Pith. sign in

REVIEW 5 major objections 8 minor 76 references

OmniESI: A unified framework for enzyme-substrate interaction prediction with progressive conditional deep learning

T0 review · 5 major / 8 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Enzyme-substrate prediction improves when general embeddings are progressively conditioned toward catalytic features.

desk verdict A solid empirical paper with a genuinely new two-stage conditioning architecture, but its headline 'consistently superior' claim is not fully controlled because baselines are quoted from prior work without reruns. read the letter →

arxiv 2506.17963 v1 pith:PKZBZNSU submitted 2025-06-22 q-bio.BM cs.AI

Zhiwei Nie , Hongyu Zhang , Hao Jiang , Yutian Liu , Xiansong Huang , Fan Xu , Jie Fu , Zhixiang Ren
show 3 more authors
Yonghong Tian Wen-Bin Zhang Jie Chen
This is my paper
classification q-bio.BMcs.AI
keywords enzyme-substrateinteractionpredictionconditionaldeeplearningprogressivefeaturemodulationenzymekineticparametersactivesiteannotationmutationeffectESM-2graphconvolutionalnetwork
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

OmniESI sets out to establish that enzyme-substrate interactions are predicted better when the latent features of an enzyme and its substrate are progressively steered from general protein-molecule representations into a catalysis-aware representation by two conditional networks. The paper argues that existing methods either train encoders from scratch or freeze pretrained embeddings without imposing catalytic priors, so their features miss reaction-relevant residues and substrate functional groups and carry catalytically irrelevant noise. With one unified architecture and only task-specific output heads, OmniESI is reported to match or beat specialized state-of-the-art methods on seven benchmarks spanning kinetic parameter regression, enzyme-substrate pairing, active-site annotation, and mutation effect classification. The two conditioning modules together add roughly 0.16% of the parameters, and ablation studies attribute the gains to them. If the claim holds, a single parameter-efficient framework could replace a collection of task-specific enzyme prediction models.

What carries the argument

The machinery is a pair of conditional modulation modules inserted between frozen ESM-2/GCN encoders and task-specific heads. BCFM (bidirectional conditional feature modulation) uses a two-sided conditioning block that generates layer-normalization parameters from the partner's pooled representation, together with multi-scale convolutions, so that the enzyme's representation is conditioned on the substrate and vice versa. CCFM (catalysis-aware conditional feature modulation) uses the concatenated pair representation as a query to compute attention weights that highlight catalytically relevant residues and atoms. The load-bearing design choice is that the substrate conditions the enzyme representation and the enzyme conditions the substrate representation in the first module, and then the joint interaction representation conditions both sides in the second, so the latent features are gradually shifted from general protein-molecule space to catalysis-aware space.

What would settle it

Run DLKcat, UniKP, CatPred, ESP, and the EasIF variants under OmniESI's exact preprocessing, encoders, and evaluation scripts; if the quoted metric gaps shrink to noise or reverse, the superiority claim is falsified. A second decisive check is an equal-parameter ablation in which BCFM and CCFM are replaced by plain cross-attention: if the $k_{\text{cat}}$ and active-site gains disappear, the progressive conditioning is not the cause.

Watch

Extended reading notes

Core claim

The central discovery is that decomposing enzyme-substrate interaction modeling into two conditional feature-modulation stages creates a catalysis-aware latent space that transfers across tasks. In the first stage, the BCFM module makes the enzyme and substrate embeddings condition each other bidirectionally, emphasizing reaction specificity; in the second, the CCFM module derives a rough interaction vector from the modulated pair and uses it to re-weight enzyme residues and substrate atoms, emphasizing the contacts that matter for catalysis. The authors report that this architecture consistently outperforms specialized baselines: for $k_{\text{cat}}$ in the in-distribution CatPred-DB split it improves $R^2$ by 6.0% and MAE by 6.2% relative to CatPred; on the ESP overall test set it raises MCC by 0.079; on SwissProt E-RXN ASA active-site annotation it improves F1 by 4.4% and MCC by 2.7% relative to EasIF A-ESM; and on single-point CTX-M mutations it improves AUPRC and AUROC by 0.334 and 0.219 over UniKP. They also report that attention weights from the second conditional module highlight reactive bonds and functional groups consistent with known catalytic mechanisms, which they read as evidence that the network internalized patterns of catalytic efficiency rather than memorizing benchmark statistics.

