Learning Chemical Reaction Representation with Reactant-Product Alignment

Kaipeng Zeng; Xianbin Liu; Xiaokang Yang; Yanyan Xu; Yaohui Jin; Yu Zhang

REVIEW 2 major objections 8 minor 64 references

Learning Chemical Reaction Representation with Reactant-Product Alignment

T0 review · 2 major / 8 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read RAlign learns a chemical reaction representation by aligning each product atom with its reactant counterpart, then reports accuracy gains over existing reaction encoders on condition, yield, and selectivity tasks.

desk verdict Solid applied paper with a real architectural idea; the headline gain is plausible but the reaction-center mask is an extra input baselines lack, and the missing ablation on the main task is the one thing a referee should push on. read the letter →

arxiv 2411.17629 v2 pith:4C2QRWCJ submitted 2024-11-26 cs.LG cs.AI

Kaipeng Zeng , Xianbin Liu , Yu Zhang , Xiaokang Yang , Yaohui Jin , Yanyan Xu This is my paper · ORCID

classification cs.LGcs.AI

keywords chemicalreactionrepresentationlearningatommappingreactant-productalignmentcenterconditionpredictionyieldselectivitygraphneuralnetwork

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

RAlign is a graph-based encoder-decoder for chemical reactions that uses atom-level correspondence between reactants and products as its core inductive bias, rather than treating a reaction as a bag of molecules or a SMILES string. The paper argues that fusing features of corresponding atoms lets the model expose which bonds break and form, and that a reaction-center-aware decoder focuses attention on the atoms that actually change. On the USPTO CONDITION benchmark, RAlign reaches 34.30% top-1 condition-combination accuracy versus 27.42% for the strongest baseline, and it posts the best or near-best results on most yield and selectivity splits without large-scale reaction pretraining. The intended payoff is a reusable reaction representation backbone that can absorb reaction conditions in different formats and transfer across synthesis-related tasks.

What carries the argument

The Atom Aligned Encoder is the load-bearing component: it runs two message-passing networks for reactants and products and, after each layer, concatenates the intermediate features of every atom pair matched by atom mapping and passes them through a feedforward network, so each side's representation is informed by its counterpart's local context. A condition adapter uses multi-head attention to inject encoded reaction conditions into node features at every layer, making the encoder modular across datasets whose conditions are reagents, temperatures, or experimental protocols. The RC-aware decoder splits its cross-attention heads: half attend normally and half attend only to reaction-center atoms, defined as atoms whose bonds change or hydrogen counts change, their one-hop neighbors, and leaving-group atoms. The reaction-center heads give the model a direct focus on the functional groups that determine the reaction outcome.

What would settle it

Perturb the atom mappings on the USPTO CONDITION test set by randomly swapping which product atoms are paired with reactant atoms, or replace the mapping tool with another one, and retrain; if top-1 condition accuracy stays roughly the same, the alignment and reaction-center inputs are not what drives the reported gains.

Watch

Extended reading notes

Core claim

The paper's central claim is that a chemical reaction representation is substantially improved when the encoder is built around the known atomic correspondence between reactants and products. RAlign maps each product atom to its reactant counterpart, runs separate message-passing layers on the two molecular graphs, fuses the paired atom features at every layer, and keeps leaving-group atoms as special nodes. The decoder then mixes standard cross-attention heads with heads that can only attend to reaction centers, defined as atoms whose bonds or hydrogen counts change plus their one-hop neighbors and leaving groups. On the USPTO CONDITION task this yields 34.30% top-1 accuracy, an absolute gain of 6.88 points over the Parrot-LM E baseline, with similar gains in components such as solvents; on Buchwald-Hartwig yield and C-H functionalization selectivity, RAlign achieves the best R2 among compared methods on most splits. The authors interpret these results as evidence that modeling bond changes and focusing on reaction centers are stronger inductive biases than fingerprints, pretrained SMILES encoders, or simple aggregation of per-molecule features.

Load-bearing premise

The model is given reaction centers computed from atom mappings at every inference, so its reported advantage assumes reliable atom mapping is available for every reaction and that supplying this reaction-center information is a fair, transferable input.

Editorial extensions

If this is right

Reaction condition combination prediction on USPTO CONDITION improves to 34.30% top-1 overall accuracy, with per-component gains in catalyst, solvent, and reagent prediction; the same architecture also improves de novo condition generation on USPTO 500MT over non-pretrained baselines.
On Buchwald-Hartwig yield prediction, RAlign achieves the best R2 on all four out-of-sample ligand splits and competitive random-split results without large-scale reaction pretraining.
On C-H functionalization selectivity, RAlign attains the best MAE, RMSE, and R2 among compared methods; on the small thiol addition selectivity dataset it outperforms other deep learning models but trails hand-crafted fingerprints.
Ablations removing the atom alignment fusion or replacing RC-aware attention with vanilla cross-attention reduce performance on both tested tasks, indicating that both components contribute to the reported gains.
The condition adapter allows pretrained molecular condition encoders to be plugged in, so the architecture can use existing molecular models rather than requiring reaction-specific pretraining.

