Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

A Multi-view Divergence-Convergence Feature Augmentation Framework for Drug-related Microbes Prediction

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read The paper claims that its multi-view divergence-convergence framework, DCFA_DMP, outperforms existing state-of-the-art methods for predicting drug-microbe associations on the MDAD benchmark, reporting AUROC of 0.9894 and AUPR of 0.9856…

desk verdict Strong empirical claims and a sensible architecture, but the training objective is undefined and the tuning appears to use the test set, so the SOTA result cannot be verified as written. read the letter →

arxiv 2506.18797 v1 pith:D6F7QBDO submitted 2025-06-23 cs.LG

classification cs.LG
keywords drug-microbeassociationpredictionmulti-viewlearningadversarialgraphneuralnetworkstransformerattentionmechanismheterogeneousMDAD
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

Drug-microbe associations matter because they influence how well a drug works and how it is metabolized; knowing them can guide drug development and precision medicine. The paper proposes DCFA_DMP, a deep-learning framework that predicts such associations by deliberately separating, then recombining, two views of the data: the known drug-microbe association network and the drug-drug and microbe-microbe similarity graphs. In its divergence phase, adversarial learning pushes the two views' feature representations apart to preserve their distinct information; in the convergence phase, a bidirectional synergistic attention mechanism blends them into a fused representation, while transformer and graph-convolution layers alternately refine node embeddings on the heterogeneous graph. On the MDAD benchmark, the paper reports that DCFA_DMP surpasses four existing methods on AUROC (0.9894), AUPR (0.9856), and F1, with statistically significant differences, and maintains an AUROC around 0.95-0.96 for new drugs and microbes in cold-start tests.

What carries the argument

The load-bearing mechanism is the divergence-convergence feature augmentation strategy. In the divergence phase, adversarial learning maximizes the Euclidean distance between the association-view and similarity-view embeddings against a margin γ, explicitly preserving and even enlarging the differences between heterogeneous and similarity information. In the convergence phase, the Bidirectional Synergistic Attention Mechanism (BSAM) projects both views through tanh layers, computes two view-specific compatibility scores over the concatenated projections, and forms a fused vector by attention-weighted summation, so view-specific patterns are kept while complementary information is combined. The surrounding mechanism is the alternating Transformer-GNN module: the Transformer expands which nodes each drug or microbe attends to, the GNN aggregates local structure, and a message-passing sample-update layer refreshes the attention samples. Together these components produce the fused features fed to a three-layer MLP that outputs the association score.

What would settle it

Re-run DCFA_DMP on the same MDAD data and splits but select all hyperparameters on a held-out validation fold and evaluate the untouched test fold once; if AUROC and AUPR do not remain above the reported baseline values (for example, NGMDA's AUROC of 0.9761), the claimed state-of-the-art advantage fails.

Watch

Extended reading notes

Core claim

DCFA_DMP is a three-stage pipeline for drug-microbe association prediction: multi-view graph representation learning, a divergence-convergence feature enhancement strategy, and a classifier. The central discovery claim is that treating the association view and the similarity views as competing rather than redundant sources of information improves prediction. The divergence phase uses an adversarial objective with a margin γ to maximize the distance between drug (and microbe) features from the association view and the similarity view; the convergence phase then uses the Bidirectional Synergistic Attention Mechanism (BSAM) to compute view-specific compatibility scores and aggregate the two views through normalized attention weights. Transformer layers widen the receptive field of the GNN on the drug-microbe heterogeneous graph, and GNN message passing updates the attention samples. The paper reports that on MDAD the full model reaches AUROC 0.9894 ± 0.0063 and AUPR 0.9856 ± 0.0162, beating the compared methods SCSMDA, NGMDA, DHDMP, and GACNNMDA, and that ablations show each component (transformer, adversarial learning, BSAM) contributes to the gain. The method also performs well in cold-start settings: AUROC ≈ 0.95 for entirely new microbes and ≈ 0.96 for entirely new drugs.

Load-bearing premise

The headline numbers assume the model's hyperparameters were chosen without using the test set, yet the paper describes tuning them across the full dataset and reports no validation split, so the performance margins may be optimistically biased.

