Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

This paper proposes HQDeepDTAF, a hybrid quantum-classical network that predicts protein-ligand binding affinity as accurately as a leading classical model while using fewer classical parameters, and claims its angle-embedding variant is fe

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 17:10 UTC pith:JG4RBRVU

load-bearing objection Solid hybrid QML study with a broken NISQ feasibility proof: the empirical core is worth reviewing, but the 300-circuits-vs-depth argument needs to be fixed. the 3 major comments →

arxiv 2509.11046 v1 pith:JG4RBRVU submitted 2025-09-14 cs.ET cs.LGq-bio.BM

Hybrid Quantum Neural Networks for Efficient Protein-Ligand Binding Affinity Prediction

Seon-Geun Jeong , Kyeong-Hwan Moon , Won-Joo Hwang This is my paper
classification cs.ET cs.LGq-bio.BM
keywords protein-ligand binding affinityhybrid quantum neural networkdata re-uploadingangle embeddingNISQ feasibilityparameter efficiencyquantum machine learningPDBbind
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tries to establish that hybrid quantum-classical neural networks can substitute for classical neural networks in protein-ligand binding affinity prediction, a task central to drug discovery. It proposes HQDeepDTAF, which uses a classical embedding network to project features into a low-dimensional latent space, a data re-uploading quantum circuit with angle embedding to process that latent vector, and a classical regression head for the final prediction. On the PDBbind core 2016 benchmark, the 9-qubit NN-Angle variant reports MAE 1.082, RMSE 1.368, and R 0.783, slightly better than the classical DeepDTAF baseline (MAE 1.109, R 0.769) while using about 100K classical parameters versus 154K. The authors also argue, via complexity analysis and noise simulations, that this angle-embedding variant is implementable on current NISQ hardware and degrades gracefully under low noise.

Core claim

The central claim is that the nonlinear expressive power of a quantum neural network can be unlocked without deep or many-qubit circuits by combining a classical embedding network with a data re-uploading quantum circuit. The authors show empirically that this hybrid (HQNN) fits univariate and multivariate damping functions more accurately than either a pure QNN or a classical NN of comparable parameter count, and that the same design improves binding-affinity prediction over a purely classical architecture. The paper further claims that the angle-embedding variant, with 9 qubits and 20 re-uploading layers, has circuit depth about 260 and is thus within the reach of current NISQ devices, whe

What carries the argument

The load-bearing object is the hybrid quantum embedding: a classical embedding network compresses the input to a latent vector h, which is encoded via angle embedding into a data re-uploading QNN that alternates trainable PQC blocks (single-qubit rotations plus CNOT entanglers) with data-encoding layers, followed by a classical regression network. The trainable classical embedding effectively optimizes the quantum feature map during training, allowing few qubits to represent task-relevant information; expressibility and entangling-capability metrics are used to justify the 9-qubit, 20-layer configuration.

Load-bearing premise

The NISQ feasibility claim assumes that current quantum computers accept circuits of depth 260, because the paper equates the supported maximum of 300 circuits per job with circuit depth; it also assumes 20 epochs is a fair training budget for all baselines.

What would settle it

Check the actual maximum circuit depth supported by current NISQ hardware for 9 qubits; if it is below 260, the NN-Angle model cannot run as claimed. Alternatively, re-train all baselines until convergence (or under matched parameter budgets) and compare; if the classical model matches or exceeds the hybrid, the stated advantage disappears.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If correct, hybrid QML models can match or exceed classical predictors on binding affinity with substantially fewer classical parameters, lowering training and inference cost.
  • The angle-embedding variant is implementable on current NISQ hardware under the stated depth assumptions, making the approach testable on real quantum devices.
  • Low-rate noise simulation shows graceful degradation, suggesting error mitigation could make the model practical before fault tolerance.
  • The same hybrid-embedding design could be transferred to other regression tasks in computational biology and beyond.
  • The expressibility and entangling-capability analysis provides a heuristic rule for choosing qubit count and circuit depth before committing to full training.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the NISQ feasibility claim appears to equate a per-job circuit-count limit with circuit depth; if that metric is not a depth bound, the '260 depth is feasible' conclusion needs re-checking.
  • Editorial inference: the 20-epoch training budget is short for classical baselines; a longer or tuned schedule might close the performance gap, so the 'replacement' claim is strongest as a parameter-efficiency statement rather than an unconditional accuracy win.
  • Editorial inference: the ablation showing frozen quantum layers still outperform a purely classical NN hints that the advantage may partly come from a regularization effect rather than purely quantum expressivity.
  • Editorial inference: running the model on real quantum hardware with error mitigation would settle whether the simulated gains survive in practice; the paper leaves that as future work.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes HQDeepDTAF, a hybrid quantum-classical architecture for protein-ligand binding affinity prediction. It replaces the dense layers of DeepDTAF with a data re-uploading QNN that is preceded by a classical embedding network and followed by a classical regression network. The authors evaluate HQNNs on univariate and multivariate function approximation, study expressibility/entangling capability as a function of qubit/layer counts, compare angle vs amplitude embeddings, run ablations, and benchmark HQDeepDTAF against DeepDTAF, DeepDTA, Pafnucy, TopologyNet, AEScore, and Kdeep on PDBbind core 2016. The best configuration (9-qubit NN-Angle, 20 layers) achieves MAE 1.082, RMSE 1.368, R 0.783 with 100,016 classical and 540 quantum parameters, slightly better than the reported DeepDTAF results (MAE 1.109, R 0.769) with fewer classical parameters. The paper also provides an efficiency analysis and a NISQ feasibility argument, concluding that only the NN-Angle variant is currently implementable on IBM quantum hardware.