Reading between the lines

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

An unstated but direct consequence is that atom alignment could become a standard preprocessing step for graph-based reaction encoders, replacing implicit feature aggregation with explicit correspondence between the two molecular sides.
Because the reaction centers are computed from atom mapping, a fair comparison with baselines would require giving those baselines the same reaction-center information; without that ablation, part of the reported gap may reflect an extra input rather than a purely better learned representation.
Since the encoder is agnostic to how node and edge features are computed, it could be combined with 3D conformer or quantum-chemical descriptors for selectivity tasks where current small-data performance lags fingerprint methods.
The condition adapter points toward a natural extension: encoding experimental procedures written in natural language and injecting them through the same attention-based adapter, which would broaden the model beyond reagent and temperature inputs.

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.

Desk Editor's Note

read the letter

RAlign is a serious attempt at reaction representation learning that actually uses the structure of a reaction. Atom-mapped reactant/product pairs are fused in every encoder layer, and half the attention heads are restricted to reaction centers. That combination is new, and the evaluation is broad: condition prediction and generation, yield, and selectivity. On C-H functionalization and the out-of-sample Buchwald-Hartwig splits, RAlign is consistently strong and beats pretrained baselines like T5Chem and RXNFP. The authors disclose the main dependence (atom mapping) and include a robustness check against RXNMapper confidence. This is not hype.

The soft spots are real but not fatal. The biggest one is the reaction-center input. In the decoder, Vrc—the set of atoms that change bonds or hydrogen counts, plus leaving groups—is computed from atom mapping and supplied to the model at inference. No baseline gets this. On USPTO CONDITION, the headline task, this is an extra input that correlates with reaction type, so the 34.3% vs 27.4% top-1 gain is not a clean apples-to-apples comparison. The ablation of the RC-aware decoder is only run on USPTO 500MT and Buchwald-Hartwig, where the effect is small (roughly 0.3 top-1 and 0.003 R2). That suggests the reaction-center mask is not the whole story, but the task where the gain is largest is exactly where the control is missing. The robustness analysis correlates accuracy with mapping confidence; it does not perturb or remove Vrc. A simple control on USPTO CONDITION without the RC-aware heads would settle this.

Other issues are smaller. Tables 1–3 have no variance or significance estimates, and there are random splits available. On the Buchwald-Hartwig random splits RAlign is third (MAE 3.63 vs T5Chem 3.51), and on Test4 its R2 (0.532) is below T5Chem (0.605). The abstract says 'markedly outperforms,' which is stronger than the full set of tables. The citation of prior work is fair, though the introduction's 'first graph-based architecture tailored for chemical reactions' is overstated given Retroformer's local-global decoder.

The math is not circular: atom mapping comes from input molecules, not from labels, so the gains are not label leakage. The architecture is specified in enough detail to reproduce from the appendix, but code is not released. This is an important detail for the field.

Verdict: deserves referee time. I would send it to review, with a major-revision request for the missing ablation on the headline task, uncertainty estimates, and code or a strong reason for its absence. If the RC-aware decoder accounts for most of the USPTO CONDITION gain, the paper still has value—it becomes an input-augmentation method, and the claim needs to be scaled back accordingly.

Referee Report

2 major / 8 minor

Summary. The paper introduces RAlign, a supervised graph-based model for learning chemical reaction representations from molecular graphs together with atom-mapping-derived atomic correspondences between reactants and products. The architecture has three components: an atom-aligned encoder (Sec. 4.1) that fuses reactant and product node features for corresponding atoms through an information fusion layer; a condition-incorporating adapter (Sec. 4.2) that lets reaction conditions enter the encoder via multi-head attention; and reaction-center-aware (RC-aware) decoders (Sec. 4.3) in which half of the cross-attention heads attend only to reaction-center atoms Vrc, defined in Sec. 3.1 from bond changes, hydrogen-count changes, one-hop neighbors, and leaving groups. RAlign is evaluated on reaction condition prediction (USPTO CONDITION, Tables 1-2), condition generation (USPTO 500MT, Table 3), yield prediction (Buchwald-Hartwig, Tables 4-5), and selectivity prediction (C-H functionalization and thiol addition, Tables 6-7). Headline results include top-1 condition-prediction accuracy of 34.30% versus 27.42% for the strongest baseline Parrot-LM E, and best or near-best results on most other tasks, with third-place finishes on the Buchwald-Hartwig random splits and on thiol addition.

Significance. This paper contributes a clean, well-motivated architectural inductive bias: explicit reactant-product atomic alignment (Eq. 3) and reaction-center-constrained attention (Eq. 6), packaged as a modular encoder-decoder with an interchangeable condition adapter. The empirical validation is broad, spanning four downstream tasks including out-of-sample splits on Buchwald-Hartwig, and the authors disclose the regimes where the method does not win (third on Buchwald-Hartwig random splits and on thiol addition). Credit should also go to the ablations in Table 8, the robustness analysis in Appendix A.3, the detailed implementation appendices, and the plain-spoken limitation statement about atom-mapping dependence. If the reported numbers hold, RAlign is a competitive, pretraining-free alternative to large pretrained reaction models on several tasks. However, the two reservations below, namely the unablated reaction-center input on the headline condition-prediction task and the absence of variance or significance statistics in Tables 1-3, mean that the strength of the central claim is not yet fully established.

major comments (2)