Load-bearing premise

The central comparison assumes that taking the raw data splits from the original benchmark papers while quoting the competing methods' reported numbers gives a like-for-like evaluation; if the baselines used different preprocessing, input encodings, or evaluation scripts, the reported seven-benchmark superiority may not survive a controlled rerun.

Editorial extensions

If this is right

  • For enzyme kinetic parameters, the same network improves in-distribution predictions of $k_{\text{cat}}$, $K_m$, and $K_i$ over DLKcat, UniKP, and CatPred, with the largest out-of-distribution gains on $K_i$.
  • For enzyme-substrate pairing, OmniESI outperforms ESP on the overall out-of-distribution test set and on the 0–40% and 40–60% sequence-identity intervals, not just near the training distribution.
  • For active-site annotation, OmniESI beats both EasIF variants on the overall SwissProt E-RXN ASA test set and on four of the five sequence-identity intervals, and its attention maps align with amide-hydrolysis and glycosylation reaction centers.
  • For mutation effects, OmniESI classifies single-point CTX-M mutations with 86.1% accuracy versus close-to-random UniKP, and improves MCC on both double-point epistasis datasets.
  • Ablation and parameter counts show that BCFM and CCFM each contribute to the gains and together add only about 0.16% of total parameters, making the progressive conditioning the active ingredient rather than a larger model.

Reading between the lines

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

  • Beyond the paper, if the progressive conditioning is the active ingredient, then adding BCFM and CCFM to existing predictors such as UniKP or CatPred should lift their performance by a comparable margin; that transplant experiment would isolate the mechanism from the rest of OmniESI.
  • Beyond the paper, the attention maps that flag reactive bonds and substrate atoms suggest a direct application: use the CCFM attention weights to propose candidate catalytic residues and reactive atoms for unannotated enzyme-substrate pairs, then validate on a larger M-CSA set than the two illustrative cases.
  • Beyond the paper, the fact that one architecture serves four tasks with only output heads and training hyperparameters changed points to natural extensions such as enantioselectivity or thermostability prediction, which the paper does not evaluate.
  • Beyond the paper, the steady performance drop as sequence identity falls to 40% suggests the conditioning modules correct feature misalignment but not distribution shift; pairing them with domain-invariant or meta-learning training is a concrete testable route to close that gap.
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

5 major / 8 minor

Summary. The paper proposes OmniESI, a unified deep-learning framework for enzyme-substrate interaction prediction. It uses a frozen ESM-2-650M enzyme encoder and a trained GCN substrate encoder, followed by two conditional modules (BCFM and CCFM) that progressively modulate features toward a 'catalysis-aware' representation. The architecture is evaluated on four tasks: kinetic parameter regression (kcat, Km, Ki), enzyme-substrate pairing, active site annotation, and mutation-effect classification, using existing benchmark splits. The authors report that OmniESI outperforms or matches specialized baselines on most metrics, with the conditional modules adding only 0.16% of the total parameters.

Significance. If the reported comparisons are trustworthy, OmniESI is a valuable unified baseline and a useful architectural template: it couples strong pretrained encoders with lightweight task-agnostic conditioning, and it provides honest supplementary tables showing exceptions. The paper's strengths include public code/data links, reuse of established benchmark splits, transparent parameter accounting, and qualitative attention visualizations. However, the central 'consistently superior' claim depends on baseline numbers imported from previous papers without reruns, and the reported margins are small in several cases; the interpretability claim is also only qualitatively supported. These issues limit the current certainty of the headline claims but are addressable.