Significance. If the numerical results hold, the paper makes a useful empirical contribution: it shows that a hybrid quantum embedding with data re-uploading can match or slightly beat a strong classical baseline on a practical regression task with fewer classical parameters, and it provides a concrete recipe for choosing qubit counts and circuit depth. The public code, five-run repeats with standard deviations, noise simulations, and ablation of frozen vs trainable quantum layers are strengths. The authors appropriately disclaim theoretical universality and do not overclaim quantum advantage. However, the central NISQ-feasibility claim rests on a faulty comparison in Section 6, and the baseline comparison uses a fixed 20-epoch budget without establishing convergence. Both issues are fixable but currently weaken the paper's main contributions.

major comments (3)
  1. [§6 (Feasibility of HQDeepDTAF)] The proof that only HQDeepDTAF-NN-Angle is implementable compares circuit depths of 260 and 10,480 with the statement that 'most IBM quantum computers support a maximum of 300 circuits [73]'. This conflates circuit depth (number of sequential gate layers) with the maximum number of circuits allowed per job/request. These quantities are incommensurable, so the argument does not establish that a depth-260 circuit is feasible on current hardware. Since NISQ feasibility is a stated contribution, this step must be replaced with a correct criterion, e.g., comparing compiled circuit depth or estimated execution time against device coherence times and maximum depth limits, or reporting an actual hardware run. The subsequent 'Practical Considerations' paragraph acknowledges connectivity and native-gate overheads, but it does not repair the erroneous 300-circuits comparison.
  2. [§5.1 and §5.7] All models, including classical baselines, are trained for only 20 epochs with no convergence criterion. Section 5.7 states that 'most models converge within 20 epochs', but no quantitative evidence is given, and Figure 8 shows substantial variance and signs of overfitting for some baselines. Because Table 7 reports final accuracy, a baseline that converges in more epochs is systematically disadvantaged. The claimed improvement over DeepDTAF is therefore not yet robustly established. Please report full learning curves for all baselines and either train until a common convergence criterion is met or justify quantitatively that 20 epochs is sufficient for every model.
  3. [§5.4, Table 5] The circuit-depth values reported for Hybrid NN-Angle are internally inconsistent with the formula in Section 6. With O((4+M_in)L) and M_in=9, the depths should be 65 (L=5), 130 (L=10), 195 (L=15), and 260 (L=20), but Table 5 lists O(11), O(55), O(195), and O(260); the L=5 and L=10 entries are wrong. The 2- and 4-qubit rows also scale differently from the stated formula. Since these depth values are used to argue that NN-Angle is NISQ-feasible, the table and the formula must be reconciled.
minor comments (5)
  1. [§5.5, Table 6] The text says 'the classical NN baseline requires 1,215 parameters for comparable depth', but Table 6 lists the Classical NN as having 101,336 classical parameters. The 1,215 figure is unexplained and inconsistent with the table; please correct or derive it.
  2. [§3.5] Theorem 1 contains typos: 'Lebesgue integer function' should be 'Lebesgue integrable function', and 'neuran' should be 'neuron'. Also, the statement mixes notation (w_i ∈ R^N with x_i scalars); please make the summation indices consistent.
  3. [§5.2] 'adpoted' should be 'adopted' in two places, and the parameter counts '25,241,865,3265' should be written with separators for readability.
  4. [§5.1, Baselines] TopologyNet was reimplemented using only protein sequence inputs because the original implementation is not publicly available. This is a significant modification; please state explicitly how this affects comparability with the published TopologyNet results.
  5. [References] The reference list contains inline repository URLs in Table 1 that are not formatted as formal citations. Please move them to the references or footnotes.

