JTreeformer: Graph-Transformer via Latent-Diffusion Model for Molecular Generation

Chengxun Xie; Ji Shi; Miao Zhang; Xinming Zhang; Zhonghao Li

REVIEW 4 major objections 5 minor 38 references

JTreeformer: Graph-Transformer via Latent-Diffusion Model for Molecular Generation

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read JTreeformer claims that turning molecule graphs into junction trees lets a graph-transformer decoder generate molecules one fragment at a time, reaching 100% validity on MOSES with record diversity.

desk verdict Plausible architecture with a real MOSES diversity bump, but the QM9 table is misleading and the 'outperforms' claim doesn't hold up. read the letter →

arxiv 2504.20770 v1 pith:XECO5G4G submitted 2025-04-29 cs.LG cs.AI

Ji Shi , Chengxun Xie , Zhonghao Li , Xinming Zhang , Miao Zhang This is my paper · ORCID

classification cs.LGcs.AI

keywords moleculegenerationgraphtransformerjunctiontreelatentdiffusionDAGCNDDIMMOSESQM9

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

This paper proposes JTreeformer, a generative model that turns molecule generation into junction-tree generation. It claims that a graph transformer decoder can reconstruct a molecule one fragment at a time when a directed acyclic graph convolution feeds it information from the partially built structure, and that sampling the encoder's latent space with a diffusion model yields higher-quality, more diverse molecules than direct latent sampling. The result matters because transformer-based graph decoders have lagged behind graph-neural-network decoders for molecules, and the reported numbers suggest a way to close that gap while keeping generation chemically valid. If the claims hold, the same recipe—tree serialization, partial-graph decoding, and latent diffusion—offers a practical route to large-scale molecular exploration for drug discovery.

What carries the argument

The load-bearing object is the junction tree: each node is a chemical fragment such as a ring, bond, or atom, and a molecule is generated by predicting the next fragment plus one of four BFS position labels $P\in\{0,1,2,3\}$ that places it relative to already generated nodes. This reduces graph decoding to $2N-1$ autoregressive steps for a tree with $N$ nodes. The decoder's DAGCN layer, $K=I+\theta D^{-1/2}(D-M)D^{-1/2}$ with a mask $M=f_M(A)$, propagates information from the partial tree while preventing attention to future nodes, and it runs in parallel with masked multi-head attention. After encoder training, a DDIM diffusion model with a skip-connected network is trained in latent space; sampling starts from noise, denoises to a latent vector, and the decoder turns that vector back into a junction tree, with Monte Carlo tree search choosing among isomers during assembly.

What would settle it

Take every molecule in a held-out split of MOSES, compute its junction tree, traverse it in the paper's BFS order, and check that each node's relationship to its predecessor falls in $\{0,1,2,3\}$ and that the sequence reconstructs the tree exactly. One molecule whose tree cannot be serialized with these four labels, or that decodes to a wrong skeleton, falsifies the completeness assumption.

Watch

Extended reading notes

Core claim

The paper claims that molecule generation can be recast as junction-tree generation and that a decoder combining a Directed Acyclic Graph Convolution Network (DAGCN) with masked multi-head attention can iteratively reconstruct the molecule from the partial tree at each step. On MOSES it reports 100% validity, 98.6% uniqueness, 99.88% novelty, and an IntDiv1 of 0.8822; on QM9 it reports 78.5% uniqueness and 75.8% novelty, beating the listed baselines. It further claims that a DDIM diffusion model trained in the encoder's latent space improves diversity over direct sampling, and that the learned latent space supports meaningful interpolation and simple property-guided generation.

Load-bearing premise

The decoder's autoregressive reconstruction assumes the BFS four-position encoding ($P\in\{0,1,2,3\}$) is a complete way to serialize every junction tree, so that decoding the predicted position labels can rebuild the original molecular skeleton; the proof of completeness is deferred to the supplementary material and is not in this preprint.