major comments (5)
  1. [§2.2, Supplementary Tables S1–S8] The headline SOTA claim rests on baseline metrics quoted from prior publications without rerunning those baselines under OmniESI's preprocessing and evaluation pipeline. The paper states in the supplementary that 'the raw results of the other models are taken from a prior work,' and Table S6 itself documents protocol mismatches: AEGAN was evaluated without removing 225 test samples overlapping its training set, and BLASTp used the entire SwissProt database as its alignment database. Since the reported margins are small (e.g., ID kcat R2 0.638 vs 0.602 in Table S1; ESP overall AUROC 0.972 vs 0.956 in Table S4), the comparison is not fully controlled. Please rerun the baselines under identical test files, featurization, and evaluation scripts, or provide a sensitivity analysis that quantifies how much protocol differences could shift the quoted numbers.
  2. [Abstract; §2.2; §2.3] The abstract's claim that OmniESI 'consistently delivered superior performance than state-of-the-art specialized methods across seven benchmarks' is stronger than the tables support. OmniESI is worse than CatPred on kcat OOD at 40% identity (R2 0.342 vs 0.365, Table S1), worse than CatPred on Km OOD at 60% and 40% (Table S2), and worse than EasIF A-ESM on the active-site 40–50% interval (F1 0.826 vs 0.836, Table S7). The text later correctly says 'best or second best on all metrics,' so the abstract and Discussion should be tempered to 'best or second best on most metrics' or equivalent.
  3. [§4.6; Tables S1–S12] No uncertainty quantification is provided: every metric is a single run, with no standard deviations or significance tests. Several improvements are modest (e.g., ID kcat R2 0.638 vs 0.602, Table S1; active-site F1 0.907 vs 0.902 at 50–60%, Table S7), so the claimed advantages could be within training noise. Please report mean ± std over at least three to five seeds and, where appropriate, paired significance tests for the main comparisons.
  4. [§2.3; §2.5; Abstract] The claim that the conditional networks 'internalize the fundamental patterns of catalytic efficiency' is supported only by two qualitative case studies and PCA visualizations, not by an independent quantitative validation. This is post hoc interpretation of attention weights. Please either weaken the claim to 'attention weights are qualitatively consistent with known catalytic mechanisms in selected cases' or add quantitative evidence, such as overlap between highlighted atoms/residues and experimentally annotated catalytic sites, with appropriate baselines.
  5. [§4.1.4; Tables S9–S11] The mutation-effect evaluation is not yet convincing as a fair comparison. The test sets are carved from the same DMS experiments without an explicit position-disjoint or sequence-disjoint split, and the baseline UniKP is reported at chance level (MCC=0, ACC=0.5, Table S9). Because the 17 active sites and substrates are shared across training and test mutations, there is a risk of information leakage through shared mutation positions or substrates. Please clarify the split construction, run a position-disjoint cross-validation, and compare against at least one additional baseline beyond UniKP.
minor comments (8)
  1. [§2.2; §4.1.4; Eq. (2)] Typographical errors should be fixed: 'domian transition' in the text near Eq. (2), 'OmniESi' in §2.2, and 'ceftazidimi' in §4.1.4.
  2. [Supplementary Tables S1–S3, S7–S8] The supplementary table captions render 'Table' as 'T able'; please correct the formatting.
  3. [Table S7 vs Table S6] The baseline name 'EasIF A-Saprot' in Table S7 is inconsistent with 'EasIF A-SaProt' used in the main text and Table S6.
  4. [§4.1.2] The entry counts are arithmetically inconsistent: 212,111 + 572,282 = 784,393, but the text states 784,396 entries.
  5. [References [4] and [9]] References [4] and [9] cite the same paper by Black et al. and should be merged or one removed.
  6. [§4.6] The statement that performance is insensitive to architecture hyperparameters is not supported by any sensitivity sweep; either add a small study or soften the claim.
  7. [§2.3] Please state more prominently that the MCSA E-RXN CSA zero-shot results are used only for interpretability visualization, not as a predictive benchmark.
  8. [Data and code availability] Please provide a versioned release of the code, pretrained weights, and dataset access details to enable reproduction.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: OmniESI's benchmark claims rest on held-out test evaluation against external baselines, and the disclosed baseline-protocol differences are a comparability concern, not a circularity.

full rationale

The paper is an empirical machine-learning study: OmniESI is trained on labeled enzyme-substrate data and evaluated on held-out test splits defined by external benchmark sources (CatPred-DB, ESP-DB, SwissProt E-RXN ASA). No equation in the paper defines a predicted quantity in terms of the target label, and no fitted parameter is renamed as a prediction. The architecture derivation is constructive rather than derivational: BCFM and CCFM are proposed modules, and their contributions are tested by ablations on held-out kcat subsets. The 'internalization of catalytic efficiency' claims come from post hoc attention visualization and PCA of learned representations, which are interpretability analyses rather than predictions forced by construction; they are weak validation evidence but not circular. The baseline numbers for DLKcat, UniKP, CatPred, ESP, EasIF, etc., are quoted from prior publications, and the paper itself discloses protocol differences for AEGAN and BLASTp in Supplementary Table S6. Quoting external baseline results raises a question of comparability and verification, but it is not a circularity: the comparison does not reduce to the present paper's own fitted values or to a self-citation chain. There is no uniqueness theorem imported from the authors, no ansatz smuggled in via citation, and no renaming of a known result presented as a derivation. Accordingly, the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

