Pith. sign in

REVIEW 4 major objections 7 minor 48 references

$\text{M}^{2}$LLM: Multi-view Molecular Representation Learning with Large Language Models

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Fusing three LLM views pushes ClinTox to 99.5% AUC.

desk verdict The rule-to-code feature idea is genuinely new, but the per-molecule fusion weights are unspecified, so the paper's test-set numbers cannot be produced from the method as written. read the letter →

arxiv 2508.08657 v1 pith:OEL4GC56 submitted 2025-08-12 cs.LG cs.AIcs.CL

Jiaxin Ju , Yizhen Zheng , Huan Yee Koh , Can Wang , Shirui Pan This is my paper · ORCID
classification cs.LGcs.AIcs.CL
keywords molecularpropertypredictionmulti-viewrepresentationlearninglargelanguagemodelsembeddingrule-basedfeaturesdynamicfusionMoleculeNetbenchmarkLLMreasoning
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

The paper tries to establish that a large language model can produce richer molecular representations than SMILES-only encoding or GNNs if it is prompted to view the molecule from three angles: its structure, the task being predicted, and the domain rules chemists use. To do this, it proposes M2LLM, which combines the LLM embeddings from the first two views with a feature vector built from LLM-generated rules, then fuses all three with per-molecule learned weights. The paper reports state-of-the-art results on several MoleculeNet benchmarks, including 99.5% ROC-AUC on ClinTox and RMSE 0.44 on ESOL, and shows the multi-view fusion consistently improves over a SMILES-only baseline. If true, this suggests that pretrained scientific knowledge in LLMs can substitute for much of the dataset-driven learning in molecular property prediction.

What carries the argument

The central mechanism is the three-view dynamic fusion. The structure view concatenates the SMILES string with three fixed insight questions (3D shape, intermolecular forces, chemical equilibrium) and encodes the text through the LLM. The task view appends the specific prediction question, e.g., "Will the chemical compound penetrate the blood-brain barrier?" to the SMILES. The rules view first has the LLM generate 20 scientific rules for the task and 3 pattern rules from sampled training molecules, then a rule-to-feature code translator turns each rule into a binary or numeric function of the SMILES. The fused representation is $z^{\text{fused}}_i = \alpha^{\text{struct}}_i z^{\text{struct}}_i + \alpha^{\text{task}}_i z^{\text{task}}_i + \alpha^{\text{rule}}_i z^{\text{rule}}_i$, with non-negative per-molecule weights summing to 1, learned jointly with an MLP.

What would settle it

A concrete test: hold out a set of molecules synthesized and labeled after the training cutoff of the LLM backbones, rerun M2LLM and the baselines on these held-out molecules, and check whether the multi-view gain over the SMILES-only baseline persists; if the gain vanishes, the reported benchmark results are best explained by pretraining contamination rather than by the multi-view reasoning.

Watch

Extended reading notes

Core claim

M2LLM claims that the best molecular representations are not learned from graphs or fingerprints alone but assembled from three LLM-driven views: a structure view that encodes SMILES alongside generic chemistry questions, a task view that encodes SMILES alongside the specific prediction question, and a rules view in which the LLM generates chemical rules (from scientific knowledge and from observed training data) that are translated into executable feature functions. These three representations are combined by a weighted sum with molecule-specific, learnable weights that sum to one, and the resulting vector feeds an MLP for classification or regression. The paper reports that this fusion reaches higher ROC-AUC and lower RMSE than all tested GNN and transformer baselines on eight MoleculeNet datasets, highlighting near-perfect ClinTox scores and a 56.9% error reduction on ESOL relative to the best baseline.

Load-bearing premise

The paper's empirical claims rest on the unstated assumption that the LLaMa, Galactica, and OpenAI embedding models were not trained on the MoleculeNet benchmark molecules and their labels; if those examples appear in the pretraining corpora, the reported gains could come from memorization rather than from the multi-view fusion.