[Sec. 4.3, Tables 1-2, Table 8, Appendix A.3] The fairness of the headline condition-prediction comparison is not established because the RC-aware decoder receives the ground-truth reaction-center atom set Vrc at inference (computed from atom mapping, Sec. 3.1), an input that none of the baselines (Parrot-LM E, T5Chem, GCNN, FPRCR) receive. This is not label leakage, but it is an extra, task-correlated input channel: the case study in Sec. 5.5 states that the RC-aware heads 'provid[e] the model with information about the reaction type,' and reaction type correlates with catalyst, solvent, and reagent choices in exactly the USPTO CONDITION task. The ablation in Table 8 removes the RC-aware decoder only on USPTO 500MT and Buchwald-Hartwig, not on USPTO CONDITION, so the 6.88-point top-1 margin over Parrot-LM E has no control. The robustness analysis in Appendix A.3 correlates accuracy with RXNMapper confidence but never removes or perturbs Vrc. Please (a) report the '- Reaction-Center-Aware Decoders' variant on USPTO CONDITION; (b) add a control where Vrc is replaced by a same-size random atom set, to separate the effect of the center information from decoder capacity; and (c) add a mapping-perturbation study (e.g., shuffled atom alignments) to Appendix A.3. Until these are provided, part of the reported gain may be attributable to the privileged center input rather than to the learned representation.
[Tables 1-3 and 5-7] The statistical support for the 'markedly outperforms' claim is incomplete. Tables 1-3 report single-run accuracies with no variance or significance information, so the 6.88-point and 1.55-point top-1 gaps on USPTO CONDITION and USPTO 500MT cannot be assessed for stability. In contrast, Tables 5-7 report mean plus or minus standard deviation over ten splits, and on the Buchwald-Hartwig random splits RAlign is third (MAE 3.6331 +/- 0.1259 versus T5Chem 3.5059 +/- 0.1562; R2 0.9581 +/- 0.0049 versus 0.9662 +/- 0.0034). The text's assertion that the model is 'not significantly outperformed' by the pretrained baselines is based on overlapping standard deviations rather than a paired test, and the claim that the MAE gap is 'less than 0.1' (Sec. 5.2) is inaccurate for the strongest baseline, since the gap to T5Chem is 0.127. Please report multiple seeds with standard deviation or a significance test for the headline Tables 1-3, and run a paired test or effect-size analysis for the Buchwald-Hartwig random-split and thiol-addition comparisons.

minor comments (8)

[Sec. 4.1, Eq. (3)] The left-hand side of the reactant MPNN equation lists the intermediate node features as {hR(k)_i, hR(k)_i, ..., hR(k)_i} with a repeated index; this should range over vR_1 through vR_n as in Eq. (2).
[Sec. 4.3, Eq. (6)] The square-root-of-d normalization appears to be applied outside already-normalized softmax weights alpha1_l and alpha2_l; if scaled dot-product attention is intended, the factor should be inside the exponent. Please correct the formula for reproducibility.
[Sec. 5.2] The statement that the MAE gap to the strongest baseline is 'less than 0.1' holds only versus YieldBert; the gap to T5Chem is 3.6331 minus 3.5059, which is 0.1272, so the sentence should be corrected.
[Sec. 5.1] The margins '9.03% and 11.93%' reported against 'the strongest baseline that did not utilize pretraining' do not tie out to a single model in Table 3: 9.03 is the top-1 margin over Reagent Transformer, which is itself initialized from a pretrained Molecular Transformer checkpoint, while the top-10 margin over T5Chem-from-scratch is 11.97. Please clarify which baseline is intended and reconcile the numbers.
[Sec. 1 and Sec. 2] The contribution bullet claiming 'the first graph-based architecture specifically tailored for chemical reactions' and the first modeling of atomic correspondence overstates novelty, given GCNN (Maser et al., 2021), the condensed-graph-of-reaction encoding used by Chemprop, and prior atom-mapping-aware architectures such as Retroformer (Wan et al., 2022). Please soften these claims and position the contribution as a general reaction-representation backbone.
[Sec. 5.1 and Appendix B] For USPTO CONDITION, please state explicitly whether the train, validation, and test split is identical to the one used by Wang et al. (2023) for Parrot-LM E, since Table 2 compares directly against the reported Parrot-LM E numbers; the RXNMapper re-annotation should be described as performed within each split to avoid any ambiguity.
[Appendix A.3] For USPTO 500MT the Pearson coefficients for top-5 and top-10 accuracy are 0.367 and 0.415, which are moderate rather than negligible, so the conclusion that 'there is no strong correlation' should be tempered or the analysis extended with a perturbation test.
[Sec. 5.1 and Appendix D.2] Because the USPTO 500MT generation labels are restructured with the Type I/II/III ordering and GCNN is reformulated as multilabel classification over a predefined library, the USPTO 500MT comparison mixes task definitions across baselines; please state explicitly that the structured output ordering is an implementation choice and discuss whether it favors RAlign relative to the SMILES-generation baselines.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: RAlign is an end-to-end supervised representation model; reaction-center input is derived from atom mapping, not from the predicted labels, and self-citations are not load-bearing.