No new physical entities are introduced. The model adds learned parameters, but these are trained weights, not free parameters in the derivation sense. The free parameters listed are the hand-chosen hyperparameters. The axioms are the domain assumptions about encoders, datasets, and interpretability that the empirical claims rest on.

free parameters (2)
  • Batch size and learning rate per dataset scale = 8/1e-3 (small), 32/1e-4 (medium), 128/1e-4 (large)
    Chosen by hand per dataset scale; Supplementary Tables S15-S17. Although the paper claims architecture insensitivity, these optimization hyperparameters were adjusted per task.
  • Architecture hyperparameters = hidden dim 128; DSCNN layers 4; kernel sizes 5,9,13,17 (enzyme) and 3,5,7,9 (substrate); attention heads 4; dropout 0.1
    Determined by 'step-by-step hyperparameter tuning' (Section S7); unified across tasks but still hand-chosen numbers the results depend on.
assumptions (6)
  • domain assumption ESM-2 650M embeddings capture enzyme sequence features relevant to catalysis.
    Used as frozen enzyme encoder (Section 2.1, Eq. 1); if these features lack catalytic information, the conditional modules can only re-weight, not create, such information.
  • domain assumption 2D graph representation of substrate preserves the chemical features needed for catalysis.
    Substrate is encoded as 2D molecular graph via GCN (Section 2.1); stereochemistry and 3D conformation are not used.
  • domain assumption Benchmark datasets are accurate and the labels reflect true enzyme-substrate interactions.
    All results depend on CatPred-DB, ESP-DB, SwissProt E-RXN ASA, MCSA, and CTX-M DMS data (Section 4.1).
  • domain assumption Sequence identity to training set is a valid proxy for out-of-distribution difficulty.
    OOD evaluation is defined by max sequence identity cutoffs (Section 2.2); no other distribution shifts are considered.
  • domain assumption DMS mutation data can be split randomly into train/test without residue-level leakage.
    The single-point and double-point datasets are split by selecting 20% of positives (and matching negatives) without stated stratification by mutation position (Section 4.1.4), so mutations at the same positions may appear in both train and test.
  • ad hoc to paper Attention weights are interpretable evidence of catalytic mechanism internalization.
    The paper claims internalization from two attention visualizations (Section 2.3) without a quantitative alignment test.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OmniESI: A unified framework for enzyme-substrate interaction prediction with progressive conditional deep learning." pith.science (2026). https://pith.science/paper/PKZBZNSU

@misc{pith2026250617963,
  author       = {Pith},
  title        = {Pith review of: OmniESI: A unified framework for enzyme-substrate interaction prediction with progressive conditional deep learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PKZBZNSU}},
  note         = {Machine review of arXiv:2506.17963}
}
read the original abstract

Understanding and modeling enzyme-substrate interactions is crucial for catalytic mechanism research, enzyme engineering, and metabolic engineering. Although a large number of predictive methods have emerged, they do not incorporate prior knowledge of enzyme catalysis to rationally modulate general protein-molecule features that are misaligned with catalytic patterns. To address this issue, we introduce a two-stage progressive framework, OmniESI, for enzyme-substrate interaction prediction through conditional deep learning. By decomposing the modeling of enzyme-substrate interactions into a two-stage progressive process, OmniESI incorporates two conditional networks that respectively emphasize enzymatic reaction specificity and crucial catalysis-related interactions, facilitating a gradual feature modulation in the latent space from general protein-molecule domain to catalysis-aware domain. On top of this unified architecture, OmniESI can adapt to a variety of downstream tasks, including enzyme kinetic parameter prediction, enzyme-substrate pairing prediction, enzyme mutational effect prediction, and enzymatic active site annotation. Under the multi-perspective performance evaluation of in-distribution and out-of-distribution settings, OmniESI consistently delivered superior performance than state-of-the-art specialized methods across seven benchmarks. More importantly, the proposed conditional networks were shown to internalize the fundamental patterns of catalytic efficiency while significantly improving prediction performance, with only negligible parameter increases (0.16%), as demonstrated by ablation studies on key components. Overall, OmniESI represents a unified predictive approach for enzyme-substrate interactions, providing an effective tool for catalytic mechanism cracking and enzyme engineering with strong generalization and broad applicability.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