Editorial extensions

If this is right

  • If the framework is correct, molecular property prediction no longer needs hand-crafted fingerprints or graph-specific architectures; the same LLM backbone, prompted differently, supplies both semantic and structural information.
  • The rules view yields an interpretable feature set: each generated rule (e.g., "molecular weight < 500 Da") corresponds to a human-readable condition, so predictions can be traced back to chemical reasoning.
  • The component analysis suggests that different tasks recruit different views: classification relies more on the structure view, regression on rules and task views, implying the dynamic fusion weights capture task-relevant inductive biases automatically.
  • Since the structure view is modular and its questions can be replaced or added, the framework can be extended to new molecular properties without retraining the backbone, only reformulating the prompts.

Reading between the lines

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

  • The near-perfect ClinTox results (99.1-99.5% across backbones) are suspiciously high; if the backbones' pretraining corpora contained MoleculeNet molecules and labels, the comparison would partly measure memorization rather than the multi-view framework.
  • The rule-to-feature translator effectively performs code synthesis from natural language and could be applied outside property prediction, e.g., to generate physically interpretable descriptors for other scientific quantitative structure-activity problems.
  • The per-molecule weights alpha provide a cheap interpretability signal: examining which view dominates for a given molecule could reveal why the model is confident, turning the method into a hypothesis generator for chemists.
  • The approach is likely sensitive to the quality of the LLM's scientific knowledge; on molecules with novel chemistries the rules view might generate misleading rules, and the dynamic weights might not compensate.
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.

Referee Report

4 major / 7 minor

Summary. The paper proposes M2LLM, a multi-view molecular representation learning framework that combines three LLM-derived views: a molecular structure view (SMILES with structural questions), a molecular task view (SMILES with task-specific questions), and a molecular rules view (LLM-generated scientific and data-pattern rules converted to features). The views are combined through a per-molecule weighted fusion, Eq. (6), and an MLP performs the final classification or regression. The authors evaluate M2LLM on eight MoleculeNet datasets with scaffold splits using Galactica, LLaMa-3.1, and OpenAI embedding backbones, reporting state-of-the-art results on several classification and regression benchmarks, and they analyze per-view contributions and compare against a SMILES-only baseline.

Significance. If the method is fully specified and the empirical results hold, the paper offers a useful direction by showing that LLM semantic knowledge and rule-based reasoning can complement molecular structure encodings, and the multi-view fusion idea is potentially interesting for molecular property prediction. The paper has concrete strengths: it evaluates across multiple datasets and backbones, includes a SMILES-only ablation, and provides per-component contribution analysis. However, the central fusion mechanism is presently underspecified to the point that the test-time forward pass cannot be reproduced, and several performance claims are weaker than the text suggests when standard deviations are inspected. The significance of the contribution depends on resolving these issues.