Editorial extensions

If this is right

  • If the central claim holds, DCFA_DMP gives the best published prediction quality for drug-microbe associations on MDAD, with AUROC above 0.98 and AUPR above 0.98.
  • The divergence-convergence recipe is validated as a general fusion strategy: ablations show that removing the adversarial divergence or the BSAM attention, or replacing BSAM with additive, multiplicative, concatenation, or cross-attention fusion, all degrade performance.
  • The alternating Transformer-GNN structure is necessary to the gain: removing either layer or substituting standard attention for the Transformer and GCN for the GNN lowers the metrics.
  • Cold-start results imply that the model can give useful association scores for drugs and microbes with no prior association records, which is the practical scenario for discovering new therapies.
  • The method can be applied to other bio-information prediction tasks, such as RNA-disease association prediction, as the authors state in the conclusion.

Reading between the lines

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

  • A reader should treat the reported gains as possibly optimistic because hyperparameters (β1, β2, neighbor count N, learning rate, dropout, epochs) were chosen across all datasets with no described validation split, so part of the margin over baselines could come from selection on the test set.
  • Because the similarity matrices and association data are taken from the NGMDA paper, the comparison with NGMDA tests the new fusion and learning head more than the input features; re-running all baselines on identical input features would isolate the framework's contribution.
  • The divergence-convergence idea is a general recipe: the same two-phase push-apart-then-fuse strategy could be tested on other link-prediction tasks on heterogeneous biological networks, such as drug-target or RNA-disease prediction, and would be a direct test of whether the mechanism transfers.
  • The case study's external validation rests on one drug, E3D2o; confirming top-ranked predictions with in vitro or clinical evidence would test the practical claim more severely than literature matching.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes DCFA_DMP, a multi-view divergence-convergence framework for drug-microbe association prediction. The method combines GCN-based KNN graph representation learning, a Transformer-GNN module, an adversarial divergence phase that pushes association-view and similarity-view features apart, a Bidirectional Synergistic Attention Mechanism (BSAM) for convergence, and an MLP classifier. Experiments on the MDAD benchmark report state-of-the-art AUROC (0.9894) and AUPR (0.9856) against four baselines, plus ablation studies, parameter sensitivity analysis, cold-start experiments, and a COVID-19 case study. The central claim is that DCFA_DMP is a new SOTA predictor for drug-microbe associations.

Significance. If the reported results hold, DCFA_DMP would be a meaningful advance in drug-microbe association prediction, combining several components (graph Transformer with GNN, adversarial view separation, and bidirectional attention fusion) in a way not previously packaged together. The cold-start experiments and case study add practical relevance. The paper does not provide code or machine-checked proofs, so the contribution rests entirely on the reproducibility of the reported numbers. The central SOTA claim is currently undercut by an invalid loss equation and undefined loss terms, which prevent independent verification of the trained model. With those issues corrected and a clear validation protocol supplied, the work could be a solid empirical contribution.

major comments (3)
  1. [III.C, Eq. (19)] The binary cross-entropy loss is written incorrectly. The term for negative samples should be (1 - y_ij) log(1 - sigma(z_i)), not (1 - y_ij) log(sigma(z_i)). As written, both the positive and negative terms push sigma(z_i) toward 1, which is not a valid classification objective. In addition, the subscript in z_i is inconsistent with the pair index (i,j); the logit should be a function of the drug-microbe pair, presumably z_ij, and should be obtained from the MLP output in Eq. (18). This error makes the primary training objective unusable as specified.
  2. [III.C, Eq. (20)] The overall loss L_total = L_rel + beta1 * L_con,drug + beta2 * L_con,mico references L_con,drug and L_con,mico, but these terms are never defined in the paper. The only divergence losses introduced in Section III.B.1 are L_adv,drug and L_adv,microbe (Eqs. 11 and 12), which are hinge-style separation losses, not contrastive losses. If the intended losses are indeed L_adv,drug and L_adv,microbe, the notation is misleading; if separate contrastive losses exist, their definitions are omitted. Either way, the exact objective minimized to produce the results in Table I cannot be reconstructed, blocking reproduction.
  3. [IV.C, Section IV.A] The hyperparameters beta1, beta2, N, learning rate, dropout, and number of epochs are selected by empirical tuning, with the text stating that values were chosen to maximize performance on the evaluation metrics, but no validation split or nested cross-validation procedure is described. Because the same test set is used for both parameter selection and final performance reporting, the numbers in Table I may be optimistically biased by selection on the test data. A clear separation between training, validation, and test sets, or an explicit nested-cross-validation protocol, is needed to support the claimed significant outperformance over baselines.
minor comments (5)
  1. [II] The heading 'RELATED MATERIALS' is followed by an empty section; either the section should be removed or filled with a proper related-work discussion.
  2. [III.A.2] The phrase 'reducing in the discriminative power of node representations' should be 'reducing the discriminative power'.
  3. [III.C, Eq. (20)] The subscript 'mico' is a typo and should be 'microbe'.
  4. [I, Contributions] The phrase 'prove the practicality of our model in drug localization tasks' appears to be a typo; the case study is about drug-microbe association prediction for COVID-19, not drug localization.
  5. [IV.C] The text says parameters were tuned 'across all datasets' but only the MDAD dataset is used in the experiments; this should be clarified.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported test-set SOTA is partly constructed by hyperparameter tuning on the test set; the model's central prediction claim is otherwise not circular.