76 extracted references · 71 canonical work pages

  1. [1]

    Conformational aspects of inhibitor design: enzyme–substrate interactions in the transition state.Bioorganic medicinal chemistry7, 647–652 (1999)

    Wolfenden, R. Conformational aspects of inhibitor design: enzyme–substrate interactions in the transition state.Bioorganic medicinal chemistry7, 647–652 (1999)

    work page 1999

  2. [2]

    C., Kieslich, C

    Ferrall-Fairbanks, M. C., Kieslich, C. A. & Platt, M. O. Reassessing enzyme kinetics: considering protease-as-substrate interactions in proteolytic networks. Proceedings of the National Academy of Sciences117, 3307–3318 (2020)

    work page 2020

  3. [3]

    Albery, W. J. & Knowles, J. R. Evolution of enzyme function and the development of catalytic efficiency.Biochemistry15, 5631–5640 (1976). 1https://huggingface.co/timm 2https://huggingface.co/docs/transformers/index 3https://docs.pytorch.org/docs/stable/notes/ddp.html 4https://www.rdkit.org/ 22

    work page 1976

  4. [5]

    Campbell, E.et al.The role of protein dynamics in the evolution of new enzyme function.Nature Chemical Biology12, 944–950 (2016)

    work page 2016

  5. [6]

    S., Arcus, V

    Newton, M. S., Arcus, V. L., Gerth, M. L. & Patrick, W. M. Enzyme evolution: innovation is easy, optimization is complicated.Current Opinion in Structural Biology48, 110–116 (2018)

    work page 2018

  6. [7]

    Sweetlove, L. J. & Fernie, A. R. The role of dynamic enzyme assemblies and substrate channelling in metabolic regulation.Nature Communications9, 2136 (2018)

    work page 2018

  7. [8]

    D., Belsham, G

    Ryan, M. D., Belsham, G. J. & King, A. M. Specificity of enzyme-substrate interactions in foot-and-mouth disease virus polyprotein processing.Virology 173, 35–45 (1989)

    work page 1989

  8. [9]

    W.et al.A high-throughput screening method for determining the substrate scope of nitrilases.Chemical Communications51, 2660–2662 (2015)

    Black, G. W.et al.A high-throughput screening method for determining the substrate scope of nitrilases.Chemical Communications51, 2660–2662 (2015)

    work page 2015

Show all 76 references
  1. [10]

    K., Labanieh, L

    Longwell, C. K., Labanieh, L. & Cochran, J. R. High-throughput screening tech- nologies for enzyme engineering.Current Opinion in Biotechnology48, 196–202 (2017)

    2017

  2. [11]

    K., Heckmann, D

    Kroll, A., Engqvist, M. K., Heckmann, D. & Lercher, M. J. Deep learning allows genome-scale prediction of michaelis constants from structural features.PLoS biology19, e3001402 (2021)

    2021

  3. [12]

    Li, F.et al.Deep learning-based kcat prediction enables improved enzyme- constrained model reconstruction.Nature Catalysis5, 662–672 (2022)

    2022

  4. [13]

    Du, B.-X.et al.Gelkcat: An integration learning of substrate graph with enzyme embedding for kcat prediction.2023 IEEE International Conference on Bioinformatics and Biomedicine408–411 (2023)

    2023

  5. [14]

    & Yang, A

    Qiu, S., Zhao, S. & Yang, A. Dltkcat: deep learning-based prediction of temperature-dependent enzyme turnover rates.Briefings in Bioinformatics25, bbad506 (2024)

    2024

  6. [15]

    Kroll, A., Rousset, Y., Hu, X.-P., Liebrand, N. A. & Lercher, M. J. Turnover number predictions for kinetically uncharacterized enzymes using machine and deep learning.Nature communications14, 4139 (2023)

    2023

  7. [16]

    Yu, H., Deng, H., He, J., Keasling, J. D. & Luo, X. Unikp: a unified framework for the prediction of enzyme kinetic parameters.Nature Communications14, 8211 (2023). 23

    2023

  8. [17]

    & Yan, M

    He, X. & Yan, M. Graphkm: machine and deep learning for km prediction of wildtype and mutant enzymes.BMC bioinformatics25, 135 (2024)

    2024

  9. [18]