major comments (4)
  1. [§3.3, Eq. (6)] The fusion mechanism is not computable as written. The per-molecule weights α_struct_i, α_task_i, and α_rule_i are described as 'learnable weights' that are 'specific to each molecule,' but no formula, network, or inference procedure is given for producing these weights for a molecule in the scaffold-split test set. If they are free scalar parameters fit on the training split, a test molecule would have no α_i. If they are the output of a learned function, that function is the core of the dynamic fusion claim and must be specified. In addition, the three view vectors have incompatible dimensionalities: z_struct_i is the concatenation of three LLM embeddings (Eq. (2)), z_task_i is a single LLM embedding (Eq. (3)), and z_rule_i is a rule-based feature vector whose length is the number of rules (Eq. (5)). No projection to a common space is described, so the weighted sum in Eq. (6) is not a well-defined vector operation as stated. Because the reported test-set numbers in Tables 1 and 2 are generated by this forward pass, the method is not reproducible from the manuscript.
  2. [§4.2, Tables 1 and 2] The claim of state-of-the-art performance is only partially supported by the reported means and standard deviations. On BACE, M2LLM (80.0) is below RF+ECFP4 (85.0), and the paper acknowledges this. On HIV, SIDER, and BBBP, the reported improvements over the best baseline are often within one standard deviation (e.g., HIV 79.5±0.7 vs Uni-Mol 78.3±1.3; SIDER 63.7±0.3 vs Uni-Mol 62.3±5.6; BBBP 75.5±1.3 vs MolCLR 73.1±1.6). No significance tests or paired comparisons are reported. The regression gains on FreeSolv for M2LLM(Galactica) (2.39±1.39) overlap substantially with the best baseline (2.38±0.40). The phrase 'surpassing existing baselines with significant improvements' should be qualified, and the state-of-the-art claim should be supported with appropriate statistical tests or restricted to the datasets where the improvement is clear.
  3. [§4.5 and Appendix A.4, Table 5] The text in §4.5 claims that M2LLM 'consistently improved the scores across all six datasets' and that on ClinTox 'our framework offering a marginal improvement.' This is contradicted by Appendix Table 5, where for most backbones the SMILES-only baseline achieves a higher ClinTox ROC-AUC than M2LLM (e.g., llama-3.1-8B: 99.68±0.09 vs 98.67±2.11; galactica-6.7b: 100.00±0.00 vs 99.93±0.05; text-embedding-3-large: 99.86±0.00 vs 99.38±0.09). The discrepancy between Figure 4, which shows ClinTox M2LLM at 1.000, and the appendix table needs to be resolved and the over-claim corrected.
  4. [§4.1 and §3.2] The evaluation does not address the possibility that the pretraining corpora of Galactica, LLaMa-3.1, and the OpenAI embedding models contain the MoleculeNet benchmark molecules and labels. Given the near-perfect ClinTox scores and the large ESOL improvement, the paper should at least discuss this contamination risk and, where feasible, provide an analysis such as checking SMILES overlap with public pretraining data or motivating why memorization is unlikely for held-out scaffold splits. This is an external-validity concern that directly affects the interpretation of the claimed state-of-the-art results.
minor comments (7)
  1. [§3.1] There is a typo in the text: 'Sructure Insight 3' should be 'Structure Insight 3.'
  2. [§4.2, Table 1] The table caption says 'with bold being the best result, and bold being the second best result,' but both are described as 'bold'; the two bold styles need to be distinguished or the caption clarified.
  3. [§4.2] The dataset name is spelled inconsistently as 'Clintox' in the text and 'ClinTox' elsewhere; please standardize.
  4. [§4.5, Figure 4] In Figure 4, the classification subplot uses a scale starting near 0.5, which visually compresses the differences for ClinTox (near 1.0); consider using a separate scale or adding numerical labels above each bar.
  5. [References] In the reference list, 'OpenAI et al., 2023' appears with a stray 'Josh' before 'OpenAI'; this should be corrected.
  6. [§3.2] The Rule-to-Feature Code Translator is described as using LLMs to convert rules into executable code, but no details are given about the programming language, execution environment, or how the code functions are validated; this is important for reproducibility and should be specified, even in an appendix.
  7. [Appendix A.2] The bar charts in Figure 5 are not very informative for the classification datasets because most values are close to 1.0; a table with numeric values or a zoomed plot would be clearer.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the empirical multi-view pipeline is not a derivation that reduces to its inputs; the underspecified fusion weights are a reproducibility gap, not a circular step.