full rationale

RAlign's central claims are empirical: a reactant-product aligned encoder plus a reaction-center-aware decoder is trained end-to-end on held-out labels (reaction conditions, yields, selectivity) and evaluated on test splits. In Section 3.1, the reaction-center set Vrc is defined from the atom correspondence between reactants and products (bond changes, hydrogen-count changes, one-hop neighbors, leaving group), and in Section 4.3 this set is used as a mask in RC-aware cross-attention; Vrc is therefore a function of the input molecular graphs and atom mapping, not of the target labels. Equation 3 and Equation 6 define feature updates and masked attention; no output is reinserted as an input and no fitted parameter is renamed as a prediction. The reported gains may partly reflect that baselines do not receive the reaction-center mask (a fairness/input-comparability caveat acknowledged in the Limitations: 'The model requires atom mappings as input'), and Appendix A.3 checks only correlation with RXNMapper confidence rather than removing Vrc; but this is not circularity under the definition used here, because Vrc is a legitimate input feature rather than the target or a parameter fit to the target. The self-citations (Zeng et al., 2024; Yang et al., 2022) occur in the related-work enumeration of GNN molecule encoders and are not used to justify the architecture's effectiveness; no uniqueness theorem or ansatz is imported from the authors' prior work. USPTO 500MT label re-annotation (Appendix D.2) is a preprocessing choice, not a construction that makes test predictions equal to training labels. Consequently, there is no step where a 'prediction' is equivalent to its input by construction; the score reflects only one minor, non-load-bearing self-citation.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim does not rest on derived constants or new physical entities. It rests on standard chemistry assumptions about atom conservation, on the quality of external atom-mapping tools, and on a hand-picked reaction-center definition. The learned network weights and hyperparameters are the main fitted quantities, and they are trained on the same benchmark labels used for evaluation.

free parameters (2)

Task-specific hyperparameters (hidden size, encoder/decoder layers, attention heads, dropout, learning rate) = hidden 512 or 128; layers 3, 5, or 6; heads 8; dropout 0.0 or 0.1; LR 1e-4, 5e-4, 5e-5, or 1.25e-4
Chosen by hand per dataset in Appendix C.5 with no sensitivity analysis; reported results could depend on these choices.
Reaction-center attention head fraction = half of the 8 heads
The decoder constrains 4 of 8 heads to attend only to reaction centers (Section 4.3); the paper does not ablate this fraction.

assumptions (3)

domain assumption Atoms are conserved in reactions and every product atom has a unique reactant counterpart (Section 3.1).
This justifies the bijection used by the atom-aligned encoder; it fails for isotope labeling or reactions with ambiguous atom provenance.
domain assumption RXNMapper (or template-based mapping) provides sufficiently accurate atom mapping for all datasets (Sections 5.1, D.1).
The full architecture depends on the mapping as input; the paper's robustness check only correlates performance with mapper confidence.
ad hoc to paper The reaction-center definition (bond-change termini, hydrogen-count changes, one-hop neighbors, leaving groups) captures the functional groups relevant to each downstream task (Section 3.1).
This definition is adapted from Retroformer and is not validated against alternative reaction-center definitions.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Learning Chemical Reaction Representation with Reactant-Product Alignment." pith.science (2026). https://pith.science/paper/4C2QRWCJ

@misc{pith2026241117629,
  author       = {Pith},
  title        = {Pith review of: Learning Chemical Reaction Representation with Reactant-Product Alignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4C2QRWCJ}},
  note         = {Machine review of arXiv:2411.17629}
}

read the original abstract

Organic synthesis stands as a cornerstone of the chemical industry. The development of robust machine learning models to support tasks associated with organic reactions is of significant interest. However, current methods rely on hand-crafted features or direct adaptations of model architectures from other domains, which lack feasibility as data scales increase or ignore the rich chemical information inherent in reactions. To address these issues, this paper introduces RAlign, a novel chemical reaction representation learning model for various organic reaction-related tasks. By integrating atomic correspondence between reactants and products, our model discerns the molecular transformations that occur during the reaction, thereby enhancing comprehension of the reaction mechanism. We have designed an adapter structure to incorporate reaction conditions into the chemical reaction representation, allowing the model to handle various reaction conditions and to adapt to various datasets and downstream tasks. Additionally, we introduce a reaction-center-aware attention mechanism that enables the model to concentrate on key functional groups, thereby generating potent representations for chemical reactions. Our model has been evaluated on a range of downstream tasks. Experimental results indicate that our model markedly outperforms existing chemical reaction representation learning architectures on most of the datasets. We plan to open-source the code contingent upon the acceptance of the paper.

Figures

Figures reproduced from arXiv: 2411.17629 by the authors.

**Figure 1.** Overview of RAlign. For a given chemical reaction, the molecule graphs of reactants and products, along with their atomic correspondence, are input into a L-layer Atom Aligned Encoder to extract reaction node features HP (L) and HR(L) . If the reaction conditions are also provided, they are encoded by a condition encoder and merged into HP (L) and HR(L) via an adapter structure proposed in this study. The resulting … view at source ↗

**Figure 2.** Different visualizations of node embeddings at each encoder layer, comparing the aligned encoder with the non-aligned encoder. and to comprehend the reaction process, thereby leading to performance improvements. 5.5. Case Study To intuitively demonstrate what our proposed module has learned, we conducted visualizations of both the embeddings learned by the atom-aligned encoder and the attention coefficients from t… view at source ↗

**Figure 3.** Visualization of RC-aware cross attention weights of each node in reactants and products for catalyst prediction such as chemical reaction condition prediction. These tasks are intrinsically linked to the design of molecular synthesis schemes and the optimization of production processes in the organic chemical industry. Our work has the potential to increase the production efficiency of the organic synthesis industr… view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: The visualization the attention weights of each node in reactants and products while predicting the reaction condition combination for the reduction of nitro groups. A.2. Embedding Visualization of Atom Aligned Encoder We have completed the reaction information for the…

**Figure 5.** Figure 5: The visualization of the RC-aware cross attention coefficients for the reaction condition combination prediction of a amide coupling reaction. 16 [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]