Circularity Check

0 steps flagged

No circular derivation; central results are empirical and self-contained, with only minor non-load-bearing self-citations.

full rationale

The paper's central claims—HQNN parameter efficiency and binding-affinity accuracy—are supported by direct numerical experiments (Tables 5–8) comparing HQDeepDTAF variants against classical baselines on PDBbind core 2016. The architecture (Definitions 3–6, Eqs. 11–14) is constructed from standard data-re-uploading and hybrid-embedding components, and the reported MAE/RMSE/R values are measured outputs, not quantities implied by the construction. The efficiency analysis in Section 6 is an explicit complexity comparison with stated assumptions (Eqs. 20–26); it does not presuppose the conclusion. The expressibility/entangling-capability analysis (Section 5.3) is used only to motivate the 9-qubit, 20-layer choice; the later benchmarks are empirical. Self-citations [15]–[17] appear only in related-work and convergence-similarity remarks and are not load-bearing for the numerical claims; no uniqueness theorem from prior work is invoked, and no ansatz is smuggled in via citation. The main caveats are non-circularity concerns: Section 6's feasibility 'proof' compares a circuit depth of 260 against IBM's limit of 300 circuits per job, which appears to conflate circuit depth with number of circuits, and the architecture search in Section 5.4 selects the best configuration from the same benchmark later reported as the headline result—both are correctness/benchmark-hygiene issues, not cases where a prediction reduces to its inputs by definition. Accordingly, no circular step is identified.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claim depends on architectural choices (qubit count, layer count) and on the assumption that the IBM '300 circuit' figure represents a depth limit. The latter is the most fragile item in the ledger.

free parameters (4)
  • Number of qubits (9) = 9
    Chosen so the NN-Amplitude model matches DeepDTAF's parameter count and as the smallest power-of-two encoding for 512D input; all headline results use 9 qubits.
  • Circuit layers (20) = 20
    Selected after expressibility/entangling-capability analysis; deeper circuits improve MAE but increase noise sensitivity. All NISQ comparisons use 20 layers.
  • Learning rate (0.005) and weight decay (0.01) = 0.005, 0.01
    Fixed for all models, including baselines; may disadvantage methods that need longer schedules.
  • Fixed sequence lengths (1000/150/63) = 1000, 150, 63
    Chosen to cover about 90% of sequences in the dataset; truncation and padding affect all models equally.
axioms (4)
  • domain assumption Data re-uploading QNNs can universally approximate univariate functions (Perez-Salinas et al. 2020; Schuld et al. 2021)
    Used to motivate why the QNN part can represent nonlinear functions; not proven in this paper.
  • ad hoc to paper IBM quantum computers support a maximum of 300 circuits
    Load-bearing for the NISQ feasibility proof; the cited source likely refers to job limits, not circuit depth.
  • domain assumption Depolarizing and amplitude damping channels capture relevant NISQ noise
    Used in the noise simulations; no hardware validation.
  • standard math PQC blocks can be decomposed via Z-Y decomposition
    Appendix B, standard result.

pith-pipeline@v1.3.0-alltime-deepseek · 29561 in / 14337 out tokens · 144880 ms · 2026-08-04T17:10:19.062833+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Hybrid Quantum Neural Networks for Efficient Protein-Ligand Binding Affinity Prediction." pith.science (2026). https://pith.science/paper/JG4RBRVU

@misc{pith2026250911046,
  author       = {Pith},
  title        = {Pith review of: Hybrid Quantum Neural Networks for Efficient Protein-Ligand Binding Affinity Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JG4RBRVU}},
  note         = {Machine review of arXiv:2509.11046}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Protein-ligand binding affinity is critical in drug discovery, but experimentally determining it is time-consuming and expensive. Artificial intelligence (AI) has been used to predict binding affinity, significantly accelerating this process. However, the high-performance requirements and vast datasets involved in affinity prediction demand increasingly large AI models, requiring substantial computational resources and training time. Quantum machine learning has emerged as a promising solution to these challenges. In particular, hybrid quantum-classical models can reduce the number of parameters while maintaining or improving performance compared to classical counterparts. Despite these advantages, challenges persist: why hybrid quantum models achieve these benefits, whether quantum neural networks (QNNs) can replace classical neural networks, and whether such models are feasible on noisy intermediate-scale quantum (NISQ) devices. This study addresses these challenges by proposing a hybrid quantum neural network (HQNN) that empirically demonstrates the capability to approximate non-linear functions in the latent feature space derived from classical embedding. The primary goal of this study is to achieve a parameter-efficient model in binding affinity prediction while ensuring feasibility on NISQ devices. Numerical results indicate that HQNN achieves comparable or superior performance and parameter efficiency compared to classical neural networks, underscoring its potential as a viable replacement. This study highlights the potential of hybrid QML in computational drug discovery, offering insights into its applicability and advantages in addressing the computational challenges of protein-ligand binding affinity prediction.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

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

  1. Exploring the Effects of Entanglement on Quantum Machine Learning of Pathogen Epitope-Receptor Binding

    quant-ph 2026-06 unverdicted novelty 3.0

    On a dataset of 80 epitopes, a high-entanglement ZZ feature map in a hybrid QNN showed lower training AUAC and higher test-to-training AUAC ratio than lower-entanglement maps while keeping competitive test accuracy.