  1. fitted input called prediction [Section IV.C, Parameter sensitivity analysis]
    "Through empirical tuning of these parameters, we set a = 4000, lr = 0.005, and dropout = 0.5 for all experiments. ... The values of the trade-off parameters β1 and β2 were chosen from {0.01, 0.02, 0.03 }, and all possible pairwise combinations were tested. Figure 4 shows the results of various metrics under different trade-off parameters. Considering the performance across five metrics, the optimal overall results were achieved when both β1 and β2 were set to 0.03."

    The dataset is split once into training and test sets (Section IV: 'The complete dataset was randomly split into a training set and a testing set with the testing set representing 10% of the total data'), and no validation split is described. Section IV.C then tunes learning rate, epoch count, dropout, β1, β2, and neighbor count N by evaluating performance metrics on the same data used to produce the reported Table I numbers. The reported test AUROC and AUPR are therefore not out-of-sample predictions from a fixed model; they are the outcome of selecting hyperparameters on the very test set used for the SOTA claim. The 'prediction' is statistically forced by the tuning procedure, so the central SOTA claim partially reduces to a fit on the test set.

full rationale

The paper's architecture and feature-enhancement derivation are not circular: the GCN, Transformer, adversarial divergence, and BSAM fusion are stated as trainable modules whose output scores are empirical predictions, not quantities equivalent to the inputs by construction. No load-bearing self-citation chain or imported uniqueness theorem appears. The one concrete circularity is evaluation circularity: hyperparameters are tuned by inspecting performance metrics without a validation split, so the test-set numbers in Table I are partly constructed by the tuning process. Separately, the training objective is internally inconsistent as written: Eq. (19) places log(sigmoid(z_i)) in both the positive and negative terms, and Eq. (20) references L_con,drug and L_con,mico, which are never defined in the paper. That is a reproducibility and correctness defect rather than a circular-derivation defect, so it does not by itself raise the circularity score, but it compounds the difficulty of verifying the SOTA claim.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a set of hyperparameters selected empirically and a standard supervised learning setup. No new entities are introduced. The main unstated assumptions are the negative sampling strategy and the representativeness of the single 10% test split.

free parameters (8)
  • beta1 = 0.03
    Trade-off weight for drug adversarial loss, chosen from {0.01,0.02,0.03} based on test performance.
  • beta2 = 0.03
    Trade-off weight for microbe adversarial loss, chosen from {0.01,0.02,0.03} based on test performance.
  • N (number of neighbors in KNN graph) = 8
    Number of neighbors in the similarity KNN graph, chosen from {1,4,8,12,16} based on test performance.
  • lr = 0.005
    Learning rate, set for all experiments without sensitivity analysis.
  • dropout = 0.5
    Dropout rate, set for all experiments without sensitivity analysis.
  • epochs = 4000
    Number of training epochs, set for all experiments.
  • gamma (adversarial separation margin)
    Margin in the adversarial hinge loss; value not reported in the paper.
  • embedding dimensions and number of layers
    Not reported; required to reproduce the architecture.
assumptions (4)
  • domain assumption Unknown drug-microbe pairs are treated as negative samples.
    Standard in link prediction but not stated in the paper; the negative sampling ratio is not given.
  • domain assumption The similarity matrices and KNN graph construction from [14] are correct and complete.
    The paper uses drug-drug and microbe-microbe similarities derived from NGMDA [14] without re-verification.
  • standard math GCN and multi-head attention are standard and correctly implemented.
    The paper relies on standard GCN, Transformer, and attention formulas.
  • domain assumption The random 10% test split is representative of the data distribution.
    Only one random split is described; no cross-validation or multiple seeds are used for the main result.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Multi-view Divergence-Convergence Feature Augmentation Framework for Drug-related Microbes Prediction." pith.science (2026). https://pith.science/paper/D6F7QBDO

