Pith. sign in

REVIEW 4 major objections 7 minor 31 references

GeoScatt-GNN: A Geometric Scattering Transform-Based Graph Neural Network Model for Ames Mutagenicity Prediction

T0 review · 4 major / 7 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 Scattering-based molecular features are claimed to lift Ames mutagenicity prediction to a 0.9812 AUC, ahead of a CNN baseline.

desk verdict Plausible scattering application, but the headline AUC is unverifiable because the pipeline never separates test molecules from feature construction and the meta-graph. read the letter →

arxiv 2411.15331 v1 pith:XFCUAHYL submitted 2024-11-22 cs.LG cs.AIeess.IVq-bio.QM

Abdeljalil Zoubir , Badr Missaoui This is my paper · ORCID
classification cs.LGcs.AIeess.IVq-bio.QM MSC 68T0768R1092E10
keywords mutagenicitygraphneuralnetworksgeometricscatteringtransformtoxicitypredictionwaveletmoleculargraphsSAGEAmestest
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 show that scattering-transform features, computed from molecular graphs and from 2D molecular images, are a more informative input for Ames mutagenicity prediction than standard chemical descriptors, fingerprints, or learned CNN image features. On the paper's benchmark (6,277 compounds after preprocessing), it reports that a LightGBM classifier fed geometric-scattering features plus GIN embeddings reaches an AUC of 0.9812, and that a new graph-of-graphs model called MOLG3-SAGE reaches 0.9301 accuracy and 0.8603 MCC. If these numbers hold under honest splitting, the practical upshot is that a training-free wavelet feature extractor can replace much of the learned featurization in toxicity screening, improving accuracy while reducing model complexity. The paper presents this as a step toward faster and more reliable computational chemical safety assessment.

What carries the argument

The load-bearing object is the geometric scattering transform on molecular graphs, built from tight Hann wavelets and diffusion wavelets, which produces multiscale, permutation-invariant node embeddings; the 2D scattering transform on molecule images plays a supporting role. The hybrid pipeline concatenates these scattering features with graph-isomorphism-network (GIN) embeddings and feeds the result into standard classifiers, while MOLG3-SAGE replaces the classifier with a graph-of-graphs: a fully connected meta-graph whose nodes are molecules, whose edge weights are Gaussian-kernel cosine similarities between scattering embeddings, and whose message passing is done by two GraphSAGE layers. The meta-graph is the mechanism that lets each molecule's prediction be informed by its similarity to every other molecule in the dataset.

What would settle it

Rebuild the meta-graph using only training molecules, assign molecules to train and test by scaffold clusters rather than at random, and rerun both MOLG3-SAGE and the LightGBM-plus-GIN pipeline; if the AUCs stay at roughly 0.9812 and 0.9622 while the CNN baseline stays at 0.954, the claim is supported, and if the gap narrows or inverts, the reported edge depends on split leakage and near-duplicate molecules.

Watch

Extended reading notes

Core claim

The central discovery claim is that multiscale wavelet scattering, applied in two versions (a 2D wavelet scattering transform on molecule images and a geometric scattering transform on molecular graphs), captures the structural and spatial information most relevant to mutagenicity better than previous representations. The paper reports that geometric scattering features alone, used with a simple classifier, already rival a CNN model (AUC 0.9507 versus 0.954), and that adding GIN embeddings pushes LightGBM to AUC 0.9812, above all listed baselines. It also claims that organizing the whole dataset as a fully connected meta-graph, in which each molecule is a node and edge weights are Gaussian kernels on cosine similarity of scattering embeddings, and then running GraphSAGE on that meta-graph, yields the best accuracy (0.9301) and MCC (0.8603). The paper frames this as evidence that scattering-based molecular representations set a new performance level for Ames mutagenicity prediction.

Load-bearing premise

The entire comparison assumes the test molecules never leak into the fully connected meta-graph used to train MOLG3-SAGE, and that an 80/20 random split is a fair approximation of external validation.