**Figure 6.** Figure 6: The visualization of the RC-aware cross attention coefficients for the reaction condition combination prediction of a olefin metathesis reaction. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

**Figure 7.** Figure 7: The visualizations of node embeddings at each encoder layer for different reactions, comparing the atom-aligned encoder with the conventional MPNN encoder. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]

**Figure 8.** Figure 8: The relation between top-k accuracy and the atom mapping confidence score on the USPTO CONDITION dataset and the USPTO 500MT dataset. evaluate the performance. A prediction is considered correct if and only if all of the molecules of it are correctly predicted. The ord…

**Figure 9.** Figure 9: The pipelines for different tasks This work does not train a multi-task unified model like T5chem (Lu & Zhang, 2022). Given the different datasets and varying inputs and outputs for each task, we have selected different modules to assemble the pipelines for the tasks d…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 46 canonical work pages

[1]

T., Estrada, J

Ahneman, D. T., Estrada, J. G., Lin, S., Dreher, S. D., and Doyle, A. G. Predicting reaction performance in c--n cross-coupling using machine learning. Science, 360 0 (6385): 0 186--190, 2018

work page 2018
[2]

Reagent prediction with a molecular transformer improves reaction data quality

Andronov, M., Voinarovska, V., Andronova, N., Wand, M., Clevert, D.-A., and Schmidhuber, J. Reagent prediction with a molecular transformer improves reaction data quality. Chemical Science, 14 0 (12): 0 3235--3246, 2023

work page 2023
[3]

L., Kiros, J

Ba, J. L., Kiros, J. R., and Hinton, G. E. Layer normalization. arXiv preprint arXiv:1607.06450, 2016

arXiv 2016
[4]

Uncertainty-aware yield prediction with multimodal molecular features

Chen, J., Guo, K., Liu, Z., Isayev, O., and Zhang, X. Uncertainty-aware yield prediction with multimodal molecular features. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pp.\ 8274--8282, 2024 a

work page 2024
[5]

Precise atom-to-atom mapping for organic reactions via human-in-the-loop machine learning

Chen, S., An, S., Babazade, R., and Jung, Y. Precise atom-to-atom mapping for organic reactions via human-in-the-loop machine learning. Nature Communications, 15 0 (1): 0 2250, 2024 b

work page 2024
[6]

and Guestrin, C

Chen, T. and Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, pp.\ 785--794, 2016

2016
[7]

S., Babu, C

Dara, S., Dhamercherla, S., Jadav, S. S., Babu, C. M., and Ahsan, M. J. Machine learning in drug discovery: a review. Artificial intelligence review, 55 0 (3): 0 1947--1999, 2022

work page 1947
[8]

BERT: pre-training of deep bidirectional transformers for language understanding

Devlin, J., Chang, M., Lee, K., and Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. In Burstein, J., Doran, C., and Solorio, T. (eds.), Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, ...

doi:10.18653/v1/n 2019

Show all 64 references

[9]

and Lenssen, J

Fey, M. and Lenssen, J. E. Fast graph representation learning with PyTorch Geometric . In ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019

2019
[10]

Description of organic reactions based on imaginary transition structures

Fujita, S. Description of organic reactions based on imaginary transition structures. 1. introduction of new concepts. Journal of Chemical Information and Computer Sciences, 26 0 (4): 0 205--212, 1986

1986
[11]

J., Coley, C

Gao, H., Struble, T. J., Coley, C. W., Wang, Y., Green, W. H., and Jensen, K. F. Using machine learning to predict suitable conditions for organic reactions. ACS Central Science, 4: 0 1465--1476, 11 2018. ISSN 23747951. doi:10.1021/acscentsci.8b00357

2018 doi
[12]

Computer software review: Reaxys, 2009

Goodman, J. Computer software review: Reaxys, 2009

2009
[13]

W., Wu, H., Ranasinghe, D., Heid, E., Struble, T

Guan, Y., Coley, C. W., Wu, H., Ranasinghe, D., Heid, E., Struble, T. J., Pattanaik, L., Green, W. H., and Jensen, K. F. Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors. Chemical science, 12 0 (6): 0 219...

2021
[14]

Improving chemical reaction yield prediction using pre-trained graph neural networks

Han, J., Kwon, Y., Choi, Y.-S., and Kang, S. Improving chemical reaction yield prediction using pre-trained graph neural networks. Journal of Cheminformatics, 16 0 (1): 0 25, 2024

2024
[15]

and Oprea, T

Hasselgren, C. and Oprea, T. I. Artificial intelligence for drug discovery: Are we there yet? Annual Review of Pharmacology and Toxicology, 64 0 (1): 0 527--550, 2024

2024
[16]

P., Chung, Y., Li, S.-C., Graff, D

Heid, E., Greenman, K. P., Chung, Y., Li, S.-C., Graff, D. E., Vermeire, F. H., Wu, H., Green, W. H., and McGill, C. J. Chemprop: A machine learning package for chemical property prediction. Journal of Chemical Information and Modeling, 64 0 (1): 0 9--17, 2024. doi:10.1021/acs...