@misc{pith2026250618797,
  author       = {Pith},
  title        = {Pith review of: A Multi-view Divergence-Convergence Feature Augmentation Framework for Drug-related Microbes Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D6F7QBDO}},
  note         = {Machine review of arXiv:2506.18797}
}
read the original abstract

In the study of drug function and precision medicine, identifying new drug-microbe associations is crucial. However, current methods isolate association and similarity analysis of drug and microbe, lacking effective inter-view optimization and coordinated multi-view feature fusion. In our study, a multi-view Divergence-Convergence Feature Augmentation framework for Drug-related Microbes Prediction (DCFA_DMP) is proposed, to better learn and integrate association information and similarity information. In the divergence phase, DCFA_DMP strengthens the complementarity and diversity between heterogeneous information and similarity information by performing Adversarial Learning method between the association network view and different similarity views, optimizing the feature space. In the convergence phase, a novel Bidirectional Synergistic Attention Mechanism is proposed to deeply synergize the complementary features between different views, achieving a deep fusion of the feature space. Moreover, Transformer graph learning is alternately applied on the drug-microbe heterogeneous graph, enabling each drug or microbe node to focus on the most relevant nodes. Numerous experiments demonstrate DCFA_DMP's significant performance in predicting drug-microbe associations. It also proves effectiveness in predicting associations for new drugs and microbes in cold start experiments, further confirming its stability and reliability in predicting potential drug-microbe associations.

Figures

Figures reproduced from arXiv: 2506.18797 by the authors.

Figure 1
Figure 1. Workflow of DCFA DMP: (1) Multi-view graph representation learning; (2) Divergence-Convergence feature enhancement strategy; (3) Drug-microbe association prediction; power of node representations and failing to effectively capture the intricate features of drug-microbe heterogeneous graphs. Although Transformer capture long-range dependencies, they typically neglect certain structural information in complex graph-st… view at source ↗
Figure 2
Figure 2. Ablation results for (A) Transformer-based Graph; (B) Adversarial Learning method. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Ablation results for Bidirectional Synergistic Attention Mechanism. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Parameter analysis for different trade-off parameters ( [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Parameter analysis for different number of neighboring nodes. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: (A) Results of performance metrics under cold start condition. (B) The top 25% of potential drugs predicted by DCFA [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. SOC-DGL: Social Interaction Behavior Inspired Dual Graph Learning Framework for Drug-Target Interaction Identification

    cs.LG 2025-06 reject novelty 4.0 of 10

    SOC-DGL combines direct and even-hop graph similarity signals with a reweighted loss to predict drug-target interactions, reporting top results in benchmark comparisons.

Reference graph

Works this paper leans on

29 extracted references · 28 canonical work pages · cited by 1 Pith paper

  1. [1]

    Chapter 12: Human microbiome analysis

    Xochitl C Morgan and Curtis Huttenhower. Chapter 12: Human microbiome analysis. PLoS computational biology, 8(12):e1002808, 2012

  2. [2]

    The healthy human microbiome

    Jason Lloyd-Price, Galeb Abu-Ali, and Curtis Hutten- hower. The healthy human microbiome. Genome medicine, 8:1–11, 2016

  3. [3]

    Mapping human microbiome drug metabolism by gut bacteria and their genes

    Michael Zimmermann, Maria Zimmermann-Kogadeeva, Rebekka Wegmann, and Andrew L Goodman. Mapping human microbiome drug metabolism by gut bacteria and their genes. Nature, 570(7762):462–467, 2019

  4. [4]

    What are the consequences of the disappearing human microbiota? Nature Reviews Microbiology, 7(12):887–894, 2009

    Martin J Blaser and Stanley Falkow. What are the consequences of the disappearing human microbiota? Nature Reviews Microbiology, 7(12):887–894, 2009