Editorial extensions

If this is right

  • If the reported results are correct, scattering features offer a training-free, multiscale alternative to learned molecular fingerprints, with simple gradient-boosted trees matching or exceeding a purpose-built CNN.
  • MOLG3-SAGE's fully connected meta-graph implies that molecule-to-molecule similarity, not just per-molecule structure, carries predictive signal for mutagenicity; retaining that relational information is what the paper credits for its best accuracy and MCC.
  • The 2D scattering result implies that molecular images, when transformed rather than fed raw to a CNN, can outperform handcrafted descriptors, suggesting image-based screening pipelines need not abandon interpretable feature engineering.
  • Because the scattering transforms are not mutagenicity-specific, the same pipeline is directly portable to other molecular property and toxicity endpoints, subject to revalidation.

Reading between the lines

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

  • Editorial inference: the most informative follow-up test, not reported in the paper, is a scaffold-split evaluation in which the meta-graph is rebuilt from training molecules only; this would separate genuine generalization from reliance on near-duplicate molecules between train and test.
  • Editorial inference: since both 2D and geometric scattering are parameter-light and training-free, the likely practical value is in small-data chemistry settings where deep models overfit; a useful extension would be measuring how the AUC gap changes as training set size shrinks.
  • Editorial inference: the 0.9812 AUC may approach the ceiling of this benchmark, so the next meaningful comparison is against recent consensus and transformer-based predictors on larger external mutagenicity sets, and on time-split or scaffold-split validation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes scattering-transform-based representations for Ames mutagenicity prediction, combining 2D image scattering, geometric graph scattering (GGS), GIN embeddings, and a fully connected meta-graph GraphSAGE model (MOLG3-SAGE). Using the Hansen et al. benchmark (6,277 compounds after preprocessing and a random 80/20 split), it reports AUC 0.9812 for Lightgbm with GGS+GIN features and ACC 0.9301/AUC 0.9622 for MOLG3-SAGE, and claims state-of-the-art performance against published baselines. The central claim is empirical: scattering features improve mutagenicity screening over descriptors and over published CNN/GNN models.

Significance. Scattering transforms are a plausible and relatively unexplored representation for molecular property prediction, and the paper's claims are concrete and falsifiable: the reported AUCs could in principle be reproduced or refuted with the same data. If the results survive a clean, train-only evaluation protocol, the contribution would be useful to computational toxicology. However, as presented, the evaluation does not rule out test information entering feature construction and meta-graph message passing, and the baseline comparisons are not apples-to-apples. The headline numbers are therefore not currently trustworthy, and the significance of the work hinges on a re-run under a rigorous protocol.

major comments (4)
  1. [Section 4.3, Eqs. (15)-(18)] The meta-graph is defined over "n molecules" without any statement that test molecules are excluded from the node set and from the edge-weight construction, including the Gaussian-kernel normalization sigma. If test nodes and their pairwise distances participate in GraphSAGE message passing and in the global normalization, the evaluation is transductive, not the inductive comparison made against published models in Table 4. This directly affects the reported MOLG3-SAGE ACC 0.9301 and AUC 0.9622. Please specify and implement a train-only meta-graph construction, with sigma computed from training pairs only, and re-report all affected results.
  2. [Section 4.2.2 and Section 5] The Chi-squared feature selection on the 11,681-dimensional 2D scattering representation and the generation of GIN embeddings are not described as restricted to the training split. If feature selection or GIN training uses all 6,277 molecules or the test labels, then the Lightgbm GGS+GIN AUC 0.9812 and the 2D-scattering AUC 0.9128 are inflated by information leakage. The manuscript must state whether the GIN is supervised by mutagenicity labels, on which split it is trained, and whether the 4000-feature subset is selected on the training portion only; the experiments should then be rerun under that strict protocol.
  3. [Tables 3 and 4] The comparison to published baselines is not apples-to-apples. Baseline AUC values in Table 4 come from papers with different preprocessing of the Hansen data, different validation schemes (often cross-validation), and possibly different class distributions, whereas this paper uses a single random 80/20 split after its own filtering to 6,277 compounds. A higher AUC than a published number under different protocols does not establish superiority. Baselines should be rerun under the same split and metrics, or standard benchmark splits (e.g., scaffold split) should be used for all models.
  4. [Table 3] The central results are reported as point estimates from a single random split, with no standard deviations, no number of seeds, and no code. Given that the best reported AUC (0.9812) is close to other values in the same table (0.9796, 0.9748), the state-of-the-art claim is not statistically supported. Please report means and variances over multiple random splits, and ideally scaffold splits, and make code, seeds, and the exact data split available.
minor comments (7)
  1. [Abstract and Section 2] The text repeatedly refers to the "ZINC dataset" and "zinc toxicity," but the experiments use the Hansen et al. Ames mutagenicity benchmark; these terms should be corrected throughout.
  2. [Section 4.2.1 and Table 1] The text says the tight Hann wavelet transform is used with three scales (j = 3), while Table 1 lists the number of tight Hann scales as J = 4; the text and table should be aligned, and the meaning of "No. Tight HANN coefficients 7" should be explained.
  3. [Section 5] The paper says hyperparameters were optimized with 10-fold cross-validation on the training portion, and Table 3 is labeled "on a test set"; please clarify whether all Table 3 numbers come from the held-out 20% and, if so, state that the test set was used only once after model selection.
  4. [Section 3.2, Eq. (7)] Equation (7) contains an incomplete or malformed trigonometric expression, and the variables R and e_max are not defined in the text; please fix the equation and define all symbols.
  5. [Section 5] The description of the GIN architecture is unclear: it says three sequential graph isomorphism layers, then mentions input dimension 7 and "64 hidden layers," followed by fully connected layers yielding 128-dimensional embeddings. Please give the exact layer counts and dimensions.
  6. [References] Reference [28] is not the original Graph Isomorphism Network paper; the GIN method should be cited to Xu et al., "How Powerful are Graph Neural Networks?" (ICLR 2019).
  7. [Section 5] The descriptor name "EPCF6" appears to be a typo for "ECFP6."