2024 doi
[17]

Open graph benchmark: Datasets for machine learning on graphs

Hu, W., Fey, M., Zitnik, M., Dong, Y., Ren, H., Liu, B., Catasta, M., and Leskovec, J. Open graph benchmark: Datasets for machine learning on graphs. Advances in neural information processing systems, 33: 0 22118--22133, 2020 a

2020
[18]

Strategies for pre-training graph neural networks

Hu, W., Liu, B., Gomes, J., Zitnik, M., Liang, P., Pande, V., and Leskovec, J. Strategies for pre-training graph neural networks. In International Conference on Learning Representations, 2020 b . URL https://openreview.net/forum?id=HJlWWJSFDH

2020
[19]

Irwin, R., Dimitriadis, S., He, J., and Bjerrum, E. J. Chemformer: a pre-trained transformer for computational chemistry. Machine Learning: Science and Technology, 3 0 (1): 0 015022, 2022

2022
[20]

Graph neural networks with multiple feature extraction paths for chemical property estimation

Ishida, S., Miyazaki, T., Sugaya, Y., and Omachi, S. Graph neural networks with multiple feature extraction paths for chemical property estimation. Molecules, 26 0 (11): 0 3125, 2021

2021
[21]

Drugood: Out-of-distribution dataset curator and benchmark for ai-aided drug discovery--a focus on affinity prediction problems with noise annotations

Ji, Y., Zhang, L., Wu, J., Wu, B., Li, L., Huang, L.-K., Xu, T., Rong, Y., Ren, J., Xue, D., et al. Drugood: Out-of-distribution dataset curator and benchmark for ai-aided drug discovery--a focus on affinity prediction problems with noise annotations. In Proceedings of the AAA...

2023
[22]

Predicting organic reaction outcomes with weisfeiler-lehman network

Jin, W., Coley, C., Barzilay, R., and Jaakkola, T. Predicting organic reaction outcomes with weisfeiler-lehman network. Advances in neural information processing systems, 30, 2017

2017
[23]

A substructure-based screening approach to uncover n-nitrosamines in drug substances

Kao, Y.-T., Wang, S.-F., Wu, M.-H., Her, S.-H., Yang, Y.-H., Lee, C.-H., Lee, H.-F., Lee, A.-R., Chang, L.-C., and Pao, L.-H. A substructure-based screening approach to uncover n-nitrosamines in drug substances. Journal of Food & Drug Analysis, 30 0 (1), 2022

2022
[24]

M., Maser, M

Kearnes, S. M., Maser, M. R., Wleklinski, M., Kast, A., Doyle, A. G., Dreher, S. D., Hawkins, J. M., Jensen, K. F., and Coley, C. W. The open reaction database. Journal of the American Chemical Society, 143 0 (45): 0 18820--18826, 2021

2021
[25]

W., Zhang, X., Krenske, E

Keto, A., Guo, T., Underdue, M., Stuyver, T., Coley, C. W., Zhang, X., Krenske, E. H., and Wiest, O. Data-efficient, chemistry-aware machine learning predictions of diels–alder reaction outcomes. Journal of the American Chemical Society, 6 2024. ISSN 0002-7863. doi:10.1021/jac...

2024 doi
[26]

A., Thiessen, P

Kim, S., Chen, J., Cheng, T., Gindulyte, A., He, J., He, S., Li, Q., Shoemaker, B. A., Thiessen, P. A., Yu, B., Zaslavsky, L., Zhang, J., and Bolton, E. E. PubChem in 2021: new data content and improved web interfaces . Nucleic Acids Research, 49 0 (D1): 0 D1388--D1395, 11 202...

2021 doi
[27]

Kingma, D. P. and Ba, J. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

2014 arXiv
[28]

Uncertainty-aware prediction of chemical reaction yields with graph neural networks

Kwon, Y., Lee, D., Choi, Y.-S., and Kang, S. Uncertainty-aware prediction of chemical reaction yields with graph neural networks. Journal of Cheminformatics, 14: 0 1--10, 2022

2022
[29]

Landrum, G. et al. Rdkit: A software suite for cheminformatics, computational chemistry, and predictive modeling. Greg Landrum, 8: 0 31, 2013

2013
[30]

Reaction performance prediction with an extrapolative and interpretable graph model based on chemical knowledge

Li, S.-W., Xu, L.-C., Zhang, C., Zhang, S.-Q., and Hong, X. Reaction performance prediction with an extrapolative and interpretable graph model based on chemical knowledge. Nature Communications, 14 0 (1): 0 3569, 2023

2023
[31]

Predicting regioselectivity in radical c- h functionalization of heterocycles through machine learning

Li, X., Zhang, S.-Q., Xu, L.-C., and Hong, X. Predicting regioselectivity in radical c- h functionalization of heterocycles through machine learning. Angewandte Chemie International Edition, 59 0 (32): 0 13253--13259, 2020

2020
[32]

and Zhang, Y

Lu, J. and Zhang, Y. Unified deep learning model for multitask reaction predictions with explanation. Journal of chemical information and modeling, 62 0 (6): 0 1376--1387, 2022

2022
[33]

The development of pharmacophore modeling: Generation and recent applications in drug discovery

Lu, X., Yang, H., Chen, Y., Li, Q., He, S.-y., Jiang, X., Feng, F., Qu, W., and Sun, H. The development of pharmacophore modeling: Generation and recent applications in drug discovery. Current pharmaceutical design, 24 0 (29): 0 3424--3439, 2018

2018
[34]

R., Cui, A

Maser, M. R., Cui, A. Y., Ryou, S., DeLano, T. J., Yue, Y., and Reisman, S. E. Multilabel classification models for the prediction of cross-coupling reaction conditions. Journal of Chemical Information and Modeling, 61 0 (1): 0 156--166, 2021

2021
[35]

