Pith. sign in

REVIEW 4 major objections 5 minor 47 references

OTMol: Robust Molecular Structure Comparison via Optimal Transport

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Atom mapping, not rotation, is the real RMSD problem—OTMol solves it with optimal transport.

desk verdict A promising application of supervised Gromov-Wasserstein to molecular alignment, but the reported guarantees are softer than the abstract implies and the validation protocol cherry-picks hyperparameters. read the letter →

arxiv 2509.01550 v1 pith:XY6WIUU5 submitted 2025-09-01 q-bio.BM math.OC

Xiaoqi Wei , Xuhang Dai , Yaqi Wu , Yanxiang Zhao , Yingkai Zhang , Zixuan Cang This is my paper
classification q-bio.BMmath.OC
keywords RMSDsupervisedoptimaltransportGromov-Wassersteinmolecularsuperimpositionatommappingchiralitypreservationbondconnectivityconsistencywaterclusters
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

OTMol claims that the molecular RMSD problem can be solved properly by treating atom matching as a fused supervised Gromov-Wasserstein optimal transport problem, rather than as a linear assignment with hand-built cost matrices. The method replaces heuristic costs with pairwise intra-molecular distance comparisons—Euclidean plus graph-geodesic distances—and hard atom-label constraints, then applies the Kabsch rotation without reflection. On benchmarks spanning ATP, Imatinib, cyclic peptides, lipids, sugars, DNA, water clusters, and neon clusters, OTMol reports zero bond-connectivity inconsistencies and chirality-preserving alignments, and lower RMSD than ArbAlign for most tested pairs, while running in O(n^3) time. The paper argues that RMSD comparisons are only meaningful when chemical integrity is preserved, so a slightly higher RMSD from OTMol can be the correct answer when the lower value from a baseline comes from an atom mapping that breaks bonds or inverts stereochemistry.

What carries the argument

Load-bearing object: the fused supervised Gromov-Wasserstein objective (Eq. 4), combining a label cost (Cij = 0 for matching atom labels, infinity otherwise) with a Gromov-Wasserstein term that penalizes mismatches between pairwise distances inside each molecule. Distance matrices are a convex combination of Euclidean distance and graph-geodesic distance, so the plan respects geometry and covalent topology. OTMol scans alpha in [0,1], hard-thresholds each plan into a permutation, filters to the lowest-BCI assignments, and applies Kabsch without reflection. For clusters, the mechanism is hierarchical: molecule-level assignment via perturbed GW on representative coordinates, then block-structu

What would settle it

Take a reference molecule, randomly permute its atom indices and apply a random rotation, and ask OTMol to align the permuted copy to the original; a correct method must recover the identity assignment with near-zero RMSD and zero BCI. Repeating this over many permutations and flexible conformers settles the claim—any non-identity assignment with nonzero RMSD on such constructed examples falsifies the method.

Watch

Extended reading notes

Core claim

Central claim: atom-to-atom correspondence, not the rigid-body fit, is the real bottleneck in RMSD, and the fused supervised Gromov-Wasserstein problem finds that correspondence. OTMol solves Eq. (4) with hard atom-label costs and D = (1 - c)DE + cDG; for each alpha it converts the transport plan to a permutation, keeps assignments with the lowest bond-connectivity inconsistency, and returns the lowest-RMSD one. Reflection is never applied, so chirality is preserved. Clusters of identical atoms use a GW-plus-OT refinement; clusters of identical molecules use a two-level hierarchical match. Result: chemically meaningful RMSD for single molecules and clusters without manual cost matrices.

Load-bearing premise

The pipeline assumes that the optimal-transport score built from atom labels and pairwise graph distances ranks chemically correct atom matchings above incorrect ones, so that the lowest-RMSD assignment surviving the bond-connectivity filter is the right one.

Editorial extensions

If this is right

  • Atom-order-independent RMSD: structures from different software or databases can be compared directly, since the same assignment is found regardless of input ordering.
  • Low RMSD becomes interpretable: a BCI-minimal, chirality-preserving mapping means a low RMSD reflects genuine similarity rather than broken bonds or a reflected mirror image.
  • Cluster comparisons stop splitting molecules: the hierarchical match ensures each water molecule maps as a whole, removing a failure mode of the baseline.
  • The fsGW formulation is reusable for molecular analog comparison and maximum-common-substructure search, as the paper notes.
  • O(n^3) scaling keeps the method practical for larger systems such as neon clusters up to 1000 atoms.