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmark study with no self-citation chain and no definitional reduction of predictions to inputs.

full rationale

The paper is an empirical modeling paper, not a derivation. Its central claims (Lightgbm with GGS+GIN embeddings reaching AUC 0.9812, and MOLG3-SAGE reaching ACC 0.9301) are measured on a held-out 20% test split, not derived from the feature definitions. The scattering transforms (2D WST and geometric graph scattering) are defined independently of the mutagenicity label in Eqs. (1)-(3), (5)-(11), and the machine learning and GNN models are trained rather than algebraically forced. The meta-graph edge weights in Eqs. (15)-(18) reuse the scattering embeddings that also serve as node features; this is feature reuse within a model, not a circular definition, because mutagenicity is not defined in terms of those embeddings. The paper contains no self-citations that carry the argument: citations [20]-[25] are external mathematical and methodological sources for wavelets and graph scattering. The main weaknesses that can be inferred from the text--ambiguous train/test separation for feature selection, the potentially transductive MOLG3-SAGE meta-graph, and the random 80/20 split without scaffold separation--are evaluation-protocol concerns, not circularity in the sense of a fitted parameter renamed as a prediction or an equation reducing to its own input. No circular step can be exhibited with the required quote and explicit reduction, so the appropriate score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The reported improvements rest on several domain assumptions: scattering features encode mutagenicity-relevant structure, cosine similarity between scattering embeddings is a valid chemical similarity, and the meta-graph evaluation does not leak label information. The paper also introduces many fitted hyperparameters, including wavelet scales, orientations, layer counts, and the feature subset size, all selected by experimentation on the same benchmark without nested validation.