  5. [5]

    Semi-supervised classification with graph convolutional networks

    Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907, 2016

  6. [6]

    Predicting human microbe–drug associ- ations via graph convolutional network with conditional random field

    Yahui Long, Min Wu, Chee Keong Kwoh, Jiawei Luo, and Xiaoli Li. Predicting human microbe–drug associ- ations via graph convolutional network with conditional random field. Bioinformatics, 36(19):4918–4927, 2020

  7. [7]

    Gnaemda: microbe-drug associations prediction on graph normalized convolutional network

    Haonan Huang, Yuping Sun, Meijing Lan, Huizhe Zhang, and Guobo Xie. Gnaemda: microbe-drug associations prediction on graph normalized convolutional network. IEEE Journal of Biomedical and Health Informatics , 27 (3):1635–1643, 2023

  8. [8]

    Microbe-disease association prediction using rgcn through microbe-drug- disease network

    Yueyue Wang, Xiujuan Lei, and Yi Pan. Microbe-disease association prediction using rgcn through microbe-drug- disease network. IEEE/ACM Transactions on Compu- tational Biology and Bioinformatics , 20(6):3353–3362, 2023

Show all 29 references
  1. [9]

    Graph attention networks

    Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, Yoshua Bengio, et al. Graph attention networks. stat, 1050(20):10–48550, 2017

  2. [10]

    Ensembling graph attention networks for human microbe–drug association predic- tion

    Yahui Long, Min Wu, Yong Liu, Chee Keong Kwoh, Jiawei Luo, and Xiaoli Li. Ensembling graph attention networks for human microbe–drug association predic- tion. Bioinformatics, 36(Supplement 2):i779–i786, 2020

  3. [11]

    Mgatmda: Predicting microbe-disease associations via multi-component graph attention network

    Dayun Liu, Junyi Liu, Yi Luo, Qihua He, and Lei Deng. Mgatmda: Predicting microbe-disease associations via multi-component graph attention network. IEEE/ACM Transactions on computational biology and bioinformat- ics, 19(6):3578–3585, 2021

  4. [12]

    Heterogeneous graph attention network

    Xiao Wang, Houye Ji, Chuan Shi, Bai Wang, Yanfang Ye, Peng Cui, and Philip S Yu. Heterogeneous graph attention network. In The world wide web conference , pages 2022–2032, 2019

  5. [13]

    Mdmd: a computational model for predicting drug- related microbes based on the aggregated metapaths from a heterogeneous network

    Jiajie Xing, Yuan Zhang, Jiaxuan Wang, and Juan Wang. Mdmd: a computational model for predicting drug- related microbes based on the aggregated metapaths from a heterogeneous network. In 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) , pages 6306–6...

  6. [14]

    Multi-scale topology and position feature learning and relationship- aware graph reasoning for prediction of drug-related microbes

    Ping Xuan, Jing Gu, Hui Cui, Shuai Wang, Nakaguchi Toshiya, Cheng Liu, and Tiangang Zhang. Multi-scale topology and position feature learning and relationship- aware graph reasoning for prediction of drug-related microbes. Bioinformatics, 40(2):btae025, 2024

  7. [15]

    Dynamic category- sensitive hypergraph inferring and homo-heterogeneous neighbor feature learning for drug-related microbe pre- JOURNAL OF LATEX CLASS FILES, VOL

    Ping Xuan, Zelong Xu, Hui Cui, Jing Gu, Cheng Liu, Tiangang Zhang, and Peiliang Wu. Dynamic category- sensitive hypergraph inferring and homo-heterogeneous neighbor feature learning for drug-related microbe pre- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2024 10 dict...

