Pith. sign in

REVIEW 3 major objections 4 minor 36 references

Q-Score: A Quantum-Native Scoring Function for Molecular Docking

T0 review · 3 major / 4 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read Q-Score ranks protein-ligand binding by orbital donor-acceptor energies solved as a quantum clique problem, not by summing empirical atom-pair contacts.

desk verdict Solid engineering of GNN-NBO weights into an existing MWVCP/QAOA docking pipeline, with honest NISQ scaling data; the headline orthogonality claim is real but the orbital-physics interpretation is under-validated. read the letter →

arxiv 2607.09737 v1 pith:GEP4XMG7 submitted 2026-07-02 physics.chem-ph cs.LG

Kangyu Zheng , Yidong Zhou , Ruihao Li , Zixin Ding , Zhiding Liang , Shaohua Li This is my paper
classification physics.chem-phcs.LG PACS 03.67.Lx87.15.Aa31.15.ae
keywords moleculardockingscoringfunctionQAOAorbitalinteractiongraphNBONISQdrugdiscoverymaximum-weightclique
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

Classical docking scores add up pairwise contacts and therefore favor larger molecules and miss the orbital charge-transfer effects that often decide binding specificity. This paper replaces that additive sum with Q-Score: a graph neural network predicts second-order orbital interaction energies, those energies become node weights on a compatibility graph, and a maximum-weight clique is found by digitized-counterdiabatic QAOA so that only mutually realizable anchors contribute. Across eleven co-crystal structures the quantum solver recovers the exact clique on eight targets at ten qubits. On one thousand AI-generated molecules the resulting ranks are statistically uncorrelated with a standard classical score, track orbital quality almost perfectly, show no molecular-weight bias, and enrich for strong orbital contacts at nearly twice the random rate. Hardware runs on an IBM Eagle processor confirm that six-qubit instances remain solvable on present-day noisy devices.

What carries the argument

The Orbital Interaction Graph (OIG): each node is an atom-pair interaction anchor weighted by summed GNN-predicted E(2) energies, edges encode rigid-body geometric compatibility, and the Q-Score is the total weight of the maximum clique found by DC-QAOA.

What would settle it

Replace the GNN-predicted E(2) node weights with high-level ab initio NBO energies on a held-out set of protein-ligand complexes and check whether the Spearman correlation with classical scores remains near zero and the interaction-quality enrichment factor remains near 2.

Watch

Extended reading notes

Core claim

Encoding GNN-predicted NBO second-order energies as node weights of an Orbital Interaction Graph and solving the resulting maximum-weight vertex clique with DC-QAOA yields a docking score that is orthogonal to classical empirical scoring, driven by orbital interaction quality, free of molecular-weight bias, and capable of recovering exact optimal cliques on most redocking targets at ten qubits.

Load-bearing premise

The graph neural network, trained only on small-molecule quantum datasets, must produce chemically trustworthy intermolecular orbital energies for real protein-ligand interfaces; if those predicted energies are systematically wrong, both the orthogonality and the enrichment results become artifacts of the model rather than of orbital physics.

Editorial extensions

If this is right

  • Docking pipelines can rank candidates by stereoelectronic complementarity instead of atom count, reducing size bias in virtual screens.
  • Q-Score and classical scores can be used jointly as complementary filters because they disagree on top-ranked molecules.
  • NISQ devices can already solve the six-qubit clique instances that arise from compact binding-site graphs.
  • As qubit counts grow, larger Orbital Interaction Graphs become feasible while classical exact solvers face exponential growth.

Reading between the lines

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

  • If the GNN energies prove reliable, the same OIG construction could serve as a drop-in rescoring stage for any docking engine that already produces poses.
  • Adaptive or instance-dependent penalty schedules would be required before the method can be applied blindly across chemically diverse targets without manual retuning.
  • The approach supplies a concrete test-bed for error-mitigation techniques: any method that restores ten-qubit bitstring fidelity would immediately enlarge the usable pocket size.
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 / 4 minor

Summary. The manuscript introduces Q-Score, a docking scoring function that replaces empirical pairwise sums with GNN-predicted NBO second-order perturbation energies E(2). These are aggregated into an Orbital Interaction Graph (OIG) whose maximum-weight vertex clique is solved by Digitized-Counterdiabatic QAOA; the clique weight is the score. Redocking on 11 co-crystal structures shows DC-QAOA recovers the exact classical optimum on 8 of 11 targets at 10 qubits (default penalty P=6). On 1000 Pocket2Mol-generated molecules across 10 targets, Q-Score is uncorrelated with Vina (mean Spearman ρ=0.05), strongly correlated with mean E(2) (ρ=0.90), free of molecular-weight bias, and yields an Interaction Quality Enrichment Factor of 1.96. Simulation mean approximation ratio is 0.94 (52% exact); 1000 circuits on IBM Eagle r3 establish a practical 6-qubit hardware boundary.