Reading between the lines

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

  • The alpha-grid plus BCI filter compensates for the nonconvexity of Gromov-Wasserstein; a direct optimization of BCI/RMSD could replace the grid and likely improve robustness.
  • The same distance-preserving matching principle may transfer to point-cloud, surface, or density-map alignment problems where correspondence is unknown.
  • Because the implementation relies on a generic GW solver, the practical ceiling is set by solver local optima; a permuted-input recovery test would quantify this ceiling and could guide solver choice.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces OTMol, an optimal-transport-based method for RMSD calculation that is invariant to atom ordering. It casts the atom-assignment problem as a fused supervised Gromov-Wasserstein problem (Eq. 4) with label constraints and distance matrices combining Euclidean and graph-geodesic distances, solves it with the POT library, converts the soft transport plan to a permutation by row-wise argmax with a validity check, and then applies Kabsch alignment. For homogeneous atom clusters and molecular clusters, it uses GW/OT with Hadamard powers and perturbation-based sampling of molecule-level assignments. The method is benchmarked against ArbAlign on ATP, imatinib, cyclic peptides, lipids, sugars, DNA, water clusters, neon clusters, and FGG/hydrate systems using RMSD, bond-connectivity inconsistency (BCI), and atom-mapping inconsistency (AMI). The paper claims order-independent, chirality-preserving, one-to-one mapping with low RMSD and O(n^3) scaling.

Significance. OTMol addresses a practical need: order-independent RMSD with chemical fidelity. Strengths include public code and data, use of established OT solvers, the introduction of BCI/AMI as complementary quality metrics, and explicit attention to chirality and molecular-cluster integrity. The comparison to ArbAlign on external benchmark sets is valuable. However, the reported results are weakened by post hoc selection over a large hyperparameter grid and by the lack of a hard permutation constraint in the optimization. The central claims are plausible but require stronger validation before the method can be regarded as a general-purpose tool.

major comments (4)
  1. [§2.2, Algorithms 1–4] The abstract claims OTMol 'enforces one-to-one mappings', but the algorithms do not enforce this. After solving the fsGW/GW/OT problem, P′ is obtained by taking j = argmax_k P_ik and accepted only if it is a permutation. The optimization itself has no constraint preventing two rows from sharing the same argmax; the algorithm merely filters. Consequently, for some α no valid P′ may exist, and Algorithm 4 can return an empty list, making the final 'Take the lowest RMSD' undefined. No failure counts are reported for the 101 α values, 9 p-values, or 100 perturbations. The stated guarantee is therefore unsupported. Either the method should solve a constrained assignment problem (e.g., Hungarian rounding with proven permutation output) or the claim should be weakened and failure rates reported.
  2. [§2.5.1 vs Algorithm 1] The selection rule used in the experiments differs from the pseudocode. Algorithm 1 returns the lowest RMSD over α, while §3.1.1 and §2.5.1 state that the user chooses the lowest-RMSD assignment among the set of assignments with the lowest BCI values. This BCI filtering is central to the reported 0% bond-mismatch results, yet it is not part of the provided algorithm listing. The method as described is therefore not fully reproducible; the pseudocode should include the BCI filter or the text should explain how the filter is applied.
  3. [§3.1.1, §3.2.1–3.2.3] The evaluation protocol selects the best result over L_α = {0, 0.01, ..., 1}, dataset-specific c, L_p = {0.5, ..., 4.5}, and l = 100 after BCI filtering. The reported RMSD for each pair is thus a minimum over a large hyperparameter grid, not the performance of a fixed model. ArbAlign has no analogous selection step, so the comparison in Figures 5D/E is not apples-to-apples. Moreover, a single α curve is shown for only one FGG pair (Fig. 4B) and one water-cluster pair (Fig. 5A), so the reader cannot assess sensitivity. To support the 'consistently achieves low RMSD' claim, the paper should report fixed-hyperparameter results, cross-validation, or the distribution of RMSD across the grid.
  4. [§3.1.5, Table 2] Several molecular classes in the biomedical dataset consist of a single conformer pair: DNA (215D), DLP, EIC, BGC, and BGCGLC. Table 2 reports ArbAlign BCI statistics across Boltz-2 models, but for OTMol only aggregate statements are given; there is no RMSD distribution with sample sizes. The broad claim that OTMol 'consistently' outperforms across 'a wide range' of systems is not supported by n=1 per class. Additional independent pairs per class, or a clearly qualified statement, are needed.
minor comments (5)
  1. [Abstract] The statement that OTMol 'eliminates the need for manually defined cost functions' overstates the contribution; the method still requires choosing atom-label schemes, α, c, p, and l, and uses a label cost matrix in Eq. (4). Suggest revising the wording.
  2. [§4.3, Eq. (6)] The 'bilevel optimization' formula introduces a threshold ρ and an outer maximization over s that are never defined or used in the algorithms; it is unclear how Eq. (6) relates to the implemented fsGW of Eq. (4). Please clarify or remove.
  3. [Algorithms 2–3] Several typos and notation issues appear: 'Calulate' in Algorithm 3, 'RMSD P′' instead of 'RMSD_P′', and reuse of P for different plans in Algorithm 2. Using distinct symbols would improve readability.
  4. [§2.5.1] The BCI definition 'the same atom pair is found in the aligned structure' is ambiguous: does this refer to atom indices after permuting B? It should be specified whether BCI is computed on the mapping P′ or on the aligned coordinates.
  5. [Table 1] There is a typo in the cyclic peptides row: 'smller dihedral variation' should be 'smaller dihedral variation'. Similar typos occur elsewhere, e.g., 'decribed' in §3.1.5.