Editorial extensions

If this is right

On MOSES, JTreeformer reports 100% validity, 98.6% uniqueness, 99.88% novelty, and an IntDiv1 of 0.8822, surpassing the listed baselines on internal diversity.
On QM9, it reports 78.5% uniqueness and 75.8% novelty, higher than the baselines taken from [15].
Switching from direct latent sampling to the diffusion sampler raises IntDiv1 from 0.8518 to 0.8822, so the diffusion component is what buys the diversity gain.
Removing DAGCN drops uniqueness to 0.176, showing the partial-tree convolution is load-bearing for the decoder.
Using Monte Carlo tree search for isomer selection removes the need for a separately trained isomer-prediction network used by earlier junction-tree models.

Reading between the lines

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

If the four-position BFS encoding is complete, the same serialization trick should transfer to other tree-structured generation tasks, such as syntax trees, scene graphs, or retrosynthesis plans, where a partial-tree decoder could generate structure one node at a time.
Because the model drops the KL term and lets diffusion supply the prior distribution, the paper implicitly suggests that latent diffusion can substitute for KL regularization in other discrete VAE-style generators; this is worth testing on non-chemical graph domains.
The property-guided experiments are presented as ongoing work; a natural extension is conditional diffusion with continuous property embeddings, then measuring whether latent interpolation controls logP or TPSA as precisely as the visual examples suggest.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Quick take: the paper is worth a serious referee, but the authors need to fix a load-bearing evaluation error before the claims should be trusted. The architecture—GCN+transformer encoder, DAGCN decoder, latent DDIM—is a sensible engineering contribution, and the MOSES IntDiv improvement (0.8822 vs. best prior ~0.8738) is the kind of result that would be useful if it reproduces. The DAGCN decoder is the most original piece, though its stability proof and the completeness proof for the BFS positional encoding are only referenced to a supplementary that isn't in this preprint.

The soft spots are real and in proportion. The QM9 comparison in Table 2 is seriously misleading: FCD and KL Div are lower-is-better by standard practice, including in the cited baselines, yet they're marked with up-arrows. Under the standard reading JTreeformer is worse than GraphDF and MARS on FCD, and worse than GCN and MARS on KL Div. So the abstract's broad 'outperforms existing molecule generation methods' is not supported by that table. The MOSES table also omits the standard FCD, SNN, Frag, and filter metrics, and no error bars are reported, so the only unambiguous win is diversity. The ablation table has a glaring anomaly—removing DAGCN while keeping features collapses uniqueness to 0.176—which suggests either an undocumented failure mode or a bug; the authors should explain it.

That said, I don't see circularity or fitted constants. The latent diffusion is tested by ablation, and the comparisons are against external tables. The main issues are reporting and verification, not fraud. The paper's own text acknowledges the supplement as the location of proofs; the problem is it's absent.

If I were the editor, I'd send it out. The architecture is plausible enough and the MOSES diversity number is interesting enough that a careful referee could separate signal from noise. But I'd expect major revisions: correct the QM9 metric directions, report full MOSES metrics with standard deviations, include the supplementary material, and explain the ablation anomaly. It's a conditionally acceptable paper, not a desk reject.

Referee Report

4 major / 5 minor

Summary. The paper proposes JTreeformer, a variational-autoencoder-style framework for molecular graph generation that converts molecules into junction trees, encodes them with a parallel GCN and multi-head attention encoder, decodes them with a novel Directed Acyclic Graph Convolution Network (DAGCN) combined with masked attention, and trains a DDIM latent diffusion model on the encoder's latent space. Molecular assembly is handled by Monte Carlo Tree Search. The authors report results on MOSES and QM9 and claim state-of-the-art performance, supported by ablation studies of the diffusion component and of the proposed modules.