Significance. If the GNN-derived node weights are chemically faithful, the work supplies a compact, quantum-native scoring primitive that is demonstrably orthogonal to classical empirical scores and free of the well-known molecular-weight bias. The hybrid pipeline (SIMG* → OIG → DC-QAOA → Kabsch pose) is fully automated, the quantum solver is benchmarked at scale (1000 instances) with transparent approximation-ratio tables, and hardware execution on a 127-qubit Eagle processor is reported with match rates and circuit metrics. These elements make the paper a concrete data point for both quantum-chemistry-informed docking and NISQ combinatorial optimization, even though classical solvers remain competitive at the 10-qubit scale examined.

major comments (3)
  1. Section IV-B and the scoring-comparison claims (Table II, ρ(E(2),Q)=0.90, IQEF=1.96): SIMG* is trained exclusively on QM9/GEOM small-molecule data and applied zero-shot to protein–ligand interfaces after a 5 Å cut. No comparison of predicted intermolecular E(2) against true NBO (or even DFT interaction energies) is provided for any of the 11 co-crystal complexes or the 1000 generated poses. Because Q-Score is defined as the sum of those same E(2) weights (Eq. 9), the reported orbital-selectivity metrics are tautological once a clique is found; without external chemical validation of the node weights, the orthogonality and enrichment results cannot be attributed to genuine stereoelectronic physics rather than to properties of the GNN representation.
  2. Section V-C, Table I and Table VI: under the default penalty P=6, three of eleven redocking targets (FXR, PRMT5, SRC) return invalid cliques whose scores exceed the classical optimum. The subsequent penalty sweep shows that the optimal P is instance-dependent (P=10 recovers two targets but degrades others). Because the Q-Score itself is the weight of the recovered clique, an unreliable constraint-enforcement mechanism directly undermines score fidelity on a non-negligible fraction of targets; an adaptive or theoretically justified penalty schedule is needed before the method can be treated as a robust scorer.
  3. Section V-C (8SKH case study) and the redocking evaluation: Kabsch reconstruction yields all-atom RMSD of 3.34–3.45 Å. While the paper correctly notes that prior graph-based docking formulations often omit pose reconstruction, these RMSD values remain only modest by conventional docking standards and are reported for a single target. Without a systematic RMSD table across all 11 co-crystals (or a comparison against Vina poses on the same OIG anchors), it is difficult to judge whether the selected cliques correspond to chemically useful binding modes.
minor comments (4)
  1. Abstract and Section V-D: the phrase “enriching for strong orbital interactions at twice the random rate” should be qualified by the paper’s own acknowledgment that IQEF is expected by construction once E(2) weights are used.
  2. Figure 3(b) heatmap uses “inv” for invalid cliques; a brief legend or caption note would improve readability.
  3. Table VII reports wall-clock times for a single BRAF instance; adding mean ± std over the 1000-molecule set (already mentioned as 235 s) would make the scaling claim more precise.
  4. The distance tolerance τ is listed as target-specific in Table VI but the selection criterion is not described; a short sentence on how τ was chosen would aid reproducibility.

Circularity Check

2 steps flagged · score 6.0 of 10

Q-Score is defined as the sum of GNN-supplied E(2) weights (Eq. 9); the headline rho=0.90 and IQEF=1.96 are therefore tautological once a clique is found, as the paper itself notes.

  1. self definitional [Section IV-F, Eq. (9) and Eq. (3); Section V-D]
    "Q-Score = sum_{v_k in C*} w_k = sum_{v_k in C*} W(a_L^{(k)}, a_P^{(k)}) (9) ... W(a_L, a_P) = sum |E^{(2)}_{oL->oP}| (3) ... We note that the strong Q-Score/E^{(2)} correlation (rho=0.90) and high IQEF are expected by construction, since Q-Score is defined as a weighted sum of E^{(2)} energies; these metrics confirm internal consistency rather than independent docking validation."