full rationale

The paper's central claims are empirical results on held-out scaffold splits rather than formal derivations, and I could not exhibit any equation or fitted parameter that is equivalent to the reported predictions by construction. The structure and task embeddings (Eqs. 1-3) are LLM encodings of SMILES and prompts; the rules view (Eqs. 4-5) is supervised feature construction from training-label subsets, which is standard feature engineering and does not use test labels; and the fused prediction (Eqs. 6-7) is a weighted sum followed by an MLP, so no test label enters the training procedure. The dynamic per-molecule weights alpha_i in Eq. (6) are underspecified for test molecules, which makes the method incomplete as written, but under-specification is a reproducibility gap rather than a circular reduction. The self-citations (e.g., [Zheng et al., 2025] for LLM pattern-recognition abilities) support background premises and are not load-bearing uniqueness or equivalence claims; LLM pretraining contamination would be an external-validity threat, not a circularity. Therefore the paper is not circular in any of the seven monitored senses.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method adds no new physical or theoretical entities. Its load-bearing premises are domain assumptions about LLM capabilities and data hygiene, plus two hand-chosen rule-generation settings. The per-molecule fusion weights are the main fitted parameters of the framework.

free parameters (3)
  • per-molecule fusion weights alpha_i (alpha_struct, alpha_task, alpha_rule) = not reported
    Learnable per-molecule scalars in Eq. (6) with a simplex constraint; no parameterization or update rule is specified in Section 3.3.
  • number of scientific rules per task = 20
    LLM prompt in Section 3.2 requests 20 rules; hand-chosen, and the number of data-pattern rules depends on unreported K and m.
  • rule subset count K and subset size m = not reported
    Section 3.2 defines R_data(t) over K subsets of m molecules but never states K or m; these control the rule feature dimension.
assumptions (4)
  • domain assumption LLM embeddings and generated rules are chemically meaningful for the target properties when prompted with SMILES and questions.
    Core assumption implicit in Sections 3.1 and 3.2; if false, the multi-view representations carry no signal.
  • domain assumption LLM pretraining corpora do not contain the MoleculeNet benchmark test molecules or labels.
    Unflagged in Section 4; if contamination occurred, reported test scores would be inflated.
  • domain assumption The scaffold split is applied correctly and identically across all methods.
    Section 4.1 states scaffold splitting; if the split or data processing differs, comparisons could be biased.
  • domain assumption Rule-to-feature code generated by the LLM is correct with respect to the molecular structures and returns reliable features.
    Section 3.2 relies on LLM-generated executable functions; no verification of code correctness is described.

how reviews work

0 comments
Cite this review

Pith. "Pith review of $\text{M}^{2}$LLM: Multi-view Molecular Representation Learning with Large Language Models." pith.science (2026). https://pith.science/paper/OEL4GC56

@misc{pith2026250808657,
  author       = {Pith},
  title        = {Pith review of: $\textM^2$LLM: Multi-view Molecular Representation Learning with Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OEL4GC56}},
  note         = {Machine review of arXiv:2508.08657}
}
abstract

Accurate molecular property prediction is a critical challenge with wide-ranging applications in chemistry, materials science, and drug discovery. Molecular representation methods, including fingerprints and graph neural networks (GNNs), achieve state-of-the-art results by effectively deriving features from molecular structures. However, these methods often overlook decades of accumulated semantic and contextual knowledge. Recent advancements in large language models (LLMs) demonstrate remarkable reasoning abilities and prior knowledge across scientific domains, leading us to hypothesize that LLMs can generate rich molecular representations when guided to reason in multiple perspectives. To address these gaps, we propose $\text{M}^{2}$LLM, a multi-view framework that integrates three perspectives: the molecular structure view, the molecular task view, and the molecular rules view. These views are fused dynamically to adapt to task requirements, and experiments demonstrate that $\text{M}^{2}$LLM achieves state-of-the-art performance on multiple benchmarks across classification and regression tasks. Moreover, we demonstrate that representation derived from LLM achieves exceptional performance by leveraging two core functionalities: the generation of molecular embeddings through their encoding capabilities and the curation of molecular features through advanced reasoning processes.

Figures

Figures reproduced from arXiv: 2508.08657 by the authors.