    & Lercher, M

    Kroll, A., Ranjan, S. & Lercher, M. J. A multimodal transformer network for protein-small molecule interactions enhances predictions of kinase inhibition and enzyme-substrate relationships.PLOS Computational Biology20, e1012100 (2024)

    2024

  10. [19]

    Wang, J.et al.Mpek: a multitask deep learning framework based on pretrained language models for enzymatic reaction kinetic parameters prediction.Briefings in Bioinformatics25, bbae387 (2024)

    2024

  11. [20]

    Wang, Z.et al.Robust enzyme discovery and engineering with deep learning using catapro.Nature communications16, 2736 (2025)

    2025

  12. [21]

    Boorla, V. S. & Maranas, C. D. Catpred: a comprehensive framework for deep learning in vitro enzyme kinetic parameters.Nature communications16, 2072 (2025)

    2025

  13. [22]

    Kroll, A., Ranjan, S., Engqvist, M. K. & Lercher, M. J. A general model to predict small molecule substrates of enzymes based on machine and deep learning. Nature Communications14, 2787 (2023)

    2023

  14. [23]

    Wang, X.et al.Multi-modal deep learning enables efficient and accurate annotation of enzymatic active sites.Nature Communications15, 7348 (2024)

    2024

  15. [24]

    E.et al.Uniref clusters: a comprehensive and scalable alternative for improving sequence similarity searches.Bioinformatics31, 926–932 (2015)

    Suzek, B. E.et al.Uniref clusters: a comprehensive and scalable alternative for improving sequence similarity searches.Bioinformatics31, 926–932 (2015)

    2015

  16. [25]

    Meier, J.et al.Language models enable zero-shot prediction of the effects of mutations on protein function.Advances in neural information processing systems 34, 29287–29303 (2021)

    2021

  17. [26]

    Lin, Z.et al.Evolutionary-scale prediction of atomic-level protein structure with a language model.Science379, 1123–1130 (2023)

    2023

  18. [27]

    Elnaggar, A.et al.Prottrans: toward understanding the language of life through self-supervised learning.IEEE Transactions on Pattern Analysis and Machine Intelligence44, 7112–7127 (2021)

    2021

  19. [28]

    & Ramsundar, B

    Chithrananda, S., Grand, G. & Ramsundar, B. Chemberta: large-scale self- supervised pretraining for molecular property prediction.arXiv preprint arXiv:2010.09885(2020)

    2020 arXiv

  20. [29]

    & Barati Farimani, A

    Wang, Y., Wang, J., Cao, Z. & Barati Farimani, A. Molecular contrastive learning of representations via graph neural networks.Nature Machine Intelligence4, 279–287 (2022). 24

    2022

  21. [30]

    Xia, J.et al.Mole-bert: rethinking pre-training graph neural networks for molecules.The Eleventh International Conference on Learning Representations (ICLR)(2023)

    2023

  22. [31]

    Mechanism of enzyme specificity in the domain of carbohydrates

    Gottschalk, A. Mechanism of enzyme specificity in the domain of carbohydrates. Nature160, 113–115 (1947)

    1947

  23. [32]

    & Boddy, C

    Wang, M. & Boddy, C. N. Examining the role of hydrogen bonding interactions in the substrate specificity for the loading step of polyketide synthase thioesterase domains.Biochemistry47, 11793–11803 (2008)

    2008

  24. [33]

    Gong, Q.et al.Application of cation-πinteractions in enzyme-substrate binding: Design, synthesis, biological evaluation, and molecular dynamics insights of novel hydrophilic substrates for nqo1.European Journal of Medicinal Chemistry221, 113515 (2021)

    2021

  25. [34]

    Kipf, T. N. & Welling, M. Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907(2016)

    2016 arXiv

  26. [35]

    Xception: deep learning with depthwise separable convolutions

    Chollet, F. Xception: deep learning with depthwise separable convolutions. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition1251–1258 (2017)

    2017

  27. [36]

    & Wang, X

    Yu, W., Zhou, P., Yan, S. & Wang, X. Inceptionnext: When inception meets convnext.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition5672–5683 (2024)

    2024

  28. [37]

    Cai, X.et al.Poly kernel inception network for remote sensing detection.Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 27706–27716 (2024)

    2024

  29. [38]

    R., Fan, Q

    Chen, C.-F. R., Fan, Q. & Panda, R. Crossvit: Cross-attention multi-scale vision transformer for image classification.Proceedings of the IEEE/CVF international conference on computer vision357–366 (2021)