Circularity Check

2 steps flagged · score 6.0 of 10

Headline performance claims are partly circular: OTMol selects alignments by the same BCI/RMSD metrics it then reports as achievements.

  1. fitted input called prediction [Section 2.3 (Algorithm 1) and Section 3.1.1]
    ""For a given set of α, denoted by Lα, we solve an fsGW problem with C, DA, DB for each α and choose the atom assignment that has the lowest RMSD. ... We choose the lowest RMSD assignment from the set of atom assignments that have the lowest percentage of mismatched bonds." (Sec. 2.3); "We only choose the assignment with the lowest RMSD from the set of assignments that have the lowest BCI values. ... OTMol is able to match all edges for all pairs of conformers." (Sec. 3.1.1)"

    The final assignment is selected, over the α grid, by first minimizing BCI and then RMSD. Reporting 'OTMol is able to match all edges' (BCI=0) and 'consistently achieves low RMSD' is therefore restating the selection rule, not an independent property of the fsGW solution. ArbAlign is not given this post-hoc filter or 101-value hyperparameter search, so the headline comparison is forced in OTMol's favor. The central chemical-fidelity claim reduces, for these experiments, to the choice rule in Algorithm 1 rather than to the OT formulation.

  2. fitted input called prediction [Section 3.2.3, Algorithms 3-4]
    ""L ← PerturbationBeforeGW(RA, RB, l)" and "Take the lowest RMSDP′" (Algorithm 3); "The number of perturbations l on the centroid coordinates is 100 for all pairs. The influence of l on OTMol RMSD of the pair 10-PP1 and 10-PP2 is shown in Figure 5A-C." (Sec. 3.2.3)"

    OTMol's water-cluster result is the minimum RMSD over l random perturbations/OT solves. The observation that 'OTMol RMSD decreases as the number of perturbations increases' (Figure 5A) is a mathematical consequence of taking a minimum over a larger sample, not evidence that the OT assignment improves. The reported low RMSD is thus the selection statistic itself, so the benchmark comparison with ArbAlign on clusters is partially circular.

full rationale

Equation-level derivation is self-contained: Eq. (4) is a standard fsGW objective; no parameter is defined as the RMSD it later predicts. However, the evaluation protocol selects the reported metrics by construction. Algorithm 1 tries 101 α values and retains the assignment with lowest BCI, then lowest RMSD; Algorithms 3-4 similarly keep the lowest-RMSD over perturbations. Consequently the paper's headline 'enforces bond connectivity' and 'low RMSD' claims are, for these benchmarks, partly the selection rule restated. The external ArbAlign comparison does not remove the bias because ArbAlign is not given the same filter/search. Separately, the abstract's 'enforcing one-to-one mappings' is not delivered by the pseudocode: P' is only accepted if it happens to be a permutation; the optimization does not constrain rows to have distinct argmaxes, and no failure counts are reported. This is a correctness/robustness gap rather than an equation-level circularity. Overall, circularity is partial: the OT method itself is not vacuous, but the central performance claims reduce in part to the metric-based selection, giving score 6.

