural template extraction-learning the language of smirks. 2024

    2024

  10. [18]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017. 14

    2017

  11. [19]

    Molecular transformer: a model for uncertainty-calibrated chemical reaction prediction.ACS central science, 5(9):1572–1583, 2019

    Philippe Schwaller, Teodoro Laino, Théophile Gaudin, Peter Bolgar, Christopher A Hunter, Costas Bekas, and Alpha A Lee. Molecular transformer: a model for uncertainty-calibrated chemical reaction prediction.ACS central science, 5(9):1572–1583, 2019

    2019

  12. [20]

    Paroutes: towards a framework for benchmarking retrosynthesis route predictions.Digital Discovery, 1(4):527–539, 2022

    Samuel Genheden and Esben Bjerrum. Paroutes: towards a framework for benchmarking retrosynthesis route predictions.Digital Discovery, 1(4):527–539, 2022

    2022

  13. [21]

    Aizynthfinder: a fast, robust and flexible open-source software for retrosyn- thetic planning.Journal of cheminformatics, 12(1):70, 2020

    Samuel Genheden, Amol Thakkar, Veronika Chadimová, Jean-Louis Reymond, Ola Engkvist, and Esben Bjerrum. Aizynthfinder: a fast, robust and flexible open-source software for retrosyn- thetic planning.Journal of cheminformatics, 12(1):70, 2020

    2020

  14. [22]

    Retro*: learning retrosynthetic planning with neural guided a* search

    Binghong Chen, Chengtao Li, Hanjun Dai, and Le Song. Retro*: learning retrosynthetic planning with neural guided a* search. InInternational conference on machine learning, pages 1608–1616. PMLR, 2020

    2020

  15. [23]

    Models matter: The impact of single-step retrosynthesis on synthesis planning.Digital Discovery, 3(3):558–572, 2024

    Paula Torren-Peraire, Alan Kai Hassen, Samuel Genheden, Jonas Verhoeven, Djork-Arné Clevert, Mike Preuss, and Igor V Tetko. Models matter: The impact of single-step retrosynthesis on synthesis planning.Digital Discovery, 3(3):558–572, 2024

    2024

  16. [24]

    Junren Li, Kangjie Lin, Jianfeng Pei, and Luhua Lai. Challenging complexity with simplicity: Rethinking the role of single-step models in computer-aided synthesis planning.Journal of Chemical Information and Modeling, 64(14):5470–5479, 2024

    2024

  17. [25]

    Re-evaluating retrosynthesis algorithms with syntheseus

    Krzysztof Maziarz, Austin Tripp, Guoqing Liu, Megan Stanley, Shufang Xie, Piotr Gai ´nski, Philipp Seidl, and Marwin HS Segler. Re-evaluating retrosynthesis algorithms with syntheseus. Faraday Discussions, 256:568–586, 2025

    2025

  18. [26]

    Retrocomposer: composing templates for template-based retrosynthesis prediction.Biomolecules, 12(9):1325, 2022

    Chaochao Yan, Peilin Zhao, Chan Lu, Yang Yu, and Junzhou Huang. Retrocomposer: composing templates for template-based retrosynthesis prediction.Biomolecules, 12(9):1325, 2022

    2022

  19. [27]

    Site-specific template generative approach for retrosynthetic planning.Nature Communications, 15(1):7818, 2024

    Yu Shee, Haote Li, Pengpeng Zhang, Andrea M Nikolic, Wenxin Lu, H Ray Kelly, Vidhyadhar Manee, Sanil Sreekumar, Frederic G Buono, Jinhua J Song, et al. Site-specific template generative approach for retrosynthetic planning.Nature Communications, 15(1):7818, 2024

    2024

  20. [28]

    Retrosynformer: Planning multi-step chemical synthesis routes via a decision transformer

    Emma Granqvist, Rocío Mercado, and Samuel Genheden. Retrosynformer: Planning multi-step chemical synthesis routes via a decision transformer. 2025

    2025

  21. [29]

    Directmultistep: Direct route generation for multistep retrosynthesis.Journal of Chemical Information and Modeling, 65(8): 3903–3914, 2025

    Yu Shee, Anton Morgunov, Haote Li, and Victor S Batista. Directmultistep: Direct route generation for multistep retrosynthesis.Journal of Chemical Information and Modeling, 65(8): 3903–3914, 2025

    2025

  22. [30]

    PhD thesis, 2012

    Daniel Mark Lowe.Extraction of chemical structures and reactions from the literature. PhD thesis, 2012