    Q-Score is literally the sum of the GNN-predicted E(2) energies that form the node weights. Once DC-QAOA (or any solver) returns a clique, the Spearman correlation with mean E(2) and the Interaction Quality Enrichment Factor are forced by arithmetic identity; they cannot falsify or independently confirm that the score captures orbital physics. The paper states this openly yet still leads with rho=0.90 and IQEF=1.96 as headline results.

  2. fitted input called prediction [Abstract; Section V-D, Table II; Conclusion]
    "On 1000 AI-generated molecules, Q-Score is orthogonal to classical scoring with Spearman rho of 0.05, driven by orbital quality with rho of 0.90, and free of molecular-weight bias, enriching for strong orbital interactions at twice the random rate. ... Q-Score achieves a mean IQEF of 1.96, nearly 2x the rate of random selection for high orbital-energy molecules"

    The abstract and conclusion present 'driven by orbital quality with rho of 0.90' and 'enriching ... at twice the random rate' as empirical discoveries of the scoring function. Because those quantities are definitional consequences of Eq. 9, the language of discovery overstates what is actually measured. The only non-tautological claim in the same sentence is the orthogonality to Vina.

full rationale

The paper's strongest numerical claims about orbital selectivity (Spearman rho(E(2), Q)=0.90 and IQEF_Q=1.96) reduce by construction to the definition of Q-Score. Equation 9 defines Q-Score as the sum of the very W(a_L,a_P) node weights that are themselves sums of the GNN-predicted |E(2)| values (Eq. 3). Consequently any ranking by Q-Score is forced to correlate with mean E(2) per anchor and to enrich for high-E(2) molecules; the paper explicitly acknowledges this in Section V-D. The non-circular content that remains is the orthogonality to Vina (rho=0.05) and the absence of molecular-weight bias; those comparisons rest on an independent classical scorer and are not definitional. No self-citation chain or uniqueness theorem is load-bearing. The circularity is therefore partial and confined to the orbital-selectivity metrics that the abstract and conclusion still present as primary findings.

Assumptions & free parameters 4 free parameters · 4 assumptions · 2 invented entities

The central claims rest on the fidelity of the SIMG* GNN for intermolecular E(2), on the geometric-compatibility edge definition, on a hand-chosen penalty P that must be large enough to enforce clique constraints, and on the assumption that a rigid-body Kabsch reconstruction from a small clique is a meaningful docking pose. None of these are derived from first principles inside the paper; they are modeling choices or external model outputs.

free parameters (4)
  • constraint penalty P = 6.0 (default)
    Default P=6.0; three targets require P=10 to recover feasible cliques, while larger P degrades other targets. Instance-dependent and chosen by hand (Section V-C, Table I).
  • distance cutoff d_cut / d_max = 5.0 A
    Fixed at 5.0 A for binding-site extraction and node selection; directly controls which atoms enter the OIG.
  • distance tolerance tau = per-target (3.5-6.5 A)
    Per-target geometric compatibility threshold (Table VI); ranges 3.5-6.5 A and is not derived from first principles.
  • number of qubits / anchors N = 6/10/12
    Tunable parameter that truncates the interaction graph; experiments use N in {6,10,12}.
assumptions (4)
  • domain assumption SIMG* GNN predictions of NBO E(2) are chemically accurate for protein-ligand intermolecular contacts even though the model was trained on QM9/GEOM small molecules.
    Section IV-B; no independent validation against high-level QM for the 11 protein targets is supplied.
  • domain assumption Two interaction anchors are geometrically compatible under a rigid-body pose if and only if their ligand-ligand and protein-protein distances differ by at most tau (Eq. 4).
    Core of the OIG edge definition; assumes rigidity and ignores induced fit.
  • standard math A sufficiently large quadratic penalty P converts the constrained MWVCP into an unconstrained QUBO whose ground state is a valid maximum-weight clique.
    Standard QUBO encoding (Eq. 6); correctness depends on P being larger than any feasible clique weight.
  • domain assumption The Kabsch rigid-body alignment of the selected anchors yields a docking pose whose quality can be judged by RMSD to the crystal structure.
    Section IV-E; ignores ligand flexibility and side-chain motion.
invented entities (2)
  • Orbital Interaction Graph (OIG)
    purpose: Weighted graph whose nodes are atom-pair anchors carrying aggregated E(2) and whose edges encode rigid-body geometric compatibility; the docking score is the weight of its maximum clique.
    Constructed in Section IV-C; intermediate representation that packages GNN outputs for QAOA. No independent physical existence outside the pipeline.
  • Q-Score
    purpose: Scalar docking score defined as the total weight of the optimal clique returned by DC-QAOA (Eq. 9).
    Defined by construction as sum of selected E(2) terms; the quantity being optimized is identical to the reported score.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Q-Score: A Quantum-Native Scoring Function for Molecular Docking." pith.science (2026). https://pith.science/paper/GEP4XMG7