free parameters (6)
  • Tight Hann wavelet scale count = 4 in Table 1; text says j=3 gave best results
    Section 4.2.1 says the authors found three scales (j=3) provided the best results, but Table 1 lists J=4; selected by experimentation on this dataset.
  • Diffusion wavelet layer count = 3
    Table 1 sets three diffusion wavelet layers; chosen heuristically for molecular graphs.
  • 2D WST scales and orientations = J=9, L=8
    Section 4.2.2 states these were chosen after extensive experimentation to represent local and global structures.
  • 2D scattering feature subset size = 4000
    Section 4.2.2 selected 4000 of 11,681 Chi-squared selected features among tested sizes 1000, 2000, 4000, 6000, and 8000.
  • Gaussian diffusion kernel sigma = standard deviation of the distance matrix
    Equation 18 sets sigma from the data, which controls the edge weights in the MOLG3-SAGE meta-graph.
  • MOLG3-SAGE hidden and embedding dimensions = hidden 128, embedding 64
    Table 2 lists these architecture hyperparameters with no sensitivity analysis.
assumptions (5)
  • standard math The wavelet scattering transform theory (Mallat) provides stable, informative representations.
    Sections 3.1 and 3.2 rely on established scattering theory without proving it.
  • domain assumption Scattering coefficients preserve the chemical structural information needed for mutagenicity prediction.
    The whole method assumes these features are more predictive than conventional descriptors; no biological or chemical mechanism is offered.
  • domain assumption Cosine similarity between scattering embeddings is a valid molecular similarity measure for the meta-graph edges.
    Equations 15-18 construct the fully connected graph from this similarity without independent validation.
  • domain assumption Hansen et al. labels, after the clear-evidence rule, are reliable ground truth.
    Section 4.1 resolves conflicting Ames outcomes always toward the positive call, assuming the positive outcome is correct.
  • ad hoc to paper Test molecules are not part of the meta-graph used to train MOLG3-SAGE.
    A fair inductive evaluation requires this, but Section 4.3 describes the graph over all molecules without specifying how the split interacts with graph construction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GeoScatt-GNN: A Geometric Scattering Transform-Based Graph Neural Network Model for Ames Mutagenicity Prediction." pith.science (2026). https://pith.science/paper/XFCUAHYL

@misc{pith2026241115331,
  author       = {Pith},
  title        = {Pith review of: GeoScatt-GNN: A Geometric Scattering Transform-Based Graph Neural Network Model for Ames Mutagenicity Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XFCUAHYL}},
  note         = {Machine review of arXiv:2411.15331}
}
read the original abstract

This paper tackles the pressing challenge of mutagenicity prediction by introducing three ground-breaking approaches. First, it showcases the superior performance of 2D scattering coefficients extracted from molecular images, compared to traditional molecular descriptors. Second, it presents a hybrid approach that combines geometric graph scattering (GGS), Graph Isomorphism Networks (GIN), and machine learning models, achieving strong results in mutagenicity prediction. Third, it introduces a novel graph neural network architecture, MOLG3-SAGE, which integrates GGS node features into a fully connected graph structure, delivering outstanding predictive accuracy. Experimental results on the ZINC dataset demonstrate significant improvements, emphasizing the effectiveness of blending 2D and geometric scattering techniques with graph neural networks. This study illustrates the potential of GNNs and GGS for mutagenicity prediction, with broad implications for drug discovery and chemical safety assessment.

Figures

Figures reproduced from arXiv: 2411.15331 by the authors.

Figure 1
Figure 1. 2D Morlet Wavelet visualization at various orientations and parameters 3.1.2. Scattering Transform Let X be the signal to be analysed. A WST is implemented with a deep convolution network that iterates over traditional wavelet transform, nonlinear modulus, and averaging operators. The zeroth-order scattering coefficient is obtained by averaging the input signal with a low-pass filter ϕJ: S 0 x = I ∗ ϕJ, (1) where ϕJ… view at source ↗
Figure 2
Figure 2. Graph Scattering Transform with J = 3 and L = 3 for multiscale signal decomposition. 3.3. Graph Neural Network Graph Neural Networks [26] are deep learning models that analyze graph-structured data. They work by iteratively updating node representations via message transmission among surrounding nodes. The basic premise is that each node’s attributes are updated using both its own qualities and aggregated informatio… view at source ↗
Figure 3
Figure 3. Multi-Modal Pipeline for Mutagenicity Prediction Using Molecule Representations as Graphs and Images. The pipeline utilizes the Hansen et al. dataset of 6,512 compounds divided into mutagens and non-mutagens. Molecules are transformed into graph representations for geometric scattering (a) using Diffusion and Tight HANN wavelets, along with embeddings from a Graph Isomorphism Network (GIN), and into 2D molecular ima… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Fully Connected Layer Architecture for Binary Mutagenicity Classification in MolG³-SAGE Framework Our method uses a message-passing framework in which each node accumulates information from its neighbors using a mean-pooling process. The aggregated neighborhood feature…
Figure 5
Figure 5. Figure 5: Comparative Performance Analysis of Machine Learning Models for Ames Mutagenecity Using Multiple Evalu￾ation Metrics demonstrates the benefits of 2D scattering for molecular image analysis, as the scattering features can capture finer structural details, spatial relati…
Figure 6
Figure 6. Figure 6: ROC Curves for Mutagenicity Prediction Using GGS-GIN Embeddings: Performance Comparison of Random￾Forest, SVC, XGBoost, and Lightgbm Models Across 10-Fold Cross-Validation setting a new standard in the field. These findings demonstrate the efficacy of our multi-perspec…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 26 canonical work pages