Reference graph

Works this paper leans on

77 extracted references · 1 canonical work pages · cited by 1 Pith paper

  1. [1]

    Nature Reviews Drug Discovery3(11), 935–949 (2004)

    Kitchen, D.B., Decornez, H., Furr, J.R., Bajorath, J.: Docking and scoring in virtual screening for drug discovery: methods and applications. Nature Reviews Drug Discovery3(11), 935–949 (2004)

    2004

  2. [2]

    ACS omega4(14), 15956–15965 (2019)

    Zheng, L., Fan, J., Mu, Y.: Onionnet: a multiple-layer intermolecular-contact- based convolutional neural network for protein–ligand binding affinity prediction. ACS omega4(14), 15956–15965 (2019)

    2019

  3. [3]

    Nucleic Acids Research47(D1), 464–474 (2019)

    Burley, S.K., Berman, H.M., Bhikadiya, C., Bi, C., Chen, L., Di Costanzo, L., Christie, C., Dalenberg, K., Duarte, J.M., Dutta, S.,et al.: Rcsb protein data bank: biological macromolecular structures enabling research and education in fundamental biology, biomedicine, biotechnology and energy. Nucleic Acids Research47(D1), 464–474 (2019)

    2019

  4. [4]

    Proteins: Structure, Function, and Bioinformatics68(3), 726–737 (2007)

    Zhao, Y., Sanner, M.F.: Flipdock: docking flexible ligands into flexible receptors. Proteins: Structure, Function, and Bioinformatics68(3), 726–737 (2007)

    2007

  5. [5]

    Journal of Chemical Information and Modeling60(4), 2189–2198 (2020)

    Guterres, H., Im, W.: Improving protein-ligand docking results with high- throughput molecular dynamics simulations. Journal of Chemical Information and Modeling60(4), 2189–2198 (2020)

    2020

  6. [6]

    ACS Publications (2013) 37

    Brylinski, M.: Nonlinear scoring functions for similarity-based ligand docking and binding affinity prediction. ACS Publications (2013) 37

    2013

  7. [7]

    Journal of Chemical Information and Modeling53(12), 3399–3409 (2013)

    Cobanoglu, M.C., Liu, C., Hu, F., Oltvai, Z.N., Bahar, I.: Predicting drug– target interactions using probabilistic matrix factorization. Journal of Chemical Information and Modeling53(12), 3399–3409 (2013)

    2013

  8. [8]

    Bioinformatics34(21), 3666–3674 (2018)

    Stepniewska-Dziubinska, M.M., Zielenkiewicz, P., Siedlecki, P.: Development and evaluation of a deep learning model for protein–ligand binding affinity prediction. Bioinformatics34(21), 3666–3674 (2018)

    2018

  9. [9]

    ChemRxiv (2019) https://doi.org/10.26434/chemrxiv.9866912

    Rezaei, M., Li, Y., Li, X., Li, C.: Improving the accuracy of protein-ligand binding affinity prediction by deep learning models: benchmark and model. ChemRxiv (2019) https://doi.org/10.26434/chemrxiv.9866912

    work page doi:10.26434/chemrxiv.9866912 2019

  10. [10]

    PLoS computational biology13(7), 1005690 (2017)

    Cang, Z., Wei, G.-W.: Topologynet: Topology based deep convolutional and multi- task neural networks for biomolecular property predictions. PLoS computational biology13(7), 1005690 (2017)

    2017

  11. [11]

    Briefings in Bioinformatics22(5), 117 (2021)

    Zeng, Y., Chen, X., Luo, Y., Li, X., Peng, D.: Deep drug-target binding affinity prediction with multiple attention blocks. Briefings in Bioinformatics22(5), 117 (2021)

    2021

  12. [12]

    arXiv preprint arXiv:1902.04166 (2019) arXiv:1902.04166

    ¨Ozt¨ urk, H., Ozkirimli, E.,¨Ozg¨ ur, A.: Widedta: prediction of drug-target binding affinity. arXiv preprint arXiv:1902.04166 (2019) arXiv:1902.04166

    Pith/arXiv arXiv 1902

  13. [13]

    Briefings in Bioinformatics22(5), 072 (2021)

    Wang, K., Zhou, R., Li, Y., Li, M.: Deepdtaf: a deep learning method to predict protein–ligand binding affinity. Briefings in Bioinformatics22(5), 072 (2021)

    2021

  14. [14]

    Artificial Intelligence Review56(7), 5975–6037 (2023)