Significance. The architectural combination is genuinely novel: a graph-transformer encoder with a DAGCN-based decoder and latent diffusion is a reasonable design direction for molecular generation, and the interpolation and property-optimization experiments offer suggestive evidence that the latent space is semantically structured. If the reported results are reproduced under the standard MOSES and QM9 protocols, the work would be a useful contribution to the graph-transformer and molecule-generation literature. The availability of an anonymized code repository is a plus for reproducibility. However, the current evidence base is not strong enough to support the abstract's unconditional claim of outperforming existing methods, because the QM9 comparison table is internally inconsistent, the MOSES evaluation omits standard protocol metrics, and key correctness proofs are deferred to an unavailable supplementary file.

major comments (4)

[Table 2] Table 2 labels the QM9 columns 'KL Div (↑)' and 'FCD (↑)' and uses these columns to support the claim that JTreeformer outperforms baselines. Under the standard QM9 evaluation protocol used by [15], FCD (Frechet ChemNet Distance) and KL divergence are lower-is-better metrics; under that reading JTreeformer's FCD=0.312 is worse than GraphDF=0.137 and MARS=0.123, and its KL Div=0.910 is worse than GCN=0.552 and MARS=0.547. If the up-arrows are instead taken literally, the metric definitions are nonstandard and PS-VAE is the FCD leader, not JTreeformer. In either interpretation, Table 2 does not support the abstract's claim that JTreeformer outperforms existing methods on QM9, so the table and the conclusion need to be corrected.
[§4, Tables 1 and 2] The evaluation is missing standard protocol components for both datasets. The MOSES benchmark normally reports FCD, SNN, Frag, and Filters in addition to the four metrics shown in Table 1; no QM9 validity is reported in Table 2; and although the Table 1 caption says 'Reported mean over three independent model initializations', no standard deviations or per-run values are given. Baseline numbers are drawn from other papers rather than measured under a common experimental protocol in this work, with Table 2 explicitly attributing baseline data to [15]. Without these components, the central claim of state-of-the-art performance is not established.
[Table 4] In the ablation study, removing DAGCN while retaining the proposed features collapses uniqueness from 0.986 to 0.176 and IntDiv1 from 0.8822 to 0.6059, whereas removing both features and DAGCN gives Unique=0.918 and IntDiv1=0.8795. This non-monotonic pattern is not explained and suggests a training instability or a bug in the 'features without DAGCN' configuration rather than a clean modular contribution. Please investigate and report what occurred in that setting, since this anomaly undermines the ablation-based attribution of the performance gain to DAGCN.
[§3.1 and §3.3] The decoder's correctness relies on two foundational claims that are deferred to a supplementary file that is not included in this preprint: (i) that a junction tree can be traversed in BFS order with only the four position types P∈{0,1,2,3} shown in Fig. 2 and generated autoregressively, and (ii) that the DAGCN layer in Eq. (8) provides stable training for arbitrary directed acyclic graphs. If either claim fails, the reconstructed molecular skeletons may be biased and the reported diversity and validity may not transfer to broader chemical space. The manuscript should include these proofs or a precise, accessible reference to the supplementary material.

minor comments (5)

[§4.1] Typos and incomplete sentences appear in the text, e.g., 'outperforms previous modela in molecule generation quality on MOSES se' and 'showaes' in §4.2; these should be corrected.
[§3.3, Eq. (10)] The loss notation is confusing: the reconstruction term is written as E[log p_theta(x|z)] and then expanded as alpha L_father + beta L_curr, while the text refers to 'the father's node y and the current junction u' rather than the p_J and p_pos factorization introduced in Eq. (7). The variables y, u, y_hat, and u_hat need precise definitions.
[§3.4, Eq. (12)] The DDIM noise-scale parameter eta appears in Eq. (12) but its value is never reported in the training details, and no sensitivity analysis is given; please state the value used.
[References] Reference [15] is listed simply as 'CoRR (2021)' without a title or identifier, and several other references are incomplete; the bibliography should be brought to a consistent, complete format.
[§4.2 and §4.4] The interpolation material is presented twice: Fig. 7 is referenced in §4.2 and then the same experiment is described again in §4.4; consolidate these sections to avoid redundancy.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: JTreeformer's results are benchmark comparisons and ablations, not fitted constants or self-citation chains.