Figure 1
Figure 1. (a) Traditional Method: Converts SMILES into fixed [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The M2LLM Molecular Representation Learning Framework. The framework integrates three molecular views to create comprehensive and adaptable representations. Embedding generation includes two views: the molecular structure view and the molecular task view, both processed using LLMs to produce semantic representations. Feature curation provides one view: the molecular rules view, where rules are generated using LLMs b… view at source ↗
Figure 4
Figure 4. Comparison of the performance between M2LLM and the SMILES-only representation across 10 random seeds on six datasets. In the case of the ClinTox dataset, the SMILES-only base￾line achieves nearly perfect results, with our framework of￾fering a marginal improvement. This suggests that in tasks where the LLM already possesses sufficient understanding of the molecular domain through SMILES-based encoding alone, the ad… view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Evaluation of Different Backbone Models on 5 classification datasets and 3 regression datasets.In the classification datasets, all [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 31 canonical work pages

  1. [1]

    Random forests

    [Breiman, 2001] Leo Breiman. Random forests. Machine learning, 45:5–32,

    2001

  2. [6]

    The llama 3 herd of models

    [Dubey et al., 2024] Abhimanyu Dubey, Abhinav Jauhri, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783,

    arXiv 2024

  3. [7]

    Molecular representation learning with language mod- els and domain-relevant auxiliary tasks

    [Fabian et al., 2020] Benedek Fabian, Thomas Edlich, et al. Molecular representation learning with language mod- els and domain-relevant auxiliary tasks. arXiv preprint arXiv:2011.13230,

    arXiv 2020

  4. [10]

    Strategies for pre-training graph neural networks

    [Hu et al., 2019] Weihua Hu, Bowen Liu, et al. Strategies for pre-training graph neural networks. arXiv preprint arXiv:1905.12265,

    arXiv 2019

  5. [12]

    Bert: Pre-training of deep bidirectional transformers for lan- guage understanding

    [Kenton and Toutanova, 2019] Jacob Devlin Ming- Wei Chang Kenton and Lee Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for lan- guage understanding. In Proceedings of naacL-HLT , volume

    2019

  6. [13]

    Physicochemi- cal graph neural network for learning protein–ligand inter- action fingerprints from sequence data

    [Koh et al., 2024] Huan Yee Koh, Anh TN Nguyen, Shirui Pan, Lauren T May, and Geoffrey I Webb. Physicochemi- cal graph neural network for learning protein–ligand inter- action fingerprints from sequence data. Nature Machine Intelligence, pages 1–15,

    work page 2024

  7. [14]

    Large language models are zero-shot reason- ers

    [Kojima et al., 2022] Takeshi Kojima, Shixiang Shane Gu, et al. Large language models are zero-shot reason- ers. Advances in neural information processing systems , 35:22199–22213,

    work page 2022

  8. [16]

    Combining ic50 or k i values from different sources is a source of significant noise

    [Landrum and Riniker, 2024] Gregory A Landrum and Sere- ina Riniker. Combining ic50 or k i values from different sources is a source of significant noise. Journal of Chemi- cal Information and Modeling, 64(5):1560–1567,

    work page 2024

Show all 48 references
  1. [17]

    Pre- training molecular graph representation with 3d geometry

    [Liu et al., 2022] Shengchao Liu, Hanchen Wang, et al. Pre- training molecular graph representation with 3d geometry. In International Conference on Learning Representations,

    2022

  2. [18]

    Learning multi-view molecular representations with structured and unstructured knowledge

    [Luo et al., 2024] Yizhen Luo, Kai Yang, et al. Learning multi-view molecular representations with structured and unstructured knowledge. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 2082–2093,

    2024

  3. [19]

    A bayesian approach to in silico blood-brain bar- rier penetration modeling

    [Martins et al., 2012] Ines Filipa Martins, Ana L Teixeira, et al. A bayesian approach to in silico blood-brain bar- rier penetration modeling. Journal of chemical informa- tion and modeling, 52(6):1686–1697,

    2012

  4. [21]