    Askr, H., Elgeldawi, E., Aboul Ella, H., Elshaier, Y.A., Gomaa, M.M., Hassanien, A.E.: Deep learning in drug discovery: an integrative review and future challenges. Artificial Intelligence Review56(7), 5975–6037 (2023)

    2023

  15. [15]

    EPJ Quantum Technology12(1), 32 (2025)

    Moon, K.-H., Jeong, S.-G., Hwang, W.-J.: Qsegrnn: quantum segment recurrent neural network for time series forecasting. EPJ Quantum Technology12(1), 32 (2025)

    2025

  16. [16]

    IEEE Access11, 113726–113739 (2023)

    Jeong, S.-G., Do, Q.-V., Hwang, H.-J., Hasegawa, M., Sekiya, H., Hwang, W.- J.: Hybrid quantum convolutional neural networks for uwb signal classification. IEEE Access11, 113726–113739 (2023)

    2023

  17. [17]

    ICT Express10(3), 608–613 (2024)

    Jeong, S.-G., Do, Q.V., Hwang, W.-J.: Short-term photovoltaic power forecast- ing based on hybrid quantum gated recurrent unit. ICT Express10(3), 608–613 (2024)

    2024

  18. [18]

    Quantum4, 226 (2020)

    P´ erez-Salinas, A., Cervera-Lierta, A., Gil-Fuster, E., Latorre, J.I.: Data re- uploading for a universal quantum classifier. Quantum4, 226 (2020)

    2020

  19. [19]

    Schuld, M., Sweke, R., Meyer, J.J.: Effect of data encoding on the expressive power of variational quantum-machine-learning models. Phys. Rev. A103, 032430 38 (2021)

    2021

  20. [20]

    Physical Review Research3(3), 032049 (2021)

    Wu, Y., Yao, J., Zhang, P., Zhai, H.: Expressivity of quantum neural networks. Physical Review Research3(3), 032049 (2021)

    2021

  21. [21]

    IBM Journal of Research and Development62(6), 6–1620 (2018)

    Cao, Y., Romero, J., Aspuru-Guzik, A.: Potential of quantum computing for drug discovery. IBM Journal of Research and Development62(6), 6–1620 (2018)

    2018

  22. [22]

    Journal of Chemical Theory and Computation18(12), 7001–7023 (2022)

    Blunt, N.S., Camps, J., Crawford, O., Izs´ ak, R., Leontica, S., Mirani, A., Moylett, A.E., Scivier, S.A., Sunderhauf, C., Schopf, P.,et al.: Perspective on the current state-of-the-art of quantum computing for drug discovery applications. Journal of Chemical Theory and Computation18(12), 7001–7023 (2022)

    2022

  23. [23]

    Advances in Neural Information Processing Systems 35, 27810–27823 (2022)

    Yu, Z., Yao, H., Li, M., Wang, X.: Power and limitations of single-qubit native quantum neural networks. Advances in Neural Information Processing Systems 35, 27810–27823 (2022)

    2022

  24. [24]

    Nature Communications14(1), 6001 (2023)

    Chen, S., Cotler, J., Huang, H.-Y., Li, J.: The complexity of nisq. Nature Communications14(1), 6001 (2023)

    2023

  25. [25]

    IEEE Internet of Things Journal, 1–1 (2025)

    Jeong, S.-G., Duc, P.D.A., Do, Q.V., Noh, D.-I., Tung, N.X., Chien, T.V., Pham, Q.-V., Hasegawa, M., Sekiya, H., Hwang, W.-J.: Quantum annealing-based sum rate maximization for multi-uav-aided wireless networks. IEEE Internet of Things Journal, 1–1 (2025)

    2025

  26. [26]

    IEEE Trans- actions on Computer-Aided Design of Integrated Circuits and Systems42(10), 3301–3314 (2023)

    Sun, X., Tian, G., Yang, S., Yuan, P., Zhang, S.: Asymptotically optimal circuit depth for quantum state preparation and general unitary synthesis. IEEE Trans- actions on Computer-Aided Design of Integrated Circuits and Systems42(10), 3301–3314 (2023)

    2023

  27. [27]

    Physical Review A102(3), 032420 (2020)

    LaRose, R., Coyle, B.: Robust data encodings for quantum classifiers. Physical Review A102(3), 032420 (2020)

    2020

  28. [28]

    Accounts of Chemical Research50(2), 302–309 (2017)

    Liu, Z., Su, M., Han, L., Liu, J., Yang, Q., Li, Y., Wang, R.: Forging the basis for developing protein–ligand interaction scoring functions. Accounts of Chemical Research50(2), 302–309 (2017)

    2017

  29. [29]

    https://arxiv

    Cao, Y., Guerreschi, G.G., Aspuru-Guzik, A.: Quantum Neuron: an elementary building block for machine learning on quantum computers (2017). https://arxiv. org/abs/1711.11240