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

The method postulates that intra-molecular Euclidean plus graph-geodesic distances and atom labels are sufficient to define chemically valid atom assignments (Eq. 4, Algorithm 1), and that cluster assignments can be obtained from perturbed centroid matching with a block cost (Algorithms 3 and 4). Hyperparameters alpha, c, p, l, and the label scheme are tuned per dataset or per pair, so they are honest free parameters. No new physical entities are introduced.

free parameters (5)
  • alpha (fsGW balance weight) = Selected per pair from L_alpha = {0, 0.01, ..., 1}; e.g., alpha = 0.62 for FGG 55/FGG 470 at c = 0.5
    Balances the label cost term and the Gromov-Wasserstein structure term in Eq. 4. Swept in 0.01 increments and chosen by lowest RMSD after BCI filtering, so it is fitted to the benchmark pairs.
  • c (geodesic vs Euclidean distance weight) = 1 for most biomedical molecules and S1-MA-W1; 0.5 for cyclic peptides and FGG
    Sets D = (1-c)DE + cDG in Algorithm 1. Hand-chosen per dataset with no principled selection rule, and it changes the entire alignment objective.
  • p (Hadamard power on distance matrix) = L_p = {0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5}; per-pair selected
    Used in Algorithm 2 for clusters of identical atoms. The best p is chosen by lowest RMSD for each pair, so it is fit to the test data.
  • l (number of perturbations in cluster alignment) = 100 for water clusters
    Controls how many random Gaussian perturbations are used in Algorithm 4. Figure 5 shows RMSD improves as l increases, indicating this parameter was tuned to improve results.
  • atom label scheme = Element names for most datasets; SYBYL types or atom connectivity for FGG and S1-MA-W1
    The cost matrix C in Eq. 4 depends entirely on the label scheme. The choice is made per dataset and changes which assignments are allowed.
assumptions (5)
  • domain assumption Pairwise intra-molecular distances, using Euclidean and graph-geodesic terms, are sufficient to encode chemically meaningful atom assignment.
    Eq. 4 and Algorithm 1 replace manual cost matrices with DA and DB computed from coordinates and adjacency. If these distances do not capture chemistry, the reported correct mappings are not guaranteed.
  • ad hoc to paper The continuous optimal transport plan P from a nonconvex fsGW solver can be converted into a valid permutation P' by taking row-wise argmax.
    Section 2.2 defines P' by j = arg max_k P_ik and only checks whether P' is a permutation. No proof is given that the heuristic always produces a permutation or preserves the objective quality.
  • ad hoc to paper The POT fGW and EMD solvers return solutions good enough for the claims, despite the nonconvexity of the Gromov-Wasserstein objective.
    Section 3 states all computations use POT solvers, but Eq. 4 is nonconvex in general. The paper provides no optimality guarantees or initialization analysis.
  • domain assumption Perturbing molecule centroids and taking the best RMSD over l trials recovers the correct molecule-level assignment and preserves one-to-one molecular mapping.
    Algorithms 3 and 4 explicitly acknowledge that optimal atom-level assignments may not arise from the optimal molecule-level assignment, then rely on random perturbations without a guarantee. The block-cost sOT avoids cross-molecule atom spillover only if the molecule-level assignment is correct.
  • domain assumption A rigid-body Kabsch transformation without reflection preserves the chirality of the molecule being aligned.
    Eq. 1 restricts T to non-reflective rigid transformations, but the atom assignment itself comes from reflection-invariant distance matrices. Chirality preservation therefore depends on the assignment not being a mirror-matched assignment, which is nontrivial.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OTMol: Robust Molecular Structure Comparison via Optimal Transport." pith.science (2026). https://pith.science/paper/XY6WIUU5

@misc{pith2026250901550,
  author       = {Pith},
  title        = {Pith review of: OTMol: Robust Molecular Structure Comparison via Optimal Transport},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XY6WIUU5}},
  note         = {Machine review of arXiv:2509.01550}
}
read the original abstract