Pistachio: Search and faceting of large reaction databases

Mayfield, J., Lowe, D., and Sayle, R. Pistachio: Search and faceting of large reaction databases. In ABSTRACTS OF PAPERS OF THE AMERICAN CHEMICAL SOCIETY, volume 254. AMER CHEMICAL SOC 1155 16TH ST, NW, WASHINGTON, DC 20036 USA, 2017

2017
[36]

Morgan, H. L. The generation of a unique machine description for chemical structures-a technique developed at chemical abstracts service. Journal of chemical documentation, 5 0 (2): 0 107--113, 1965

1965
[37]

T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models

Mou, C., Wang, X., Xie, L., Wu, Y., Zhang, J., Qi, Z., and Shan, Y. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pp.\ 4296--4304, 2024

2024
[38]

Emerging computational approaches for the study of regio-and stereoselectivity in organic synthesis

Nakliang, P., Yoon, S., and Choi, S. Emerging computational approaches for the study of regio-and stereoselectivity in organic synthesis. Organic Chemistry Frontiers, 8 0 (18): 0 5165--5181, 2021

2021
[39]

Recent developments in the reduction of aromatic and aliphatic nitro compounds to amines

Orlandi, M., Brenna, D., Harms, R., Jost, S., and Benaglia, M. Recent developments in the reduction of aromatic and aliphatic nitro compounds to amines. Organic Process Research & Development, pp.\ acs.oprd.6b00205, 2016

2016
[40]

Pytorch: An imperative style, high-performance deep learning library

Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32, 2019

2019
[41]

Reaction classification and yield prediction using the differential reaction fingerprint drfp

Probst, D., Schwaller, P., and Reymond, J.-L. Reaction classification and yield prediction using the differential reaction fingerprint drfp. Digital Discovery, 2022

2022
[42]

Molecule edit graph attention network: modeling chemical reactions as sequences of graph edits

Sacha, M., B az, M., Byrski, P., Dabrowski-Tumanski, P., Chrominski, M., Loska, R., W odarczyk-Pruszynski, P., and Jastrzebski, S. Molecule edit graph attention network: modeling chemical reactions as sequences of graph edits. Journal of Chemical Information and Modeling, 61 0...

2021
[43]

A structure-based platform for predicting chemical reactivity

Sandfort, F., Strieth-Kalthoff, F., Kühnemund, M., Beecks, C., and Glorius, F. A structure-based platform for predicting chemical reactivity. Chem, 6 0 (6): 0 1379--1390, 2020. ISSN 2451-9294. doi:https://doi.org/10.1016/j.chempr.2020.02.017. URL https://www.sciencedirect.com/...

2020 doi
[44]

A., Bekas, C., and Lee, A

Schwaller, P., Laino, T., Gaudin, T., Bolgar, P., Hunter, C. A., Bekas, C., and Lee, A. A. Molecular transformer: a model for uncertainty-calibrated chemical reaction prediction. ACS central science, 5 0 (9): 0 1572--1583, 2019

2019
[45]

Extraction of organic chemistry grammar from unsupervised learning of chemical reactions

Schwaller, P., Hoover, B., Reymond, J.-L., Strobelt, H., and Laino, T. Extraction of organic chemistry grammar from unsupervised learning of chemical reactions. Science Advances, 7 0 (15): 0 eabe4166, 2021 a

2021
[46]

C., Nair, V

Schwaller, P., Probst, D., Vaucher, A. C., Nair, V. H., Kreutter, D., Laino, T., and Reymond, J.-L. Mapping the space of chemical reactions using attention-based neural networks. Nature Machine Intelligence, 3 0 (2): 0 144--152, 2021 b

2021
[47]

C., Laino, T., and Reymond, J.-L

Schwaller, P., Vaucher, A. C., Laino, T., and Reymond, J.-L. Prediction of chemical reaction yields using deep learning. Machine learning: science and technology, 2 0 (1): 0 015016, 2021 c

2021
[48]

Seeman, J. I. The curtin-hammett principle and the winstein-holness equation: new definition and recent extensions to classical concepts. Journal of Chemical Education, 63 0 (1): 0 42, 1986

1986
[49]

Catalytic Transfer Hydrogenation of Nitronarenes under Mild Conditions

Shi, Q. Catalytic Transfer Hydrogenation of Nitronarenes under Mild Conditions. PhD thesis, Dalian University of Technology, 2007

2007
[50]

Prediction of chemical reaction yields with large-scale multi-view pre-training

Shi, R., Yu, G., Huo, X., and Yang, Y. Prediction of chemical reaction yields with large-scale multi-view pre-training. Journal of Cheminformatics, 16 0 (1): 0 22, 2024

2024
[51]

N., Kaiser, ., and Polosukhin, I

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, ., and Polosukhin, I. Attention is all you need. Advances in neural information processing systems, 30, 2017

2017
[52]

Graph attention networks

Veli c kovi \'c , P., Cucurull, G., Casanova, A., Romero, A., Li \`o , P., and Bengio, Y. Graph attention networks. In International Conference on Learning Representations, 2018

2018
[53]

Retroformer: Pushing the limits of end-to-end retrosynthesis transformer

Wan, Y., Hsieh, C.-Y., Liao, B., and Zhang, S. Retroformer: Pushing the limits of end-to-end retrosynthesis transformer. In Chaudhuri, K., Jegelka, S., Song, L., Szepesvari, C., Niu, G., and Sabato, S. (eds.), Proceedings of the 39th International Conference on Machine Learnin...