    Are large language models superhuman chemists? arXiv preprint arXiv:2404.01475,

    [Mirza et al., 2024] Adrian Mirza, Nawaf Alampara, et al. Are large language models superhuman chemists? arXiv preprint arXiv:2404.01475,

    2024 arXiv

  5. [22]

    Freesolv: a database of experimental and cal- culated hydration free energies, with input files

    [Mobley and Guthrie, 2014] David L Mobley and J Peter Guthrie. Freesolv: a database of experimental and cal- culated hydration free energies, with input files. Journal of computer-aided molecular design, 28:711–720,

    2014

  6. [25]

    Extended-connectivity fingerprints

    [Rogers and Hahn, 2010] David Rogers and Mathew Hahn. Extended-connectivity fingerprints. Journal of chemical information and modeling, 50(5):742–754,

    2010

  7. [27]

    Self- supervised graph transformer on large-scale molecular data

    [Rong et al., 2020] Yu Rong, Yatao Bian, et al. Self- supervised graph transformer on large-scale molecular data. Advances in neural information processing systems, 33:12559–12571,

    2020

  8. [28]

    Large-scale chemical language representations capture molecular structure and properties

    [Ross et al., 2022] Jerret Ross, Brian Belgodere, et al. Large-scale chemical language representations capture molecular structure and properties. Nature Machine In- telligence, 4(12):1256–1264,

    2022

  9. [29]

    Comparative analysis of llama and chatgpt embeddings for molecule embedding

    [Sadeghi et al., 2024] Shaghayegh Sadeghi, Alan Bui, Ali Forooghi, Jianguo Lu, and Alioune Ngom. Comparative analysis of llama and chatgpt embeddings for molecule embedding. arXiv preprint arXiv:2402.00024,

    2024 arXiv

  10. [30]

    A multi-view mixture-of-experts based on language and graphs for molecular properties prediction

    [Shirasuna et al., 2024] Victor Yukio Shirasuna, Eduardo Soares, et al. A multi-view mixture-of-experts based on language and graphs for molecular properties prediction. In ICML 2024 AI for Science Workshop,

    2024

  11. [31]

    3d infomax improves gnns for molecular property prediction

    [St¨ark et al., 2022] Hannes St ¨ark, Dominique Beaini, Gabriele Corso, Prudencio Tossou, Christian Dallago, Stephan G ¨unnemann, and Pietro Li `o. 3d infomax improves gnns for molecular property prediction. In International Conference on Machine Learning , pages 20479–20502. PMLR,

    2022

  12. [32]

    Com- putational modeling of β-secretase 1 (bace-1) inhibitors using ligand based approaches

    [Subramanian et al., 2016] Govindan Subramanian, Bharath Ramsundar, Vijay Pande, and Rajiah Aldrin Denny. Com- putational modeling of β-secretase 1 (bace-1) inhibitors using ligand based approaches. Journal of chemical infor- mation and modeling, 56(10):1936–1949,

    2016

  13. [33]

    Galactica: A large language model for science

    [Taylor et al., 2022] Ross Taylor, Marcin Kardas, Guillem Cucurull, Thomas Scialom, Anthony Hartshorn, Elvis Sar- avia, Andrew Poulton, Viktor Kerkez, and Robert Stojnic. Galactica: A large language model for science. arXiv preprint arXiv:2211.09085,

    2022 arXiv

  14. [34]

    Smiles- bert: large scale unsupervised pre-training for molecular property prediction

    [Wang et al., 2019] Sheng Wang, Yuzhi Guo, et al. Smiles- bert: large scale unsupervised pre-training for molecular property prediction. In Proceedings of the 10th ACM in- ternational conference on bioinformatics, computational biology and health informatics, pages 429–436,

    2019

  15. [35]

    Molecular contrastive learning of representations via graph neural networks

    [Wang et al., 2022] Yuyang Wang, Jianren Wang, Zhonglin Cao, and Amir Barati Farimani. Molecular contrastive learning of representations via graph neural networks. Na- ture Machine Intelligence, 4(3):279–287,