  1. [1]

    B. N. Ames, W. E. Durston, E. Yamasaki, F. D. Lee, Carcinogens are mutagens: a simple test system combining liver homogenates for activation and bacteria for detection, Proceedings of the National Academy of Sciences 70 (1973) 2281–2285

    work page 1973

  2. [2]

    Honma, A

    M. Honma, A. Kitazawa, A. Cayley, R. V . Williams, C. Barber, T. Hanser, R. Saiakhov, S. Chakravarti, G. J. Myatt, K. P. Cross, et al., Improvement of quantitative structure–activity relationship (qsar) tools for predicting ames mutagenicity: outcomes of the ames/qsar inter- national challenge project, Mutagenesis 34 (2019) 3–16

    work page 2019

  3. [3]

    S. V . Feeney, R. Lui, D. Guan, S. Matthews, Multiple instance learning improves ames mutagenicity prediction for problematic molecular species, Chemical Research in Toxicology 36 (2023) 1227–1237

    work page 2023

  4. [4]

    T. Li, Z. Liu, S. Thakkar, R. Roberts, W. Tong, Deepames: A deep learning-powered ames test predictive model with potential for regulatory application, Regulatory Toxicology and Pharmacology 144 (2023) 105486

    work page 2023

  5. [5]

    S. Li, L. Zhang, H. Feng, J. Meng, D. Xie, L. Yi, I. T. Arkin, H. Liu, Mutagenpred-gcnns: a graph convolutional neural network-based classification model for mutagenicity prediction with data-driven molecular fingerprints, Interdisciplinary Sciences: Computational Life Sciences 13 (2021) 25–33

    work page 2021

  6. [6]

    T. T. Van Tran, H. Tayara, K. T. Chong, Ampred-cnn: Ames mutagenicity prediction model based on convolutional neural networks, Computers in Biology and Medicine 176 (2024) 108560

    work page 2024

  7. [7]

    Winter, F

    R. Winter, F. Montanari, F. No´e, D.-A. Clevert, Learning continuous and data-driven molecular descriptors by translating equivalent chemical representations, Chemical science 10 (2019) 1692–1701

    work page 2019

  8. [8]

    C. S. Chu, J. D. Simpson, P. M. O’Neill, N. G. Berry, Machine learning–predicting ames mutagenicity of small molecules, Journal of Molecular Graphics and Modelling 109 (2021) 108011

    work page 2021

Show all 31 references
  1. [9]

    Venkatraman, Fp-admet: a compendium of fingerprint-based admet prediction models, Journal of cheminformatics 13 (2021) 1–12

    V . Venkatraman, Fp-admet: a compendium of fingerprint-based admet prediction models, Journal of cheminformatics 13 (2021) 1–12

    2021

  2. [10]

    N. K. Shinada, N. Koyama, M. Ikemori, T. Nishioka, S. Hitaoka, A. Hakura, S. Asakura, Y . Matsuoka, S. K. Palaniappan, Optimizing machine-learning models for mutagenicity prediction through better feature selection, Mutagenesis 37 (2022) 191–202

    2022

  3. [11]

    C. Lou, H. Yang, H. Deng, M. Huang, W. Li, G. Liu, P. W. Lee, Y . Tang, Chemical rules for optimization of chemical mutagenicity via matched molecular pairs analysis and machine learning methods, Journal of Cheminformatics 15 (2023) 35

    2023

  4. [12]

    Kumar, F

    R. Kumar, F. U. Khan, A. Sharma, M. H. Siddiqui, I. B. Aziz, M. A. Kamal, G. M. Ashraf, B. S. Alghamdi, M. S. Uddin, A deep neural network–based approach for prediction of mutagenicity of compounds, Environmental Science and Pollution Research 28 (2021) 47641– 47650

    2021

  5. [13]

    R. Lui, D. Guan, S. Matthews, Mechanistic task groupings enhance multitask deep learning of strain-specific ames mutagenicity, Chemical Research in Toxicology 36 (2023) 1248–1254