    2021

  30. [39]

    & Kaur, B

    Sharma, A., Gupta, G., Ahmad, T., Mansoor, S. & Kaur, B. Enzyme engineering: current trends and future perspectives.Food Reviews International37, 121–154 (2021)

    2021

  31. [40]

    M., Edgar, S

    Woolston, B. M., Edgar, S. & Stephanopoulos, G. Metabolic engineering: past and future.Annual review of chemical and biomolecular engineering4, 259–288 (2013)

    2013

  32. [41]

    Conformational flexibility of enzyme active sites.Science262, 380– 381 (1993)

    Tsou, C.-L. Conformational flexibility of enzyme active sites.Science262, 380– 381 (1993). 25

    1993

  33. [42]

    D., Woycechowsky, K

    Toscano, M. D., Woycechowsky, K. J. & Hilvert, D. Minimalist active-site redesign: teaching old enzymes new tricks.Angewandte Chemie International Edition46, 3212–3236 (2007)

    2007

  34. [43]

    Osuna, S., Jimenez-Oses, G., Noey, E. L. & Houk, K. Molecular dynamics explo- rations of active site structure in designed and evolved enzymes.Accounts of chemical research48, 1080–1089 (2015)

    2015

  35. [44]

    Halgren, T. A. Identifying and characterizing binding sites and assessing druggability.Journal of chemical information and modeling49, 377–389 (2009)

    2009

  36. [45]

    Shen, X.et al.A highly sensitive model based on graph neural networks for enzyme key catalytic residue prediction.Journal of Chemical Information and Modeling63, 4277–4290 (2023)

    2023

  37. [46]

    F., Gish, W., Miller, W., Myers, E

    Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local alignment search tool.Journal of molecular biology215, 403–410 (1990)

    1990

  38. [47]

    L., Mitchell, J

    Holliday, G. L., Mitchell, J. B. & Thornton, J. M. Understanding the functional roles of amino acid residues in enzyme catalysis.Journal of Molecular Biology 390, 560–577 (2009)

    2009

  39. [48]

    Song, Z., Zhang, Q., Wu, W., Pu, Z. & Yu, H. Rational design of enzyme activ- ity and enantioselectivity.Frontiers in Bioengineering and Biotechnology11, 1129149 (2023)

    2023

  40. [49]

    J., Porter, C

    Bartlett, G. J., Porter, C. T., Borkakoti, N. & Thornton, J. M. Analysis of catalytic residues in enzyme active sites.Journal of molecular biology324, 105– 121 (2002)

    2002

  41. [50]

    J., Tyzack, J

    Ribeiro, A. J., Tyzack, J. D., Borkakoti, N., Holliday, G. L. & Thornton, J. M. A global analysis of function and conservation of catalytic residues in enzymes. Journal of Biological Chemistry295, 314–324 (2020)

    2020

  42. [51]

    Starr, T. N. & Thornton, J. W. Epistasis in protein evolution.Protein science 25, 1204–1218 (2016)

    2016

  43. [52]

    Miton, C. M. & Tokuriki, N. How mutational epistasis impairs predictability in protein evolution and design.Protein Science25, 1260–1272 (2016)

    2016

  44. [53]

    Drawz, S. M. & Bonomo, R. A. Three decades ofβ-lactamase inhibitors.Clinical microbiology reviews23, 160–201 (2010)

    2010

  45. [54]

    & Levesque, R

    Therrien, C. & Levesque, R. C. Molecular basis of antibiotic resistance andβ- lactamase inhibition by mechanism-based inactivators: perspectives and future directions.FEMS microbiology reviews24, 251–262 (2000). 26

    2000

  46. [55]

    & Coque, T

    Cant´ on, R. & Coque, T. M. The ctx-mβ-lactamase pandemic.Current opinion in microbiology9, 466–475 (2006)

    2006

  47. [56]

    Ambler, R. P. The structure ofβ-lactamases.Philosophical Transactions of the Royal Society of London. B, Biological Sciences289, 321–331 (1980)

    1980

  48. [57]

    P.et al.A standard numbering scheme for the class a beta-lactamases

    Ambler, R. P.et al.A standard numbering scheme for the class a beta-lactamases. Biochemical Journal276, 269 (1991)

    1991

  49. [58]