    2022

  16. [36]

    Smiles, a chemical lan- guage and information system

    [Weininger, 1988] David Weininger. Smiles, a chemical lan- guage and information system

    1988

  17. [39]

    Mole-bert: Rethinking pre-training graph neu- ral networks for molecules

    [Xia et al., 2022] Jun Xia, Chengshuai Zhao, Bozhen Hu, Zhangyang Gao, Cheng Tan, Yue Liu, Siyuan Li, and Stan Z Li. Mole-bert: Rethinking pre-training graph neu- ral networks for molecules. In The Eleventh International Conference on Learning Representations,

    2022

  18. [40]

    Analyzing learned molecular representations for property prediction

    [Yang et al., 2019] Kevin Yang, Kyle Swanson, Wengong Jin, et al. Analyzing learned molecular representations for property prediction. Journal of chemical information and modeling, 59(8):3370–3388,

    2019

  19. [41]

    Graph con- trastive learning with augmentations

    [You et al., 2020] Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, and Yang Shen. Graph con- trastive learning with augmentations. Advances in neural information processing systems, 33:5812–5823,

    2020

  20. [42]

    Kernel readout for graph neural networks

    [Yu et al., 2024] Jiajun Yu, Zhihao Wu, Jinyu Cai, Adele Lu Jia, and Jicong Fan. Kernel readout for graph neural networks. In Proceedings of the Thirty-Third Interna- tional Joint Conference on Artificial Intelligence, IJCAI- 24, pages 2505–2514,

    2024

  21. [43]

    Collab- orative expert llms guided multi-objective molecular opti- mization

    [Yu et al., 2025] Jiajun Yu, Yizhen Zheng, Huan Yee Koh, Shirui Pan, Tianyue Wang, and Haishuai Wang. Collab- orative expert llms guided multi-objective molecular opti- mization. arXiv preprint arXiv:2503.03503,

    2025 arXiv

  22. [45]

    Dynamic graph unlearning: A general and efficient post-processing method via gradient transformation

    [Zhang et al., 2025] He Zhang, Bang Wu, Xiangwen Yang, Xingliang Yuan, Xiaoning Liu, and Xun Yi. Dynamic graph unlearning: A general and efficient post-processing method via gradient transformation. In Proceedings of the ACM on Web Conference 2025, pages 931–944,

    2025

  23. [46]

    Large language models in drug discovery and development: From disease mechanisms to clinical trials

    [Zheng et al., 2024c] Yizhen Zheng, Huan Yee Koh, Mad- die Yang, Li Li, Lauren T May, Geoffrey I Webb, Shirui Pan, and George Church. Large language models in drug discovery and development: From disease mechanisms to clinical trials. arXiv preprint arXiv:2409.04481,

    arXiv

  24. [47]

    Large language models for scientific dis- covery in molecular property prediction

    [Zheng et al., 2025] Yizhen Zheng, Huan Yee Koh, Jiaxin Ju, Anh TN Nguyen, Lauren T May, Geoffrey I Webb, and Shirui Pan. Large language models for scientific dis- covery in molecular property prediction. Nature Machine Intelligence, pages 1–11,

    2025

  25. [48]

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

    [Zhou et al., 2023] Gengmo Zhou, Zhifeng Gao, et al. Uni- mol: A universal 3d molecular representation learning framework. In The Eleventh International Conference on Learning Representations,

    2023

  26. [1988]

    Moleculenet: a benchmark for molecular machine learning

    [Wu et al., 2018] Zhenqin Wu, Bharath Ramsundar, Evan N Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S Pappu, Karl Leswing, and Vijay Pande. Moleculenet: a benchmark for molecular machine learning. Chemical sci- ence, 9(2):513–530,

    2018

  27. [2000]

    Mmgnn: A molecular merged graph neural network for explainable solvation free energy prediction

    [Du et al., 2024] Wenjie Du, Shuai Zhang, Jun Xia Di Wu, Ziyuan Zhao, Junfeng Fang, and Yang Wang. Mmgnn: A molecular merged graph neural network for explainable solvation free energy prediction. In Proceedings of the Thirty-Third International Joint Conference on Artificial I...