@misc{pith2026260709737,
  author       = {Pith},
  title        = {Pith review of: Q-Score: A Quantum-Native Scoring Function for Molecular Docking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GEP4XMG7}},
  note         = {Machine review of arXiv:2607.09737}
}
read the original abstract

Molecular docking predicts how a small molecule binds to a protein and is a key bottleneck in drug discovery. Classical scoring functions sum empirical pairwise contacts, blind to quantum-mechanical effects like orbital charge transfer that govern binding specificity. We introduce Q-Score, encoding GNN-predicted orbital donor-acceptor energies into a weighted graph and scoring binding by solving a maximum-weight vertex clique problem via Digitized-Counterdiabatic QAOA. Each interaction anchor maps to one qubit and compatibility constraints become edges. Across 11 protein targets, DC-QAOA recovers the exact optimum on 8 at 10 qubits. On 1000 AI-generated molecules, Q-Score is orthogonal to classical scoring with Spearman rho of 0.05, driven by orbital quality with rho of 0.90, and free of molecular-weight bias, enriching for strong orbital interactions at twice the random rate. DC-QAOA achieves a mean approximation ratio of 0.94 with 52 percent exact. Execution of 1000 circuits on IBM Eagle confirms 6-qubit solvability on NISQ hardware.

Figures

Figures reproduced from arXiv: 2607.09737 by the authors.

Figure 1
Figure 1. Overview of the hybrid AI–quantum molecular docking pipeline. (A) Starting from a protein–ligand co-crystal structure, the binding site is extracted [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. An example DC-QAOA circuit schematic for the 6-qubit, 3-layer configuration. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 2
Figure 2. Throughout, we use a maximum pair distance [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: Redocking results across 11 co-crystal structures. (a) Classical optimal vs DC-QAOA score at 10q/3L. (b) Approximation ratio heatmap across all [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]
Figure 4
Figure 4. Figure 4: Q-Score captures orthogonal chemical information. (a) Spearman correlation heatmap across 10 targets: Q-Score is uncorrelated with Vina ( () [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: (a) Interaction Quality Enrichment Factor: Q-Score enriches for high- [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Hardware results on IBM Eagle r3 (1000 circuits, 10000 shots each). (a) Hardware vs simulator approximation ratio by configuration. (b) Per-target [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Computational cost and scaling. Simulation: (a) variational parameter count and (b) wall-clock time on a single CPU across 6, 10, and 12 qubits. Hardware: (c) transpiled circuit depth and (d) gate count on IBM Eagle r3 for 6 and 10 qubits. Solid = 3 layers, dashed = 6 …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 11 canonical work pages

  1. [1]

    Autodock vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading,

    O. Trott and A. J. Olson, “Autodock vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading,”Journal of Computational Chemistry, vol. 31, no. 2, pp. 455–461, 2010. [Online]. Available: https: //doi.org/10.1002/jcc.21334

    doi:10.1002/jcc.21334 2010

  2. [2]

    Glide: A new approach for rapid, accurate docking and scoring. 1. method and assessment of docking accuracy,

    R. A. Friesner, J. L. Banks, R. B. Murphy, T. A. Halgren, J. J. Klicic, D. T. Mainz, M. P. Repasky, E. H. Knoll, M. Shelley, J. K. Perry, D. E. Shaw, P. Francis, and P. S. Shenkin, “Glide: A new approach for rapid, accurate docking and scoring. 1. method and assessment of docking accuracy,”Journal of Medicinal Chemistry, vol. 47, no. 7, pp. 1739–1749, 200...

    doi:10.1021/jm0306430 2004

  3. [3]

    Stereoelectronic effects in biomolecules,

    D. G. Gorenstein, “Stereoelectronic effects in biomolecules,”Chemical Reviews, vol. 87, no. 5, pp. 1047–1077, 1987. [Online]. Available: https://doi.org/10.1021/cr00081a009

    work page doi:10.1021/cr00081a009 1987

  4. [4]