full rationale

JTreeformer's derivation chain is: convert molecules to BFS junction-tree sequences (Eqs. 1-2), encode with parallel GCN + attention (Eqs. 3-6), decode autoregressively with DAGCN + masked attention (Eqs. 7-8), insert DDIM latent diffusion (Eqs. 11-13), and assemble with MCTS (Sec. 3.5). Each component is either standard external machinery (JT-VAE tree ordering, DDIM, U-Net skip connections, MCTS) or an architectural choice evaluated by ablation (Tables 3-4); no equation defines a fitted quantity as a predicted result. MOSES and QM9 comparisons copy baseline numbers from external sources ([15, 24]) and are not derived from JTreeformer's own parameters, so no reported number reduces by construction. The paper contains no load-bearing author self-citation. The text's assertions that the four-position BFS encoding and DAGCN stability are 'proofed in the supplementary material' (Sec. 3.1 and Sec. 3.3) are omitted proofs in this preprint, and Table 2's up-arrows on KL Div/FCD conflict with the standard lower-is-better reading of those metrics; these are correctness/completeness concerns, not circularity under the specified definitions.

Assumptions & free parameters 5 free parameters · 6 assumptions · 2 invented entities

The central claim rests on the faithfulness of junction-tree decomposition, the completeness of a four-case BFS position encoding, and the suitability of a non-regularized latent space for diffusion; two of these are deferred to a missing supplement. The model adds no externally validated new physical or chemical entity, only architectural components.

free parameters (5)

loss weight α = 1
Appears in Eq. (10) for the father-prediction term; chosen by hand, no sensitivity analysis.
loss weight β = 1
Appears in Eq. (10) for the current-junction term; chosen by hand, no sensitivity analysis.
auxiliary loss weight δ = 0.2
Weights the property-prediction auxiliary loss in Eq. (10); chosen by hand, no ablation shown.
DDIM noise-scale parameter η = not reported
Controls stochasticity in Eq. (12); the paper never states its value, so the diffusion schedule is not fully reproducible.
preprocessing caps (maxH, max_degree, max_layers) = 50, 20, 50
Hand-set limits for the junction-tree vocabulary in §4; they affect the representable molecular space and are not justified by data.

assumptions (6)

domain assumption Junction-tree decomposition (chart-based fragment decomposition) is a faithful representation of the molecular graph.
Used in §3.1 to convert molecules into junction trees and later reassemble them; inherited from JT-VAE [12] without validation on MOSES/QM9.
ad hoc to paper BFS permutation of a junction tree can be generated using only the four position types {0,1,2,3} shown in Fig. 2.
§3.1 claims this is 'proofed in the supplementary material,' but the supplement is not included in the preprint.
ad hoc to paper The DAGCN layer (Eq. 8) provides stable training and correct node prediction on directed acyclic graphs.
§3.3 claims 'DAGCN has a more stable training process than traditional GCN,' again deferred to the missing supplementary material.
domain assumption A VAE latent space trained without KL regularization is suitable for DDIM diffusion modeling.
§3.4 replaces the KL term with latent diffusion; the only support is the ablation in Table 3, not a formal argument.
domain assumption MCTS scoring (assembled-node count, property proximity) selects chemically valid isomers.
§3.5 uses MCTS to assemble molecules; no quantitative validation of assembly accuracy is reported.
standard math Self-attention and GCN message passing follow standard definitions in Eqs. (4)-(6).
The encoder relies on these operators without proof; they are standard in the literature.

invented entities (2)

DAGCN layer
purpose: To predict the next junction node in the decoder from the partially built junction tree
Introduced in §3.3 (Eq. 8); the preprint provides no proof or external validation, only the authors' internal ablation.
virtual [JNode] token
purpose: To aggregate global junction-tree information in the encoder, analogous to a [CLS] token
Introduced in §3.2 following Graphormer [35]; it is a standard technique and lacks independent validation in this paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of JTreeformer: Graph-Transformer via Latent-Diffusion Model for Molecular Generation." pith.science (2026). https://pith.science/paper/XECO5G4G