    Pith/arXiv arXiv 2017

  30. [30]

    Europhysics Letters125(3), 30004 (2019)

    Torrontegui, E., Garc ´ ıa-Ripoll, J.J.: Unitary quantum perceptron as efficient universal approximator (a). Europhysics Letters125(3), 30004 (2019)

    2019

  31. [31]

    Quantum Information Processing21(4), 128 (2022) 39

    Maronese, M., Destri, C., Prati, E.: Quantum activation functions for quantum neural networks. Quantum Information Processing21(4), 128 (2022) 39

    2022

  32. [32]

    In: International Conference on Computational Science, pp

    Inajetovic, M.A., Orazi, F., Macaluso, A., Lodi, S., Sartori, C.: Enabling non- linear quantum operations through variational quantum splines. In: International Conference on Computational Science, pp. 177–192 (2023). Springer

    2023

  33. [33]

    Physical Review A101(1), 010301 (2020)

    Lubasch, M., Joo, J., Moinier, P., Kiffner, M., Jaksch, D.: Variational quantum algorithms for nonlinear problems. Physical Review A101(1), 010301 (2020)

    2020

  34. [34]

    Bioinformatics34(17), 821–829 (2018)

    ¨Ozt¨ urk, H.,¨Ozg¨ ur, A., Ozkirimli, E.: Deepdta: deep drug–target binding affinity prediction. Bioinformatics34(17), 821–829 (2018)

    2018

  35. [35]

    In: 2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp

    Li, Y., Rezaei, M.A., Li, C., Li, X.: Deepatom: A framework for protein- ligand binding affinity prediction. In: 2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 303–310 (2019). IEEE

    2019

  36. [36]

    Bioinformatics35(18), 3329–3338 (2019)

    Karimi, M., Wu, D., Wang, Z., Shen, Y.: Deepaffinity: interpretable deep learning of compound–protein affinity through unified recurrent and convolutional neural networks. Bioinformatics35(18), 3329–3338 (2019)

    2019

  37. [37]

    Scientific Reports 13(1), 17951 (2023)

    Domingo, L., Djukic, M., Johnson, C., Borondo, F.: Binding affinity predictions with hybrid quantum-classical convolutional neural networks. Scientific Reports 13(1), 17951 (2023)

    2023

  38. [38]

    Advanced Quantum Technologies6(9), 2300107 (2023)

    Dong, L., Li, Y., Liu, D., Ji, Y., Hu, B., Shi, S., Zhang, F., Hu, J., Qian, K., Jin, X.,et al.: Prediction of protein-ligand binding affinity by a hybrid quantum- classical deep learning algorithm. Advanced Quantum Technologies6(9), 2300107 (2023)

    2023

  39. [39]

    In: 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), vol

    Domingo, L., Chehimi, M., Banerjee, S., Yuxun, S.H., Konakanchi, S., Ogun- fowora, L., Roy, S., Selvarajan, S., Djukic, M., Johnson, C.: A hybrid quantum- classical fusion neural network to improve protein-ligand binding affinity predic- tions for drug discovery. In: 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), vol. 02, p...

    2024

  40. [40]

    Quantum Machine Intelligence3, 1–20 (2021)

    T¨ uys¨ uz, C., Rieger, C., Novotny, K., Demirk¨ oz, B., Dobos, D., Potamianos, K., Vallecorsa, S., Vlimant, J.-R., Forster, R.: Hybrid quantum classical graph neural networks for particle track reconstruction. Quantum Machine Intelligence3, 1–20 (2021)

    2021

  41. [41]

    Machine Learning: Science and Technology5(1), 015040 (2024)

    Senokosov, A., Sedykh, A., Sagingalieva, A., Kyriacou, B., Melnikov, A.: Quan- tum machine learning for image classification. Machine Learning: Science and Technology5(1), 015040 (2024)

    2024

  42. [42]

    Applied Soft Com- puting141, 110307 (2023) 40

    Ovalle-Magallanes, E., Alvarado-Carrillo, D.E., Avina-Cervantes, J.G., Cruz- Aceves, I., Ruiz-Pinales, J.: Quantum angle encoding with learnable rotation applied to quantum–classical convolutional neural networks. Applied Soft Com- puting141, 110307 (2023) 40

    2023

  43. [43]

    Quantum Machine Intelligence 7(1), 1–20 (2025)

    Sinha, A., Macaluso, A., Klusch, M.: Nav-q: quantum deep reinforcement learning for collision-free navigation of self-driving cars. Quantum Machine Intelligence 7(1), 1–20 (2025)

    2025

  44. [44]

    Nature reviews Molecular cell biology7(3), 188–197 (2006)

    Aloy, P., Russell, R.B.: Structural systems biology: modelling protein interactions. Nature reviews Molecular cell biology7(3), 188–197 (2006)