    Unbiasing scoring functions: A new normalization and rescoring strategy,

    G. Carta, A. J. S. Knox, and D. G. Lloyd, “Unbiasing scoring functions: A new normalization and rescoring strategy,”Journal of Chemical Information and Modeling, vol. 47, no. 4, pp. 1564–1571, 2007, pMID: 17552493. [Online]. Available: https://doi.org/10.1021/ci600471m

    work page doi:10.1021/ci600471m 2007

  5. [5]

    The role of quantum mechanics in structure-based drug design,

    K. Raha, M. B. Peters, B. Wang, N. Yu, A. M. Wollacott, L. M. Westerhoff, and K. M. Merz, “The role of quantum mechanics in structure-based drug design,”Drug Discovery Today, vol. 12, no. 17, pp. 725–731, 2007. [Online]. Available: https: //doi.org/10.1016/j.drudis.2007.07.006

    work page doi:10.1016/j.drudis.2007.07.006 2007

  6. [6]

    Molecular docking via quantum approximate optimization algorithm,

    Q.-M. Ding, Y .-M. Huang, and X. Yuan, “Molecular docking via quantum approximate optimization algorithm,”Phys. Rev. Appl., vol. 21, p. 034036, Mar 2024. [Online]. Available: https://doi.org/10. 1103/PhysRevApplied.21.034036

    2024

  7. [7]

    A perspective on protein structure prediction using quantum computers,

    H. Doga, B. Raubenolt, F. Cumbo, J. Joshi, F. P. DiFilippo, J. Qin, D. Blankenberg, and O. Shehab, “A perspective on protein structure prediction using quantum computers,”Journal of Chemical Theory and Computation, vol. 20, no. 9, pp. 3359–3378, 2024, pMID: 38703105. [Online]. Available: https://doi.org/10.1021/acs.jctc.4c00067

    work page doi:10.1021/acs.jctc.4c00067 2024

  8. [8]

    A quantum approximate optimization algorithm,

    E. Farhi, J. Goldstone, and S. Gutmann, “A quantum approximate optimization algorithm,” 2014. [Online]. Available: https://doi.org/10. 48550/arXiv.1411.4028

    arXiv 2014

Show all 36 references
  1. [9]

    Advancing molecular machine learning representations with stereoelectronics-infused molecular graphs,

    D. A. Boiko, T. Resch ¨utzegger, B. Sanchez-Lengeling, S. M. Blau, and G. Gomes, “Advancing molecular machine learning representations with stereoelectronics-infused molecular graphs,”Nature Machine Intelligence, vol. 7, no. 5, pp. 771–781, may 2025. [Online]. Available: https...

    2025 doi

  2. [10]

    Natural bond orbitals and extensions of localized bonding concepts,

    F. WEINHOLD and C. R. LANDIS, “Natural bond orbitals and extensions of localized bonding concepts,”Chem. Educ. Res. Pract., vol. 2, pp. 91–104, 2001. [Online]. Available: http://doi.org/10.1039/ B1RP90011K

    2001

  3. [11]

    Digitized-counterdiabatic quantum approximate optimization algorithm,

    P. Chandarana, N. N. Hegade, K. Paul, F. Albarr ´an-Arriagada, E. Solano, A. del Campo, and X. Chen, “Digitized-counterdiabatic quantum approximate optimization algorithm,”Phys. Rev. Res., vol. 4, p. 013141, Feb 2022. [Online]. Available: https://doi.org/10.1103/ PhysRevResear...

    2022

  4. [12]

    Geometric hashing: An overview,

    H. J. Wolfson and I. Rigoutsos, “Geometric hashing: An overview,” IEEE Comput. Sci. Eng., vol. 4, no. 4, p. 10–21, Oct. 1997. [Online]. Available: https://doi.org/10.1109/99.641604

    1997 doi

  5. [13]

    Ising formulations of many np problems,

    A. Lucas, “Ising formulations of many np problems,”Frontiers in Physics, vol. V olume 2 - 2014, 2014. [Online]. Available: https://doi.org/10.3389/fphy.2014.00005

    2014 doi

  6. [14]

    Quantum approximate optimization of non-planar graph problems on a planar superconducting processor,

    M. P. Harrigan, K. J. Sung, M. Neeley, K. J. Satzinger, F. Arute, K. Arya, J. Atalaya, J. C. Bardin, R. Barends, S. Boixo, M. Broughton, B. B. Buckley, D. A. Buell, B. Burkett, N. Bushnell, Y . Chen, Z. Chen, B. Chiaro, R. Collins, W. Courtney, S. Demura, A. Dunsworth, D. Eppe...