@misc{pith2026250420770,
  author       = {Pith},
  title        = {Pith review of: JTreeformer: Graph-Transformer via Latent-Diffusion Model for Molecular Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XECO5G4G}},
  note         = {Machine review of arXiv:2504.20770}
}

read the original abstract

The discovery of new molecules based on the original chemical molecule distributions is of great importance in medicine. The graph transformer, with its advantages of high performance and scalability compared to traditional graph networks, has been widely explored in recent research for applications of graph structures. However, current transformer-based graph decoders struggle to effectively utilize graph information, which limits their capacity to leverage only sequences of nodes rather than the complex topological structures of molecule graphs. This paper focuses on building a graph transformer-based framework for molecular generation, which we call \textbf{JTreeformer} as it transforms graph generation into junction tree generation. It combines GCN parallel with multi-head attention as the encoder. It integrates a directed acyclic GCN into a graph-based Transformer to serve as a decoder, which can iteratively synthesize the entire molecule by leveraging information from the partially constructed molecular structure at each step. In addition, a diffusion model is inserted in the latent space generated by the encoder, to enhance the efficiency and effectiveness of sampling further. The empirical results demonstrate that our novel framework outperforms existing molecule generation methods, thus offering a promising tool to advance drug discovery (https://anonymous.4open.science/r/JTreeformer-C74C).

Figures

Figures reproduced from arXiv: 2504.20770 by the authors.

**Figure 1.** Architecture of JTreeformer,which involves an encoder-decoder structure with attention mechanisms and graph [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** An example of prediction in junction node position. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Sampling by pre trained DDIM model in Latent space generated by JTreeformer in different diffusion steps of 50, 100, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: The leftmost one is the input molecule, followed [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Latent space generated by JTreeformer. Different colour is utilized to represent properties including the Weight, TPSA, [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Property optimization with relative molecular mass [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: The first and last columns are the starting and end [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 15 canonical work pages

[15]

Xiangzhe Kong, Zhixing Tan, and Yang Liu. 2021. GraphPiece: Efficiently Gener- ating High-Quality Molecular Graph with Substructures. CoRR (2021)

work page 2021
[1]

Rim Assouel, Mohamed Ahmed, Marwin H Segler, Amir Saffari, and Yoshua Bengio. 2018. Defactor: Differentiable edge factorization-based probabilistic graph generation. arXiv preprint arXiv:1811.09766 (2018)

arXiv 2018
[2]

Cameron B Browne, Edward Powley, Daniel Whitehouse, Simon M Lucas, Peter I Cowling, Philipp Rohlfshagen, Stephen Tavener, Diego Perez, Spyridon Samoth- rakis, and Simon Colton. 2012. A survey of monte carlo tree search methods. IEEE Transactions on Computational Intelligence and AI in games 4, 1 (2012), 1–43

2012
[3]

Nicola De Cao and Thomas Kipf. 2018. MolGAN: An implicit generative model for small molecular graphs. arXiv preprint arXiv:1805.11973 (2018)

arXiv 2018
[4]

Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convolu- tional neural networks on graphs with fast localized spectral filtering. Advances in neural information processing systems 29 (2016)

2016
[5]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding.arXiv preprint arXiv:1810.04805 (2018)

arXiv 2018
[6]

Prafulla Dhariwal and Alexander Nichol. 2021. Diffusion models beat gans on image synthesis. Advances in neural information processing systems 34 (2021), 8780–8794

2021
[7]

Michal Drozdzal, Eugene Vorontsov, Gabriel Chartrand, Samuel Kadoury, and Chris Pal. 2016. The Importance of Skip Connections in Biomedical Image Segmentation. CoRR abs/1608.04117 (2016)

arXiv 2016

Show all 38 references

[8]

Schoenholz, Patrick F

Justin Gilmer, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, and George E. Dahl. 2017. Neural Message Passing for Quantum Chemistry. In Proceedings of the 34th International Conference on Machine Learning . PMLR, 1263–1272

2017
[9]

Hirzel, Ryan P

Rafael Gómez-Bombarelli, David Duvenaud, José Miguel Hernández-Lobato, Jorge Aguilera-Iparraguirre, Timothy D. Hirzel, Ryan P. Adams, and Alán Aspuru-Guzik. 2016. Automatic chemical design using a data-driven contin- uous representation of molecules. CoRR abs/1610.02415 (2016)...