Root-mean-square deviation (RMSD) is widely used to assess structural similarity in systems ranging from flexible ligand conformers to complex molecular cluster configurations. Despite its wide utility, RMSD calculation is often challenged by inconsistent atom ordering, indistinguishable configurations in molecular clusters, and potential chirality inversion during alignment. These issues highlight the necessity of accurate atom-to-atom correspondence as a prerequisite for meaningful alignment. Traditional approaches often rely on heuristic cost matrices combined with the Hungarian algorithm, yet these methods underutilize the rich intra-molecular structural information and may fail to generalize across chemically diverse systems. In this work, we introduce OTMol, a method that formulates the molecular alignment task as a fused supervised Gromov-Wasserstein (fsGW) optimal transport problem. By leveraging the intrinsic geometric and topological relationships within each molecule, OTMol eliminates the need for manually defined cost functions and enables a principled, data-driven matching strategy. Importantly, OTMol preserves key chemical features such as molecular chirality and bond connectivity consistency. We evaluate OTMol across a wide range of molecular systems, including Adenosine triphosphate, Imatinib, lipids, small peptides, and water clusters, and demonstrate that it consistently achieves low RMSD values while preserving computational efficiency. Importantly, OTMol maintains molecular integrity by enforcing one-to-one mappings between entire molecules, thereby avoiding erroneous many-to-one alignments that often arise in comparing molecular clusters. Our results underscore the utility of optimal transport theory for molecular alignment and offer a generalizable framework applicable to structural comparison tasks in cheminformatics, molecular modeling, and related disciplines.

Figures

Figures reproduced from arXiv: 2509.01550 by the authors.

Figure 1
Figure 1. (A) Visualization of the ATP molecules from PDB and CSD databases. Only the ATP part is shown. (Left: [PITH_FULL_IMAGE:figures/full_fig_p013_1.png] view at source ↗
Figure 2
Figure 2. (A) Visualization of the Imatinib molecules from PDB and CSD databases. Only the ATP part is [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. (A, B, C, D, E) The OTMol alignment of the 4-mer pair, 5-mer pair, 6-mer pair, 7-mer pair, and 8-mer [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: (A) The OTMol alignment of FGG 55 (reference structure; colored in red) and FGG 470 (structure to be aligned; colored in blue) when c = 0 (left) or 0.5 (right), respectively. Element names are used as labels. (B) The influence of α on OTMol RMSD of the pair FGG 55 (col…
Figure 5
Figure 5. Figure 5: (A) The influence of l on OTMol RMSD of the pair 10-PP1 and 10-PP2. The OTMol RMSD decreases as the number of perturbations increases. (B) The OTMol alignment of 10-PP1(pink: hydrogen atoms; red: oxygen atoms), 10-PP2(light blue: hydrogen atoms; dark blue: oxygen atoms…
Figure 6
Figure 6. Figure 6: Comparison of OTMol and ArbAlign alignments on the FGG tripeptides. (A) RMSD comparison. A black [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Comparison of alignment results for atmospheric hydrate configurations. (A) RMSD comparison. A black [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: (A) The RMSD values of ArbAlign and OTMol for water cluster datasets. Two representative options, [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]
Figure 9
Figure 9. Figure 9: (A, B) The running time comparison of OTMol and ArbAlign for the Neon cluster dataset and the water [PITH_FULL_IMAGE:figures/full_fig_p022_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages

  1. [1]

    Johnson, Gerald M

    Mark A. Johnson, Gerald M. Maggiora, et al. Concepts and applications of molecular similarity. Wiley, 1990

    work page 1990

  2. [2]

    Algorithm for efficient superposition and clustering of molecular assem- blies using the branch-and-bound method

    Yuki Yamamoto. Algorithm for efficient superposition and clustering of molecular assem- blies using the branch-and-bound method. Journal of Chemical Information and Modeling, 65(9):4512–4530, April 2025

    work page 2025

  3. [3]

    Accurate structure prediction of biomolecular interactions with Al- phaFold3

    John Jumper and et al. Accurate structure prediction of biomolecular interactions with Al- phaFold3. Nature, 630:493–500, 2025. 23

    work page 2025

  4. [4]

    Miller III, and Animashree Anandkumar

    Zhuoran Qiao, Weili Nie, Arash Vahdat, Thomas F. Miller III, and Animashree Anandkumar. State-specific protein–ligand complex structure prediction with a multiscale deep generative model. Nature Machine Intelligence, 6:195–208, 2024

    work page 2024

  5. [5]

    Boltz-1: Democratizing biomolecular interaction modeling with open-source foundation mod- els

    Jeremy Wohlwend, Gabriele Corso, Saro Passaro, Regina Barzilay, and Tommi Jaakkola. Boltz-1: Democratizing biomolecular interaction modeling with open-source foundation mod- els. bioRxiv, 2024. Preprint

    work page 2024

  6. [6]

    Boltz-2: A biomolecular foundation model for structure and binding affinity prediction, 2025