    2006

  45. [45]

    https://arxiv.org/abs/2001.03622

    Lloyd, S., Schuld, M., Ijaz, A., Izaac, J., Killoran, N.: Quantum embeddings for machine learning (2020). https://arxiv.org/abs/2001.03622

    Pith/arXiv arXiv 2020

  46. [46]

    Advances in neural information processing systems30(2017)

    Lu, Z., Pu, H., Wang, F., Hu, Z., Wang, L.: The expressive power of neural networks: A view from the width. Advances in neural information processing systems30(2017)

    2017

  47. [47]

    Weininger, D.: Smiles, a chemical language and information system. 1. introduc- tion to methodology and encoding rules. Journal of chemical information and computer sciences28(1), 31–36 (1988)

    1988

  48. [48]

    Journal of cheminformatics3, 1–14 (2011)

    O’Boyle, N.M., Banck, M., James, C.A., Morley, C., Vandermeersch, T., Hutchi- son, G.R.: Open babel: An open chemical toolbox. Journal of cheminformatics3, 1–14 (2011)

    2011

  49. [49]

    Biotechnology Advances, 108459 (2024)

    Harding-Larsen, D., Funk, J., Madsen, N.G., Gharabli, H., Acevedo-Rocha, C.G., Mazurenko, S., Welner, D.H.: Protein representations: Encoding biological infor- mation for machine learning in biocatalysis. Biotechnology Advances, 108459 (2024)

    2024

  50. [50]

    Bioinformatics36(Supplement 2), 735–744 (2020)

    Zhang, F., Shi, W., Zhang, J., Zeng, M., Li, M., Kurgan, L.: Probselect: accu- rate prediction of protein-binding residues from proteins sequences via dynamic predictor selection. Bioinformatics36(Supplement 2), 735–744 (2020)

    2020

  51. [51]

    Bioinformatics30(18), 2592–2597 (2014)

    Magnan, C.N., Baldi, P.: Sspro/accpro 5: almost perfect prediction of protein sec- ondary structure and relative solvent accessibility using profiles, machine learning and structural similarity. Bioinformatics30(18), 2592–2597 (2014)

    2014

  52. [52]

    Proceedings of the National Academy of Sciences108(4), 1326–1330 (2011)

    Wang, L., Berne, B., Friesner, R.: Ligand binding to protein-binding pockets with wet and dry regions. Proceedings of the National Academy of Sciences108(4), 1326–1330 (2011)

    2011

  53. [53]

    Applied Acoustics148, 123–132 (2019)

    Chen, Y., Guo, Q., Liang, X., Wang, J., Qian, Y.: Environmental sound classification with dilated convolutions. Applied Acoustics148, 123–132 (2019)

    2019

  54. [54]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recogni- tion. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016)

    2016

  55. [55]

    Physical Review A101(3), 032308 (2020)

    Schuld, M., Bocharov, A., Svore, K.M., Wiebe, N.: Circuit-centric quantum 41 classifiers. Physical Review A101(3), 032308 (2020)

    2020

  56. [56]

    Cambridge University Press, Cambridge, UK (2010)

    Nielsen, M.A., Chuang, I.L.: Quantum Computation and Quantum Information. Cambridge University Press, Cambridge, UK (2010)

    2010

  57. [57]

    Physical Review A98(3), 032309 (2018)

    Mitarai, K., Negoro, M., Kitagawa, M., Fujii, K.: Quantum circuit learning. Physical Review A98(3), 032309 (2018)

    2018

  58. [58]

    Journal of Cheminformatics13(1), 59 (2021)

    Meli, R., Anighoro, A., Bodkin, M.J., Morris, G.M., Biggin, P.C.: Learning protein-ligand binding affinity with atomic environment vectors. Journal of Cheminformatics13(1), 59 (2021)

    2021

  59. [59]

    Journal of chemical information and modeling58(2), 287–296 (2018)

    Jim´ enez, J., Skalic, M., Martinez-Rosell, G., De Fabritiis, G.: K deep: protein– ligand absolute binding affinity prediction via 3d-convolutional neural networks. Journal of chemical information and modeling58(2), 287–296 (2018)

    2018

  60. [60]

    PloS one8(12), 82138 (2013)

    Zhao, M., Lee, W.-P., Garrison, E.P., Marth, G.T.: Ssw library: an simd smith- waterman c/c++ library for use in genomic applications. PloS one8(12), 82138 (2013)

    2013

  61. [61]

    https: //arxiv.org/abs/1711.05101

    Loshchilov, I., Hutter, F.: Decoupled Weight Decay Regularization (2019). https: //arxiv.org/abs/1711.05101

    Pith/arXiv arXiv 2019

  62. [62]

    https://arxiv.org/abs/1811.04968

    Bergholm, V., Izaac, J., Schuld, M., Gogolin, C., Ahmed, S., Ajith, V., Alam, M.S., Alonso-Linaje, G., AkashNarayanan, B., Asadi, A., Arrazola, J.M., Azad, U., Banning, S., Blank, C., Bromley, T.R., Cordier, B.A., Ceroni, J., Delgado, A., Matteo, O.D., Dusko, A., Garg, T., Guala, D., Hayes, A., Hill, R., Ijaz, A., Isacsson, T., Ittah, D., Jahangiri, S., J...