    Judge, A.et al.Mapping the determinants of catalysis and substrate specificity of the antibiotic resistance enzyme ctx-mβ-lactamase.Communications Biology 6, 35 (2023)

    2023

  50. [59]

    Chang, A.et al.Brenda, the elixir core data resource in 2021: new developments and updates.Nucleic acids research49, D498–D508 (2021)

    2021

  51. [60]

    & M¨ uller, W

    Wittig, U., Rey, M., Weidemann, A., Kania, R. & M¨ uller, W. Sabio-rk: an updated resource for manually curated biochemical reaction kinetics.Nucleic acids research46, D656–D660 (2018)

    2018

  52. [61]

    L.et al.Ncbi taxonomy: a comprehensive update on curation, resources and tools.Database2020, baaa062 (2020)

    Schoch, C. L.et al.Ncbi taxonomy: a comprehensive update on curation, resources and tools.Database2020, baaa062 (2020)

    2020

  53. [62]

    Consortium, T. U. Uniprot: the universal protein knowledgebase in 2025.Nucleic Acids Research53, D609–D617 (2025)

    2025

  54. [63]

    C.et al.The uniprot-go annotation database in 2011.Nucleic acids research40, D565–D570 (2012)

    Dimmer, E. C.et al.The uniprot-go annotation database in 2011.Nucleic acids research40, D565–D570 (2012)

    2012

  55. [64]

    & Bairoch, A

    Boutet, E., Lieberherr, D., Tognolli, M., Schneider, M. & Bairoch, A. Uniprotkb/swiss-prot: the manually annotated section of the uniprot knowledge- base.Plant bioinformatics: methods and protocols89–112 (2007)

    2007

  56. [65]

    Nature communications13, 964 (2022)

    Probst, D.et al.Biocatalysed synthesis planning using data-driven learning. Nature communications13, 964 (2022)

    2022

  57. [66]

    Fu, L., Niu, B., Zhu, Z., Wu, S. & Li, W. Cd-hit: accelerated for clustering the next-generation sequencing data.Bioinformatics28, 3150–3152 (2012)

    2012

  58. [67]

    Ribeiro, A. J. M.et al.Mechanism and catalytic site atlas (m-csa): a database of enzyme reaction mechanisms and active sites.Nucleic acids research46, D618– D623 (2018)

    2018

  59. [68]

    Li, Y.et al.Large selective kernel network for remote sensing object detec- tion.Proceedings of the IEEE/CVF international conference on computer vision 16794–16805 (2023). 27

    2023

  60. [69]

    & Feng, J

    Hou, Q., Lu, C.-Z., Cheng, M.-M. & Feng, J. Conv2former: A simple transformer- style convnet for visual recognition.IEEE transactions on pattern analysis and machine intelligence(2024)

    2024

  61. [70]

    Radford, A.et al.Learning transferable visual models from natural language supervision.International Conference on Machine Learning8748–8763 (2021)

    2021

  62. [71]

    Ramesh, A.et al.Zero-shot text-to-image generation.International Conference on Machine Learning8821–8831 (2021)

    2021

  63. [72]

    Ruiz, N.et al.Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation.Proceedings of the IEEE/CVF conference on computer vision and pattern recognition22500–22510 (2023)

    2023

  64. [73]

    & Courville, A

    Perez, E., Strub, F., De Vries, H., Dumoulin, V. & Courville, A. Film: visual reasoning with a general conditioning layer.Proceedings of the AAAI conference on artificial intelligence32(2018)

    2018

  65. [74]

    Vaswani, A.et al.Attention is all you need.Advances in Neural Information Processing Systems30(2017)

    2017

  66. [75]

    Wang, M.et al.Deep graph library: A graph-centric, highly-performant package for graph neural networks.arXiv preprint arXiv:1909.01315(2019)

    2019 arXiv

  67. [76]

    Li, M.et al.Dgl-lifesci: An open-source toolkit for deep learning on graphs in life science.ACS omega6, 27233–27238 (2021)

    2021

  68. [77]

    L.et al.Pymol: An open-source molecular graphics tool.CCP4 Newsl

    DeLano, W. L.et al.Pymol: An open-source molecular graphics tool.CCP4 Newsl. Protein Crystallogr40, 82–92 (2002). 28 Supplementary information S1 Prediction performance for enzyme kinetic parameters Boldindicates the best performance. Underline indicates the second-best. The r...

    2002

Pith tools

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