    2023

  6. [14]

    Xiong, Z

    G. Xiong, Z. Wu, J. Yi, L. Fu, Z. Yang, C. Hsieh, M. Yin, X. Zeng, C. Wu, A. Lu, et al., Admetlab 2.0: an integrated online platform for accurate and comprehensive predictions of admet properties, Nucleic acids research 49 (2021) W5–W14

    2021

  7. [15]

    C. Hung, G. Gini, Qsar modeling without descriptors using graph convolutional neural networks: the case of mutagenicity prediction, Molecular diversity 25 (2021) 1283–1299

    2021

  8. [16]

    J. Guo, Q. Liu, H. Guo, X. Lu, Ligandformer: A graph neural network for predicting compound property with robust interpretation, arXiv preprint arXiv:2202.10873 (2022)

    2022 arXiv

  9. [17]

    Y . Wei, S. Li, Z. Li, Z. Wan, J. Lin, Interpretable-admet: a web service for admet prediction and optimization based on deep neural representation, Bioinformatics 38 (2022) 2863–2871

    2022

  10. [18]

    J. M. Lilly, S. C. Olhede, Higher-order properties of analytic wavelets, IEEE Transactions on Signal Processing 57 (2008) 146–160

    2008

  11. [19]

    Mallat, A wavelet tour of signal processing, Third edition, Academic Press (2008)

    S. Mallat, A wavelet tour of signal processing, Third edition, Academic Press (2008)

    2008

  12. [20]

    Mallat, Group invariant scattering, Communications on Pure and Applied Mathematics 65 (2012) 1331–1398

    S. Mallat, Group invariant scattering, Communications on Pure and Applied Mathematics 65 (2012) 1331–1398

    2012

  13. [21]

    F. Gama, A. Ribeiro, J. Bruna, Diffusion scattering transforms on graphs, arXiv preprint arXiv:1806.08829 (2018)

    2018 arXiv

  14. [22]

    D. Zou, G. Lerman, Graph convolutional neural networks via scattering, Applied and Computational Harmonic Analysis 49 (2020) 1046– 1074

    2020

  15. [23]

    R. R. Coifman, M. Maggioni, Diffusion wavelets, Applied and computational harmonic analysis 21 (2006) 53–94

    2006

  16. [24]

    Hansen, S

    K. Hansen, S. Mika, T. Schroeter, A. Sutter, A. Ter Laak, T. Steger-Hartmann, N. Heinrich, K.-R. Muller, Benchmark data set for in silico prediction of ames mutagenicity, Journal of chemical information and modeling 49 (2009) 2077–2081

    2009

  17. [25]

    Jiang, Z

    Q. Jiang, Z. Shen, Tight wavelet frames in low dimensions with canonical filters, Journal of Approximation Theory 196 (2015) 55–78

    2015

  18. [26]

    Scarselli, M

    F. Scarselli, M. Gori, A. C. Tsoi, M. Hagenbuchner, G. Monfardini, The graph neural network model, IEEE transactions on neural networks 20 (2008) 61–80

    2008

  19. [27]

    Hamilton, Z

    W. Hamilton, Z. Ying, J. Leskovec, Inductive representation learning on large graphs, Advances in neural information processing systems 30 (2017)

    2017

  20. [28]

    B.-H. Kim, J. C. Ye, Understanding graph isomorphism network for rs-fmri functional connectivity analysis, Frontiers in neuroscience 14 (2020) 630

    2020

  21. [29]

    Banerjee, A

    P. Banerjee, A. O. Eckert, A. K. Schrey, R. Preissner, Protox-ii: a webserver for the prediction of toxicity of chemicals, Nucleic acids research 46 (2018) W257–W263

    2018

  22. [30]

    Karim, A

    A. Karim, A. Mishra, M. H. Newton, A. Sattar, Efficient toxicity prediction via simple features using shallow neural networks and decision trees, Acs Omega 4 (2019) 1874–1888

    2019

  23. [31]

    J. Rao, S. Zheng, Y . Lu, Y . Yang, Quantitative evaluation of explainable graph neural networks for molecular property prediction, Patterns 3 (2022). 18

    2022

Pith tools

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