    Jeremy Wohlwend, Gabriele Corso, Saro Passaro, Regina Barzilay, and Tommi Jaakkola. Boltz-2: A biomolecular foundation model for structure and binding affinity prediction, 2025. PDF available at https://jeremywohlwend.com/assets/boltz2.pdf

    work page 2025

  7. [7]

    State-aware protein-ligand complex prediction using AlphaFold3 with purified sequences

    Enming Xing, Junjie Zhang, Shen Wang, and Xiaolin Cheng. State-aware protein–ligand complex prediction using AlphaFold3 with purified sequences. arXiv, 2025. arXiv:2506.00147

    work page Pith review arXiv 2025

  8. [8]

    Predicting molecular energy using force-field optimized geometries and atomic vector representations learned from an improved deep tensor neural network

    Jianing Lu, Cheng Wang, and Yingkai Zhang. Predicting molecular energy using force-field optimized geometries and atomic vector representations learned from an improved deep tensor neural network. Journal of Chemical Theory and Computation, 15(7):4113–4121, May 2019

    work page 2019

Show all 47 references
  1. [9]

    Dataset Construction to Explore Chemical Space with 3D Geometry and Deep Learning

    Jianing Lu, Song Xia, Jieyu Lu, and Yingkai Zhang. Dataset Construction to Explore Chemical Space with 3D Geometry and Deep Learning. Journal of Chemical Information and Modeling, 61(3):1095–1104, March 2021

    2021

  2. [10]

    R. M. Shields, B. Temelso, K. A. Archer, T. E. Morrell, and G. C. Shields. Accurate predictions of water cluster formation, (h 2o)n = 2–10. J. Phys. Chem. A, 114(43):11725–11737, 2010

    2010

  3. [11]

    Temelso, K

    B. Temelso, K. A. Archer, and G. C. Shields. Benchmark structures and binding energies of small water clusters with anharmonicity corrections. J. Phys. Chem. A, 115(43):12034–12046, 2011

    2011

  4. [12]

    D. J. Wales and J. P. K. Doye. Global optimization by basin-hopping and the lowest energy structures of Lennard–Jones clusters containing up to 110 atoms. J. Phys. Chem. A, 101:5111– 5116, 1997

    1997

  5. [13]

    D. J. Wales, J. P. K. Doye, A. Dullweber, M. P. Hodges, F. Y. Naumkin, F. Calvo, J. Hern´ andez-Rojas, and T. F. Middleton. The Cambridge Cluster Database. http: //www-wales.ch.cam.ac.uk/CCD.html. Accessed March 1, 2015

    2015

  6. [14]

    Xiang, W

    Y. Xiang, W. Cai, H. Jiang, and X. Shao. An efficient method based on lattice construction and the genetic algorithm for optimization of large Lennard–Jones clusters. J. Phys. Chem. A, 108(16):3586–3592, 2004

    2004

  7. [15]

    Xiang, L

    Y. Xiang, L. Cheng, W. Cai, and X. Shao. Structural distribution of Lennard–Jones clusters containing 562 to 1000 atoms. J. Phys. Chem. A, 108:9516–9520, 2004

    2004

  8. [16]

    ArbAlign: a tool for optimal alignment of arbitrarily ordered isomers using the Kuhn–Munkres algorithm

    Berhane Temelso, Joel M Mabey, Toshiro Kubota, Nana Appiah-Padi, and George C Shields. ArbAlign: a tool for optimal alignment of arbitrarily ordered isomers using the Kuhn–Munkres algorithm. Journal of chemical information and modeling, 57(5):1045–1054, 2017

    2017

  9. [17]

    A solution for the best rotation to relate two sets of vectors

    Wolfgang Kabsch. A solution for the best rotation to relate two sets of vectors. Foundations of Crystallography, 32(5):922–923, 1976. 24

    1976

  10. [18]

    Using quaternions to calculate RMSD

    Evangelos A Coutsias, Chaok Seok, and Ken A Dill. Using quaternions to calculate RMSD. Journal of computational chemistry, 25(15):1849–1857, 2004

    2004

  11. [19]

    Implementation of the Hungarian algorithm to account for ligand symmetry and similarity in structure-based design

    William J Allen and Robert C Rizzo. Implementation of the Hungarian algorithm to account for ligand symmetry and similarity in structure-based design. Journal of chemical information and modeling, 54(2):518–529, 2014

    2014

  12. [20]

    Similarity recognition of molecular structures by op- timal atomic matching and rotational superposition

    Benjamin Helmich and Marek Sierka. Similarity recognition of molecular structures by op- timal atomic matching and rotational superposition. Journal of Computational Chemistry, 33(2):134–140, 2012