    2024

  28. [2001]

    Improving augmentation consistency for graph contrastive learning

    [Bu et al., 2024] Weixin Bu, Xiaofeng Cao, Yizhen Zheng, and Shirui Pan. Improving augmentation consistency for graph contrastive learning. Pattern Recognition , 148:110182,

    2024

  29. [2004]

    Drug discovery: a historical perspective

    [Drews, 2000] Jurgen Drews. Drug discovery: a historical perspective. science, 287(5460):1960–1964,

    2000

  30. [2010]

    Molprop: Molecular property prediction with multimodal language and graph fusion

    [Rollins et al., 2024] Zachary A Rollins, Alan C Cheng, and Essam Metwally. Molprop: Molecular property prediction with multimodal language and graph fusion. Journal of Cheminformatics, 16(1):56,

    2024

  31. [2012]

    Recurrent neural networks

    [Medsker et al., 2001] Larry R Medsker, Lakhmi Jain, et al. Recurrent neural networks. Design and Applications , 5(64-67):2,

    2001

  32. [2014]

    Gpt-4 technical report

    [OpenAI et al., 2023] Josh OpenAI, Achiam, Steven Adler, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774,

    2023 arXiv

  33. [2016]

    The iuphar/bps guide to pharmacology in

    [Harding et al., 2024] Simon D Harding, Jane F Armstrong, et al. The iuphar/bps guide to pharmacology in

    2024

  34. [2018]

    Graph learning under distribution shifts: A comprehensive sur- vey on domain adaptation, out-of-distribution, and contin- ual learning

    [Wu et al., 2024] Man Wu, Xin Zheng, Qin Zhang, Xiao Shen, Xiong Luo, Xingquan Zhu, and Shirui Pan. Graph learning under distribution shifts: A comprehensive sur- vey on domain adaptation, out-of-distribution, and contin- ual learning. arXiv preprint arXiv:2402.16374,

    2024 arXiv

  35. [2019]

    Fp2vec: a new molecular featurizer for learning molecular properties

    [Jeon and Kim, 2019] Woosung Jeon and Dongsup Kim. Fp2vec: a new molecular featurizer for learning molecular properties. Bioinformatics, 35(23):4979–4985,

    2019

  36. [2020]

    A data-driven approach to predict- ing successes and failures of clinical trials

    [Gayvert et al., 2016] Kaitlyn M Gayvert, Neel S Madhukar, and Olivier Elemento. A data-driven approach to predict- ing successes and failures of clinical trials. Cell chemical biology, 23(10):1294–1301,

    2016

  37. [2022]

    The sider database of drugs and side effects

    [Kuhn et al., 2016] Michael Kuhn, Ivica Letunic, Lars Juhl Jensen, and Peer Bork. The sider database of drugs and side effects. Nucleic acids research, 44(D1):D1075– D1079,

    2016

  38. [2023]

    New embedding models and api updates,

    [OpenAI, 2024] OpenAI. New embedding models and api updates,

    2024

  39. [2024]

    Esol: estimating aqueous solubility directly from molecular structure

    [Delaney, 2004] John S Delaney. Esol: estimating aqueous solubility directly from molecular structure. Journal of chemical information and computer sciences, 44(3):1000– 1005,

    2004

  40. [2025]

    Fast and efficient calculations of structural invariants of chirality

    [Zhang et al., 2019] He Zhang, Hanlin Mo, You Hao, Qi Li, Shirui Li, and Hua Li. Fast and efficient calculations of structural invariants of chirality. Pattern Recognit. Lett., 128:270–277,

    2019

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.