    Pith/arXiv arXiv 2022

  63. [63]

    Noise reduction in speech processing, 1–4 (2009)

    Cohen, I., Huang, Y., Chen, J., Benesty, J., Benesty, J., Chen, J., Huang, Y., Cohen, I.: Pearson correlation coefficient. Noise reduction in speech processing, 1–4 (2009)

    2009

  64. [64]

    The Clinical Biochemist Reviews 29(Suppl 1), 23 (2008)

    Chesher, D.: Evaluating assay precision. The Clinical Biochemist Reviews 29(Suppl 1), 23 (2008)

    2008

  65. [65]

    Briefings in Bioinformatics16(2), 325–337 (2015) 42

    Pahikkala, T., Airola, A., Pietil¨ a, S., Shakyawar, S., Szwajda, A., Tang, J., Ait- tokallio, T.: Toward more realistic drug–target interaction predictions. Briefings in Bioinformatics16(2), 325–337 (2015) 42

    2015

  66. [66]

    In: International Conference on Machine Learning, pp

    Tan, M., Le, Q.: Efficientnet: Rethinking model scaling for convolutional neu- ral networks. In: International Conference on Machine Learning, pp. 6105–6114 (2019). PMLR

    2019

  67. [67]

    Sim, S., Johnson, P.D., Aspuru-Guzik, A.: Expressibility and entangling capabil- ity of parameterized quantum circuits for hybrid quantum-classical algorithms. Adv. Quantum Technol. (2019)

    2019

  68. [68]

    The annals of mathematical statistics22(1), 79–86 (1951)

    Kullback, S., Leibler, R.A.: On information and sufficiency. The annals of mathematical statistics22(1), 79–86 (1951)

    1951

  69. [69]

    Meyer, D.A., Wallach, N.R.: Global entanglement in multiparticle systems. J. Math. Phys.43(9), 4273–4278 (2002)

    2002

  70. [70]

    Quantum Machine Intelligence2(1), 2 (2020)

    Henderson, M., Shakya, S., Pradhan, S., Cook, T.: Quanvolutional neural net- works: powering image recognition with quantum circuits. Quantum Machine Intelligence2(1), 2 (2020)

    2020

  71. [71]

    IEEE Transactions on Pattern Analysis and Machine Intelligence46(12), 10184– 10195 (2024)

    Zhao, R.-X., Shi, J., Li, X.: Qksan: A quantum kernel self-attention network. IEEE Transactions on Pattern Analysis and Machine Intelligence46(12), 10184– 10195 (2024)

    2024

  72. [72]

    Reviews of Modern Physics94(1), 015004 (2022)

    Bharti, K., Cervera-Lierta, A., Kyaw, T.H., Haug, T., Alperin-Lea, S., Anand, A., Degroote, M., Heimonen, H., Kottmann, J.S., Menke, T.,et al.: Noisy intermediate-scale quantum algorithms. Reviews of Modern Physics94(1), 015004 (2022)

    2022

  73. [73]

    https://arxiv.org/abs/2410.00916

    AbuGhanem, M.: IBM Quantum Computers: Evolution, Performance, and Future Directions (2024). https://arxiv.org/abs/2410.00916

    Pith/arXiv arXiv 2024

  74. [74]

    Quantum5, 412 (2021)

    Malvetti, E., Iten, R., Colbeck, R.: Quantum circuits for sparse isometries. Quantum5, 412 (2021)

    2021

  75. [75]

    Physical review letters119(18), 180509 (2017)

    Temme, K., Bravyi, S., Gambetta, J.M.: Error mitigation for short-depth quantum circuits. Physical review letters119(18), 180509 (2017)

    2017

  76. [76]

    https:// arxiv.org/abs/quant-ph/0505030

    Dawson, C.M., Nielsen, M.A.: The Solovay-Kitaev algorithm (2005). https:// arxiv.org/abs/quant-ph/0505030

    Pith/arXiv arXiv 2005

  77. [77]

    Physical Review A52(5), 3457 (1995) 43

    Barenco, A., Bennett, C.H., Cleve, R., DiVincenzo, D.P., Margolus, N., Shor, P., Sleator, T., Smolin, J.A., Weinfurter, H.: Elementary gates for quantum computation. Physical Review A52(5), 3457 (1995) 43

    1995