    2012

  23. [31]

    Machine intelligence for chemical reaction space

    Philippe Schwaller, Alain C Vaucher, Ruben Laplaza, Charlotte Bunne, Andreas Krause, Clemence Corminboeuf, and Teodoro Laino. Machine intelligence for chemical reaction space. Wiley Interdisciplinary Reviews: Computational Molecular Science, 12(5):e1604, 2022

    2022

  24. [32]

    Rdchiral: An rdkit wrapper for handling stereochemistry in retrosynthetic template extraction and application.Journal of chemical information and modeling, 59(6):2529–2537, 2019

    Connor W Coley, William H Green, and Klavs F Jensen. Rdchiral: An rdkit wrapper for handling stereochemistry in retrosynthetic template extraction and application.Journal of chemical information and modeling, 59(6):2529–2537, 2019

    2019

  25. [33]

    Higher-level strategies for computer-aided retrosynthesis

    Jihye Roh, Joonyoung F Joung, Kevin Yu, Zhengkai Tu, G Logan Bartholomew, Omar A Santiago-Reyes, Mun Hong Fong, Richmond Sarpong, Sarah E Reisman, and Connor W Coley. Higher-level strategies for computer-aided retrosynthesis. 2025

    2025

  26. [34]

    Fast prediction of distances between synthetic routes with deep learning.Machine Learning: Science and Technology, 3(1):015018, 2022

    Samuel Genheden, Ola Engkvist, and Esben Bjerrum. Fast prediction of distances between synthetic routes with deep learning.Machine Learning: Science and Technology, 3(1):015018, 2022

    2022

  27. [35]

    Clustering of synthetic routes using tree edit distance.Journal of Chemical Information and Modeling, 61(8):3899–3907, 2021

    Samuel Genheden, Ola Engkvist, and Esben Bjerrum. Clustering of synthetic routes using tree edit distance.Journal of Chemical Information and Modeling, 61(8):3899–3907, 2021. 15

    2021

  28. [36]

    Selection of cost-effective yet chemically diverse pathways from the networks of computer-generated retrosynthetic plans

    Tomasz Badowski, Karol Molga, and Bartosz A Grzybowski. Selection of cost-effective yet chemically diverse pathways from the networks of computer-generated retrosynthetic plans. Chemical science, 10(17):4640–4651, 2019

    2019

  29. [37]

    Bran, Malte Franke, Remi Schlama, Jeremy S

    Victor Sabanza Gil, Andres M. Bran, Malte Franke, Remi Schlama, Jeremy S. Luterbacher, and Philippe Schwaller. Holistic chemical evaluation reveals pitfalls in reaction prediction models, 2023

    2023

  30. [38]

    Assessing the extrapolation capability of template-free ret- rosynthesis models.arXiv preprint arXiv:2403.03960, 2024

    Shuan Chen and Yousung Jung. Assessing the extrapolation capability of template-free ret- rosynthesis models.arXiv preprint arXiv:2403.03960, 2024

    2024 arXiv

  31. [39]

    Extraction of organic chemistry grammar from unsupervised learning of chemical reactions.Science Advances, 7(15):eabe4166, 2021

    Philippe Schwaller, Benjamin Hoover, Jean-Louis Reymond, Hendrik Strobelt, and Teodoro Laino. Extraction of organic chemistry grammar from unsupervised learning of chemical reactions.Science Advances, 7(15):eabe4166, 2021

    2021

  32. [40]

    Opennmt: Open-source toolkit for neural machine translation.arXiv preprint arXiv:1701.02810, 2017

    Guillaume Klein, Yoon Kim, Yuntian Deng, Jean Senellart, and Alexander M Rush. Opennmt: Open-source toolkit for neural machine translation.arXiv preprint arXiv:1701.02810, 2017

    2017 arXiv

  33. [41]

    Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    Diederik P Kingma. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  34. [42]

    Understanding the difficulty of training deep feedfor- ward neural networks

    Xavier Glorot and Yoshua Bengio. Understanding the difficulty of training deep feedfor- ward neural networks. InProceedings of the thirteenth international conference on artificial intelligence and statistics, pages 249–256. JMLR Workshop and Conference Proceedings, 2010

    2010

  35. [43]

    hard test set

    Harry L Morgan. The generation of a unique machine description for chemical structures-a technique developed at chemical abstracts service.Journal of chemical documentation, 5(2): 107–113, 1965. 16 A Appendix A.1 Experimental Setup and Implementation Details A.1.1 General Data...

    1965

Pith tools

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