2016 arXiv
[10]

Will Hamilton, Zhitao Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs. Advances in neural information processing systems 30 (2017)

2017
[11]

Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems 33 (2020), 6840–6851

2020
[12]

Wengong Jin, Regina Barzilay, and Tommi Jaakkola. 2018. Junction tree varia- tional autoencoder for molecular graph generation. In International conference on machine learning. PMLR, 2323–2332

2018
[13]

Wengong Jin, Regina Barzilay, and Tommi Jaakkola. 2020. Hierarchical generation of molecular graphs using structural motifs. InInternational conference on machine learning. PMLR, 4839–4848

2020
[14]

Artur Kadurin, Alexander Aliper, Andrey Kazennov, Polina Mamoshina, Quentin Vanhaelen, Kuzma Khrabrov, and Alexander Zhavoronkov. 2016. The cornucopia of meaningful leads: Applying deep adversarial autoencoders for new molecule development in oncology. Oncotarget 8 (12 2016). ...

2016 doi
[16]

Devin Kreuzer, Dominique Beaini, Will Hamilton, Vincent Létourneau, and Pru- dencio Tossou. 2021. Rethinking graph transformers with spectral attention. Advances in Neural Information Processing Systems 34 (2021), 21618–21629

2021
[17]

Chunyuan Li, Xiang Gao, Yuan Li, Baolin Peng, Xiujun Li, Yizhe Zhang, and Jianfeng Gao. 2020. Optimus: Organizing Sentences via Pre-trained Modeling of a Latent Space. (2020)

2020
[18]

Yujia Li, Oriol Vinyals, Chris Dyer, Razvan Pascanu, and Peter Battaglia. 2018. Learning deep generative models of graphs. arXiv preprint arXiv:1803.03324 (2018)

2018 arXiv
[19]

Guangyi Liu, Zeyu Feng, Yuan Gao, Zichao Yang, Xiaodan Liang, Junwei Bao, Xiaodong He, Shuguang Cui, Zhen Li, and Zhiting Hu. 2023. Composable text controls in latent space with odes. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing . 1...

2023
[20]

Qi Liu, Miltiadis Allamanis, Marc Brockschmidt, and Alexander Gaunt. 2018. Constrained graph variational autoencoders for molecule design. Advances in neural information processing systems 31 (2018)

2018
[21]

Tengfei Ma, Jie Chen, and Cao Xiao. 2018. Constrained generation of semantically valid graphs via regularizing variational autoencoders. Advances in Neural Information Processing Systems 31 (2018)

2018
[22]

Erxue Min, Runfa Chen, Yatao Bian, Tingyang Xu, Kangfei Zhao, Wenbing Huang, Peilin Zhao, Junzhou Huang, Sophia Ananiadou, and Yu Rong. 2022. Trans- former for graphs: An overview from architecture perspective. arXiv preprint arXiv:2202.08455 (2022)

2022 arXiv
[23]

Joshua Mitton, Hans M Senn, Klaas Wynne, and Roderick Murray-Smith. 2021. A graph vae and graph transformer approach to generating molecular graphs. arXiv preprint arXiv:2104.04345 (2021)

2021 arXiv
[24]

Nikolenko, Alán Aspuru-Guzik, and Alex Zhavoronkov

Daniil Polykovskiy, Alexander Zhebrak, Benjamín Sánchez-Lengeling, Sergey Golovanov, Oktai Tatanov, Stanislav Belyaev, Rauf Kurbanov, Aleksey Artamonov, Vladimir Aladinskiy, Mark Veselov, Artur Kadurin, Sergey I. Nikolenko, Alán Aspuru-Guzik, and Alex Zhavoronkov. 2018. Molecu...

2018 arXiv
[25]

Mariya Popova, Mykhailo Shvets, Junier Oliva, and Olexandr Isayev. 2019. Molec- ularRNN: Generating realistic molecular graphs with optimized properties. arXiv preprint arXiv:1905.13372 (2019)

2019 arXiv
[26]

Kristina Preuer, Philipp Renz, Thomas Unterthiner, Sepp Hochreiter, and Günter Klambauer. 2018. Fréchet ChemblNet Distance: A metric for generative models for molecules. CoRR abs/1803.09518 (2018). arXiv:1803.09518 http://arxiv.org/ abs/1803.09518

2018 arXiv
[27]