    2021 doi

  7. [15]

    Quantum Computing in the NISQ era and beyond,

    J. Preskill, “Quantum Computing in the NISQ era and beyond,” Quantum, vol. 2, p. 79, Aug. 2018. [Online]. Available: https: //doi.org/10.22331/q-2018-08-06-79

    2018 doi

  8. [16]

    Molecular docking with gaussian boson sampling,

    L. Banchi, M. Fingerhuth, T. Babej, C. Ing, and J. M. Arrazola, “Molecular docking with gaussian boson sampling,”Science Advances, vol. 6, no. 23, p. eaax1950, 2020. [Online]. Available: https: //doi.org/10.1126/sciadv.aax1950

    2020 doi

  9. [17]

    Qdockbank: A dataset for ligand docking on protein fragments predicted on utility-level quantum computers,

    Y . Zhang, Y . Yang, C.-C. Lu, W. Jiang, F. Cheng, B. Fang, and Q. Guan, “Qdockbank: A dataset for ligand docking on protein fragments predicted on utility-level quantum computers,” inProceedings of the International Conference for High Performance Computing, Networking, Stora...

    2025 doi

  10. [18]

    Resource-efficient quantum algorithm for protein folding,

    A. Robert, P. K. Barkoutsos, S. Woerner, and I. Tavernelli, “Resource-efficient quantum algorithm for protein folding,”npj Quantum Information, vol. 7, p. 38, 2021. [Online]. Available: https://doi.org/10.1038/s41534-021-00368-4

    2021 doi

  11. [19]

    A quantum framework for protein binding-site structure prediction on utility-level quantum processors,

    Y . Zhang, Y . Yang, W. Martin, K. Lin, Z. Wang, C.-C. Lu, W. Jiang, R. Nussinov, J. Loscalzo, Q. Guan, and F. Cheng, “A quantum framework for protein binding-site structure prediction on utility-level quantum processors,”Advanced Science, vol. 13, no. 12, p. e13641,

  12. [20]

    Available: https://doi.org/10.1002/advs.202513641

    [Online]. Available: https://doi.org/10.1002/advs.202513641

    doi

  13. [21]

    Ligand-binding affinity estimates supported by quantum-mechanical methods,

    U. Ryde and P. S ¨oderhjelm, “Ligand-binding affinity estimates supported by quantum-mechanical methods,”Chemical Reviews, vol. 116, no. 9, pp. 5520–5566, 2016, pMID: 27077817. [Online]. Available: https://doi.org/10.1021/acs.chemrev.5b00630

    2016 doi

  14. [22]

    Exploring chemistry with the fragment molecular orbital method,

    D. G. Fedorov, T. Nagata, and K. Kitaura, “Exploring chemistry with the fragment molecular orbital method,”Phys. Chem. Chem. Phys., vol. 14, pp. 7562–7577, 2012. [Online]. Available: http: //doi.org/10.1039/C2CP23784A

    2012 doi

  15. [23]

    GNINA 1.3: the next increment in molecular docking with deep learning,

    A. T. McNutt, Y . Li, R. Meli, R. Aggarwal, and D. R. Koes, “GNINA 1.3: the next increment in molecular docking with deep learning,” Journal of Cheminformatics, vol. 17, no. 1, p. 28, 2025. [Online]. Available: https://doi.org/10.1186/s13321-025-00973-x

    2025 doi

  16. [24]

    Benchmarking the performance of portfolio optimization with qaoa,

    S. Brandhofer, D. Braun, V . Dehn, G. Hellstern, M. Huls, Y . Ji, I. Polian, A. S. Bhatia, and T. Wellens, “Benchmarking the performance of portfolio optimization with qaoa,”Quantum Information Processing, vol. 22, no. 1, p. 25, 2023. [Online]. Available: https://doi.org/10.10...