    2012

  13. [21]

    How differ- ent are two chemical structures? Journal of Chemical information and Modeling, 50(12):2129– 2140, 2010

    Jorge MC Marques, JL Llanio-Trujillo, PE Abreu, and Francisco Baptista Pereira. How differ- ent are two chemical structures? Journal of Chemical information and Modeling, 50(12):2129– 2140, 2010

    2010

  14. [22]

    Alignment of three-dimensional molecules using an image recognition algorithm

    Nicola J Richmond, Peter Willett, and Robert D Clark. Alignment of three-dimensional molecules using an image recognition algorithm. Journal of Molecular Graphics and Modelling, 23(2):199–209, 2004

    2004

  15. [23]

    Metrics for measuring distances in configuration spaces

    Ali Sadeghi, S Alireza Ghasemi, Bastian Schaefer, Stephan Mohr, Markus A Lill, and Stefan Goedecker. Metrics for measuring distances in configuration spaces. The Journal of chemical physics, 139(18), 2013

    2013

  16. [24]

    The discovery of unexpected isomers in sodium heptamers by Born–Oppenheimer molecular dynamics

    Jos´ e Manuel V´ asquez-P´ erez, Gabriel Ulises Gamboa Mart ´ ınez, Andreas M K¨ oster, and Patrizia Calaminici. The discovery of unexpected isomers in sodium heptamers by Born–Oppenheimer molecular dynamics. The Journal of chemical physics, 131(12), 2009

    2009

  17. [25]

    aRMSD: a comprehensive tool for structural analysis

    Arne Wagner and Hans-J¨ org Himmel. aRMSD: a comprehensive tool for structural analysis. Journal of Chemical Information and Modeling, 57(3):428–438, 2017

    2017

  18. [26]

    Linear assignment problems and extensions

    Rainer E Burkard and Eranda Cela. Linear assignment problems and extensions. In Handbook of combinatorial optimization: Supplement volume A, pages 75–149. Springer, 1999

    1999

  19. [27]

    The Hungarian method for the assignment problem

    Harold W Kuhn. The Hungarian method for the assignment problem. Naval research logistics quarterly, 2(1-2):83–97, 1955

    1955

  20. [28]

    Algorithms for the assignment and transportation problems

    James Munkres. Algorithms for the assignment and transportation problems. Journal of the society for industrial and applied mathematics, 5(1):32–38, 1957

    1957

  21. [29]

    A shortest augmenting path algorithm for dense and sparse linear assignment problems

    Roy Jonker and Ton Volgenant. A shortest augmenting path algorithm for dense and sparse linear assignment problems. In DGOR/NSOR: Papers of the 16th Annual Meeting of DGOR in Cooperation with NSOR/Vortr¨ age der 16. Jahrestagung der DGOR zusammen mit der NSOR, pages 622–622. S...

    1988

  22. [30]

    Gromov–Wasserstein distances and the metric approach to object matching

    Facundo M´ emoli. Gromov–Wasserstein distances and the metric approach to object matching. Foundations of computational mathematics, 11:417–487, 2011

    2011

  23. [31]

    Gromov-wasserstein averaging of kernel and distance matrices

    Gabriel Peyr´ e, Marco Cuturi, and Justin Solomon. Gromov-wasserstein averaging of kernel and distance matrices. In International conference on machine learning, pages 2664–2672. PMLR, 2016

    2016

  24. [32]

    Supervised optimal transport

    Zixuan Cang, Qing Nie, and Yanxiang Zhao. Supervised optimal transport. SIAM Journal on Applied Mathematics, 82(5):1851–1877, 2022. 25

    2022

  25. [33]

    Supervised Gromov–Wasserstein Optimal Trans- port with Metric-Preserving Constraints

    Zixuan Cang, Yaqi Wu, and Yanxiang Zhao. Supervised Gromov–Wasserstein Optimal Trans- port with Metric-Preserving Constraints. SIAM Journal on Mathematics of Data Science, 7(1):301–328, 2025

    2025

  26. [34]

    Computational optimal transport: With applications to data science

    Gabriel Peyr´ e, Marco Cuturi, et al. Computational optimal transport: With applications to data science. Foundations and Trends® in Machine Learning, 11(5-6):355–607, 2019

    2019

  27. [35]