  8. [16]

    Mkgcn: multi-modal knowledge graph convolutional network for music recommender systems

    Xiaohui Cui, Xiaolong Qu, Dongmei Li, Yu Yang, Yuxun Li, and Xiaoping Zhang. Mkgcn: multi-modal knowledge graph convolutional network for music recommender systems. Electronics, 12(12):2688, 2023

  9. [17]

    Gacnnmda: a com- putational model for predicting potential human microbe- drug associations based on graph attention network and cnn-based classifier

    Qing Ma, Yaqin Tan, and Lei Wang. Gacnnmda: a com- putational model for predicting potential human microbe- drug associations based on graph attention network and cnn-based classifier. BMC bioinformatics , 24(1):35, 2023

  10. [18]

    Predicting of microbe-drug asso- ciations via a pre-completion-based label propagation algorithm

    Haochen Zhao, Guihua Duan, Botu Yang, Suning Li, and Jianxin Wang. Predicting of microbe-drug asso- ciations via a pre-completion-based label propagation algorithm. In 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 686–691. IEEE, 2022

  11. [19]

    A data-driven approach for predicting the impact of drugs on the human microbiome

    Yadid M Algavi and Elhanan Borenstein. A data-driven approach for predicting the impact of drugs on the human microbiome. Nature Communications, 14(1):3614, 2023

  12. [20]

    Prediction of microbe-drug associations based on katz measure

    Lingzhi Zhu, Guihua Duan, Cheng Yan, and Jianxin Wang. Prediction of microbe-drug associations based on katz measure. In 2019 IEEE international conference on bioinformatics and biomedicine (BIBM) , pages 183–187. IEEE, 2019

  13. [21]

    Association mining to iden- tify microbe drug interactions based on heterogeneous network embedding representation

    Yahui Long and Jiawei Luo. Association mining to iden- tify microbe drug interactions based on heterogeneous network embedding representation. IEEE journal of biomedical and health informatics, 25(1):266–275, 2020

  14. [22]

    Predicting microbe-drug association based on similarity and semi-supervised learning

    L Zhu, J Wang, G Li, X Hu, B Ge, and B Zhang. Predicting microbe-drug association based on similarity and semi-supervised learning. Am J Biochem Biotechnol, 17(1):50–8, 2021

  15. [23]

    Gsamda: a com- putational model for predicting potential microbe–drug associations based on graph attention network and sparse autoencoder

    Yaqin Tan, Juan Zou, Linai Kuang, Xiangyi Wang, Bin Zeng, Zhen Zhang, and Lei Wang. Gsamda: a com- putational model for predicting potential microbe–drug associations based on graph attention network and sparse autoencoder. BMC bioinformatics, 23(1):492, 2022

  16. [24]

    A survey on information bottleneck

    Shizhe Hu, Zhengzheng Lou, Xiaoqiang Yan, and Yang- dong Ye. A survey on information bottleneck. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2024

  17. [25]

    Mdad: a special resource for microbe-drug associations

    Ya-Zhou Sun, De-Hong Zhang, Shu-Bin Cai, Zhong Ming, Jian-Qiang Li, and Xing Chen. Mdad: a special resource for microbe-drug associations. Frontiers in cellular and infection microbiology , 8:424, 2018

  18. [26]

    Predicting microbe–drug associations with structure-enhanced contrastive learning and self-paced negative sampling strategy

    Zhen Tian, Yue Yu, Haichuan Fang, Weixin Xie, and Maozu Guo. Predicting microbe–drug associations with structure-enhanced contrastive learning and self-paced negative sampling strategy. Briefings in Bioinformatics , 24(2):bbac634, 2023

  19. [27]

    Multimodal evaluation of drug antibacterial activity reveals cinnamaldehyde analog anti-biofilm ef- fects against haemophilus influenzae

    Javier Asensio-L ´opez, Mar ´ıa L ´azaro-D´ıez, Tania M Hern´andez-Cruz, N ´uria Blanco-Cabra, Ioritz Sorzabal- Bellido, Eva M Arroyo-Urea, Elena Buetas, Ana Gonz´alez-Paredes, Carlos Ortiz de Sol ´orzano, Saioa Bur- gui, et al. Multimodal evaluation of drug antibacterial acti...

  20. [28]

    The vibrio cholerae type vi secretion system: toxins, regulators and consequences

    Cristian V Crisan and Brian K Hammer. The vibrio cholerae type vi secretion system: toxins, regulators and consequences. Environmental microbiology, 22(10): 4112–4122, 2020

  21. [29]

    Guili Yu, Hong Yu, Qiankun Yang, Jinxin Wang, Hui Fan, Gang Liu, Lei Wang, Babatunde Kazeem Bello, Panpan Zhao, Honggang Zhang, et al. Vibrio harveyi in- fections induce production of proinflammatory cytokines in murine peritoneal macrophages via activation of p38 mapk and nf-...

Pith tools

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