    2023 doi

  17. [25]

    A case study of variational quantum algorithms for a job shop scheduling problem,

    D. Amaro, M. Rosenkranz, N. Fitzpatrick, K. Hirano, and M. Fiorentini, “A case study of variational quantum algorithms for a job shop scheduling problem,”EPJ Quantum Technology, vol. 9, p. 5, 2022. [Online]. Available: https://doi.org/10.1140/epjqt/s40507-022-00123-4

    2022 doi

  18. [26]

    Evidence for the utility of quantum computing before fault tolerance,

    Y . Kim, A. Eddins, S. Anand, K. X. Wei, E. van den Berg, S. Rosenblatt, H. Nayfeh, Y . Wu, M. Zaletel, K. Temme, and A. Kandala, “Evidence for the utility of quantum computing before fault tolerance,”Nature, vol. 618, no. 7965, pp. 500–505, 2023. [Online]. Available: https://...

    2023 doi

  19. [27]

    A purely algebraic justification of the kabsch-umeyama algorithm,

    J. Lawrence, J. Bernal, and C. Witzgall, “A purely algebraic justification of the kabsch-umeyama algorithm,”Journal of Research of the National Institute of Standards and Technology, vol. 124, Oct. 2019. [Online]. Available: http://doi.org/10.6028/jres.124.028

    2019 doi

  20. [28]

    Natural bond orbital analysis of molecular interactions: Theoretical studies of binary complexes of hf, h2o, nh3, n2, o2, f2, co, and co2 with hf, h2o, and nh3,

    A. E. Reed, F. Weinhold, L. A. Curtiss, and D. J. Pochatko, “Natural bond orbital analysis of molecular interactions: Theoretical studies of binary complexes of hf, h2o, nh3, n2, o2, f2, co, and co2 with hf, h2o, and nh3,”The Journal of Chemical Physics, vol. 84, no. 10, pp. 5...

    1986 doi

  21. [29]

    Quantum chemistry structures and properties of 134 kilo molecules,

    R. Ramakrishnan, P. O. Dral, M. Rupp, and O. A. von Lilienfeld, “Quantum chemistry structures and properties of 134 kilo molecules,” Scientific Data, vol. 1, no. 1, p. 140022, aug 2014. [Online]. Available: https://doi.org/10.1038/sdata.2014.22

    2014 doi

  22. [30]

    Geom, energy-annotated molecular conformations for property prediction and molecular generation,

    S. Axelrod and R. G ´omez-Bombarelli, “Geom, energy-annotated molecular conformations for property prediction and molecular generation,”Scientific Data, vol. 9, no. 1, p. 185, apr 2022. [Online]. Available: https://doi.org/10.1038/s41597-022-01288-4

    2022 doi

  23. [31]

    M. J. D. Powell,A Direct Search Optimization Method That Models the Objective and Constraint Functions by Linear Interpolation. Dordrecht: Springer Netherlands, 1994, pp. 51–67. [Online]. Available: https://doi.org/10.1007/978-94-015-8330-5 4

    1994 doi

  24. [32]

    Beyond affinity: A benchmark of 1d, 2d, and 3d methods reveals critical trade-offs in structure-based drug design,

    K. Zheng, K. Zhang, J. Tan, X. Chen, Y . Lu, Z. Zhang, L. Sun, M. Zitnik, T. Fu, and Z. Liang, “Beyond affinity: A benchmark of 1d, 2d, and 3d methods reveals critical trade-offs in structure-based drug design,”

  25. [33]

    Available: https://doi.org/10.48550/arXiv.2601.14283

    [Online]. Available: https://doi.org/10.48550/arXiv.2601.14283

    doi

  26. [34]

    Pocket2mol: Efficient molecular sampling based on 3d protein pockets,

    X. Peng, S. Luo, J. Guan, Q. Xie, J. Peng, and J. Ma, “Pocket2mol: Efficient molecular sampling based on 3d protein pockets,” 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2205.07249

    2025 doi arXiv

  27. [35]

    Efficient variational quantum simulator incorporating active error minimization,

    Y . Li and S. C. Benjamin, “Efficient variational quantum simulator incorporating active error minimization,”Phys. Rev. X, vol. 7, p. 021050, Jun 2017. [Online]. Available: https://doi.org/10.1103/ PhysRevX.7.021050

    2017

  28. [36]

    Error mitigation for short- depth quantum circuits,

    K. Temme, S. Bravyi, and J. M. Gambetta, “Error mitigation for short- depth quantum circuits,”Phys. Rev. Lett., vol. 119, p. 180509, Nov 2017. [Online]. Available: https://doi.org/10.1103/PhysRevLett.119.180509

    2017 doi

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.