    Optimal transport for structured data with application on graphs

    Vayer Titouan, Nicolas Courty, Romain Tavenard, and R´ emi Flamary. Optimal transport for structured data with application on graphs. In International Conference on Machine Learning, pages 6275–6284. PMLR, 2019

    2019

  28. [36]

    A purely algebraic justification of the Kabsch-Umeyama algorithm

    Jim Lawrence, Javier Bernal, and Christoph Witzgall. A purely algebraic justification of the Kabsch-Umeyama algorithm. Journal of research of the National Institute of Standards and Technology, 124:1, 2019

    2019

  29. [37]

    Least-squares estimation of transformation parameters between two point patterns

    Shinji Umeyama. Least-squares estimation of transformation parameters between two point patterns. IEEE Transactions on Pattern Analysis & Machine Intelligence, 13(04):376–380, 1991

    1991

  30. [38]

    Displacement interpolation using Lagrangian mass transport

    Nicolas Bonneel, Michiel Van De Panne, Sylvain Paris, and Wolfgang Heidrich. Displacement interpolation using Lagrangian mass transport. In Proceedings of the 2011 SIGGRAPH Asia conference, pages 1–12, 2011

    2011

  31. [39]

    Atlas of putative minima and low-lying energy networks of water clusters n= 3–25

    Avijit Rakshit, Pradipta Bandyopadhyay, Joseph P Heindel, and Sotiris S Xantheas. Atlas of putative minima and low-lying energy networks of water clusters n= 3–25. The Journal of chemical physics, 151(21), 2019

    2019

  32. [40]

    Alaya, Aur´ elie Boisbunon, Stanislas Chambon, Laetitia Chapel, Adrien Corenflos, Kilian Fatras, Nemo Fournier, L´ eo Gautheron, Nathalie T.H

    R´ emi Flamary, Nicolas Courty, Alexandre Gramfort, Mokhtar Z. Alaya, Aur´ elie Boisbunon, Stanislas Chambon, Laetitia Chapel, Adrien Corenflos, Kilian Fatras, Nemo Fournier, L´ eo Gautheron, Nathalie T.H. Gayraud, Hicham Janati, Alain Rakotomamonjy, Ievgen Redko, An- toine Ro...

    2021

  33. [41]

    POT Python Optimal Transport (version 0.9.5), 2024

    R´ emi Flamary, C´ edric Vincent-Cuaz, Nicolas Courty, Alexandre Gramfort, Oleksii Kachaiev, Huy Quang Tran, Laur` ene David, Cl´ ement Bonet, Nathan Cassereau, Th´ eo Gnassounou, Eloi Tanguy, Julie Delon, Antoine Collas, Sonia Mazelet, Laetitia Chapel, Tanguy Kerdoncuff, Xizh...

    2024

  34. [42]

    Hara and Akihiko Kondo

    Kiyotaka Y. Hara and Akihiko Kondo. ATP regulation in bioproduction. Microbial Cell Factories, 14(1):198, 2015

    2015

  35. [43]

    Distinct Conformation of ATP Molecule in Solution and on Protein

    Eri Kobayashi, Kei Yura, and Yoshinori Nagai. Distinct Conformation of ATP Molecule in Solution and on Protein. BIOPHYSICS, 9:1–12, 2013

    2013

  36. [44]

    Michael Deininger, Elisabeth Buchdunger, and Brian J. Druker. The development of imatinib as a therapeutic agent for chronic myeloid leukemia. Blood, 105(7):2640–2653, 2005

    2005

  37. [45]

    Imatinib: A Breakthrough of Targeted Therapy in Cancer

    Nida Iqbal and Naveed Iqbal. Imatinib: A Breakthrough of Targeted Therapy in Cancer. Chemotherapy Research and Practice, 2014(1):357027, 2014. 26

    2014

  38. [46]

    Important Roles of Water Clusters Confined in a Nanospace as Revealed by a Synchrotron X-ray Diffraction Study

    Tomohiko Okada, Kana Izumi, Shogo Kawaguchi, Chikako Moriyoshi, Takuya Fujimura, Ryo Sasai, and Makoto Ogawa. Important Roles of Water Clusters Confined in a Nanospace as Revealed by a Synchrotron X-ray Diffraction Study. Langmuir, 37(35):10469–10480, 2021. PMID: 34427085

    2021

  39. [47]

    Pouliquen

    Daniel L. Pouliquen. The biophysics of water in cell biology: perspectives on a keystone for both marine sciences and cancer research. Frontiers in Cell and Developmental Biology, Volume 12 - 2024, 2024. 27

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.