Oleksii Prykhodko, Simon Johansson, Panagiotis-Christos Kotsias, Esben Bjerrum, Ola Engkvist, and Hongming Chen. 2019. A De Novo Molecular Generation Method Using Latent Vector Based Generative Adversarial Network. (06 2019). doi:10.26434/chemrxiv.8299544

2019 doi
[28]

Raghunathan Ramakrishnan, Pavlo O Dral, Matthias Rupp, and O Anatole von Lilienfeld. 2014. Quantum chemistry structures and properties of 134 kilo molecules. Scientific Data 1 (2014)

2014
[29]

Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 10684–10695

2022
[30]

Olaf Ronneberger, Philipp Fischer, and Thomas Brox. 2015. U-Net: Convolutional Networks for Biomedical Image Segmentation. CoRR abs/1505.04597 (2015)

2015 arXiv
[31]

Marwin H. S. Segler, Thierry Kogej, Christian Tyrchan, and Mark P. Waller. 2017. Generating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural Networks. CoRR abs/1701.01329 (2017). arXiv:1701.01329 http://arxiv. org/abs/1701.01329

2017 arXiv
[32]

Jiaming Song, Chenlin Meng, and Stefano Ermon. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020)

2020 arXiv
[33]

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

2017
[34]

David Weininger. 1988. SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules. Journal of chemical information and computer sciences 28, 1 (1988), 31–36

1988
[35]

Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, and Tie-Yan Liu. 2021. Do transformers really perform badly for graph representation? Advances in Neural Information Processing Systems 34 (2021), 28877–28888

2021
[36]

Zhitao Ying, Jiaxuan You, Christopher Morris, Xiang Ren, Will Hamilton, and Jure Leskovec. 2018. Hierarchical graph representation learning with differentiable pooling. Advances in neural information processing systems 31 (2018)

2018
[37]

Jiaxuan You, Bowen Liu, Zhitao Ying, Vijay Pande, and Jure Leskovec. 2018. Graph convolutional policy network for goal-directed molecular graph genera- tion. Advances in neural information processing systems 31 (2018)

2018
[38]

Jiaxuan You, Rex Ying, Xiang Ren, William Hamilton, and Jure Leskovec. 2018. Graphrnn: Generating realistic graphs with deep auto-regressive models. In International conference on machine learning . PMLR, 5708–5717

2018

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [15]

Xiangzhe Kong, Zhixing Tan, and Yang Liu. 2021. GraphPiece: Efficiently Gener- ating High-Quality Molecular Graph with Substructures. CoRR (2021)

work page 2021

[2] [1]

Rim Assouel, Mohamed Ahmed, Marwin H Segler, Amir Saffari, and Yoshua Bengio. 2018. Defactor: Differentiable edge factorization-based probabilistic graph generation. arXiv preprint arXiv:1811.09766 (2018)

arXiv 2018

[3] [2]

Cameron B Browne, Edward Powley, Daniel Whitehouse, Simon M Lucas, Peter I Cowling, Philipp Rohlfshagen, Stephen Tavener, Diego Perez, Spyridon Samoth- rakis, and Simon Colton. 2012. A survey of monte carlo tree search methods. IEEE Transactions on Computational Intelligence and AI in games 4, 1 (2012), 1–43

2012

[4] [3]

Nicola De Cao and Thomas Kipf. 2018. MolGAN: An implicit generative model for small molecular graphs. arXiv preprint arXiv:1805.11973 (2018)

arXiv 2018

[5] [4]

Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convolu- tional neural networks on graphs with fast localized spectral filtering. Advances in neural information processing systems 29 (2016)

2016

[6] [5]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding.arXiv preprint arXiv:1810.04805 (2018)

arXiv 2018

[7] [6]

Prafulla Dhariwal and Alexander Nichol. 2021. Diffusion models beat gans on image synthesis. Advances in neural information processing systems 34 (2021), 8780–8794

2021

[8] [7]

Michal Drozdzal, Eugene Vorontsov, Gabriel Chartrand, Samuel Kadoury, and Chris Pal. 2016. The Importance of Skip Connections in Biomedical Image Segmentation. CoRR abs/1608.04117 (2016)

arXiv 2016