2022
[54]

Generic interpretable reaction condition predictions with open reaction condition datasets and unsupervised learning of reaction center

Wang, X., Hsieh, C.-Y., Yin, X., Wang, J., Li, Y., Deng, Y., Jiang, D., Wu, Z., Du, H., Chen, H., Li, Y., Liu, H., Wang, Y., Luo, P., Hou, T., and Yao, X. Generic interpretable reaction condition predictions with open reaction condition datasets and unsupervised learning of re...

2023 doi
[55]

Smiles, a chemical language and information system

Weininger, D. Smiles, a chemical language and information system. 1. introduction to methodology and encoding rules. Journal of chemical information and computer sciences, 28 0 (1): 0 31--36, 1988

1988
[56]

Retroxpert: Decompose retrosynthesis prediction like a chemist

Yan, C., Ding, Q., Zhao, P., Zheng, S., Yang, J., Yu, Y., and Huang, J. Retroxpert: Decompose retrosynthesis prediction like a chemist. Advances in Neural Information Processing Systems, 33: 0 11248--11258, 2020

2020
[57]

Learning substructure invariance for out-of-distribution molecular representations

Yang, N., Zeng, K., Wu, Q., Jia, X., and Yan, J. Learning substructure invariance for out-of-distribution molecular representations. Advances in Neural Information Processing Systems, 35: 0 12964--12978, 2022

2022
[58]

and Zhang, R

Yao, M. and Zhang, R. The rogress of the reduction reaction of aromatic nitro compounds. Chemical Industry and Engineering Progress, 0 (04): 0 19--25, 1984. ISSN 1000-6613

1984
[59]

Z., Zhao, Y., Xu, H., Kuramshin, A., et al

Yoshikawa, N., Skreta, M., Darvish, K., Arellano-Rubach, S., Ji, Z., Bj rn Kristensen, L., Li, A. Z., Zhao, Y., Xu, H., Kuramshin, A., et al. Large language models for chemistry robotics. Autonomous Robots, 47 0 (8): 0 1057--1086, 2023

2023
[60]

F., Henle, J

Zahrt, A. F., Henle, J. J., Rose, B. T., Wang, Y., Darrow, W. T., and Denmark, S. E. Prediction of higher-selectivity catalysts by computer-driven workflow and machine learning. Science, 363 0 (6424): 0 eaau5631, 2019

2019
[61]

Ualign: pushing the limit of template-free retrosynthesis prediction with unsupervised smiles alignment

Zeng, K., Yang, B., Zhao, X., Zhang, Y., Nie, F., Yang, X., Jin, Y., and Xu, Y. Ualign: pushing the limit of template-free retrosynthesis prediction with unsupervised smiles alignment. Journal of Cheminformatics, 16 0 (1): 0 80, 2024

2024
[62]

Adding conditional control to text-to-image diffusion models

Zhang, L., Rao, A., and Agrawala, M. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 3836--3847, 2023

2023
[63]

Uni-mol: A universal 3d molecular representation learning framework

Zhou, G., Gao, Z., Ding, Q., Zheng, H., Xu, H., Wei, Z., Zhang, L., and Ke, G. Uni-mol: A universal 3d molecular representation learning framework. In The Eleventh International Conference on Learning Representations, 2023. URL https://openreview.net/forum?id=6K2RM6wVqKu

2023
[64]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

T., Estrada, J

Ahneman, D. T., Estrada, J. G., Lin, S., Dreher, S. D., and Doyle, A. G. Predicting reaction performance in c--n cross-coupling using machine learning. Science, 360 0 (6385): 0 186--190, 2018

work page 2018

[2] [2]

Reagent prediction with a molecular transformer improves reaction data quality

Andronov, M., Voinarovska, V., Andronova, N., Wand, M., Clevert, D.-A., and Schmidhuber, J. Reagent prediction with a molecular transformer improves reaction data quality. Chemical Science, 14 0 (12): 0 3235--3246, 2023

work page 2023

[3] [3]

L., Kiros, J

Ba, J. L., Kiros, J. R., and Hinton, G. E. Layer normalization. arXiv preprint arXiv:1607.06450, 2016

arXiv 2016

[4] [4]

Uncertainty-aware yield prediction with multimodal molecular features

Chen, J., Guo, K., Liu, Z., Isayev, O., and Zhang, X. Uncertainty-aware yield prediction with multimodal molecular features. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pp.\ 8274--8282, 2024 a

work page 2024

[5] [5]

Precise atom-to-atom mapping for organic reactions via human-in-the-loop machine learning

Chen, S., An, S., Babazade, R., and Jung, Y. Precise atom-to-atom mapping for organic reactions via human-in-the-loop machine learning. Nature Communications, 15 0 (1): 0 2250, 2024 b

work page 2024

[6] [6]

and Guestrin, C

Chen, T. and Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, pp.\ 785--794, 2016

2016

[7] [7]

S., Babu, C

Dara, S., Dhamercherla, S., Jadav, S. S., Babu, C. M., and Ahsan, M. J. Machine learning in drug discovery: a review. Artificial intelligence review, 55 0 (3): 0 1947--1999, 2022

work page 1947

[8] [8]

BERT: pre-training of deep bidirectional transformers for language understanding

Devlin, J., Chang, M., Lee, K., and Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. In Burstein, J., Doran, C., and Solorio, T. (eds.), Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, ...

doi:10.18653/v1/n 2019