Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

A Physics-informed Sheaf Model

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read By modeling a protein's interaction network as a cellular sheaf, this paper proves that the ANM Hessian is a sheaf Laplacian and that its six rigid-body zero modes are exactly the sheaf's global sections.

desk verdict Solid sheaf/Hessian identification and a useful Delaunay six-mode theorem, but Algorithm 1's existence step is unproven and needs either a real proof or a downgrade. read the letter →

arxiv 2501.06197 v1 pith:W45XSX7Y submitted 2024-12-27 physics.chem-ph math.AT

Chuan-Shen Hu , Xiang Liu , Kelin Xia This is my paper · ORCID
classification physics.chem-phmath.AT MSC 55N3055U1005C5092C40
keywords normalmodeanalysisanisotropicnetworkmodelcellularsheafLaplacianglobalsectionsDelaunaytriangulationzeromodeselastic
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

This paper tries to give normal mode analysis a rigorous geometric foundation by recasting it in sheaf theory. It introduces a cellular sheaf, called the anisotropic sheaf, on the graph of atoms and interactions, and proves that with edge weights $w_{ij}=\gamma^{1/2}/s_{ij}$ the sheaf Laplacian is exactly the ANM Hessian. Under this identification, the six zero-eigenvalue rigid motions become the sheaf's global sections: local atomic displacements that agree consistently along every edge. The paper further proves that any point cloud in general position admits admissible triangulations, including the 3D Delaunay triangulation, whose 1-skeletons induce Hessians with exactly six zero modes, and gives an algorithm for an edge-minimal such graph. If true, this replaces the usual cutoff-distance heuristic in anisotropic network models with a topology-backed construction guarantee.

What carries the argument

The carrying object is the anisotropic sheaf: vertices get stalk $\mathbb{R}^3$, edges get stalk $\mathbb{R}$, and each edge's restriction map is the rank-one row vector pointing along the edge, scaled by $w_{ij}$. The 0-th sheaf coboundary matrix $C$ has one row per edge with signed copies of these maps, and the sheaf Laplacian $\Delta_0 = C^T C$ is what the proof works with. Its kernel is $H^0$, the global section space, so the six rigid motions appear as harmonic signals on the network. The dimension count $3|V|-\operatorname{rank}(C)$ then turns the six-mode question into a rank condition on $C$, and the admissibility and Delaunay arguments are precisely constructions that force $\operatorname{rank}(C) = 3|V|-6$.

What would settle it

Take five points in $\mathbb{R}^3$ in general position: an initial tetrahedron and a fifth point that is strictly closest to the tetrahedron but positioned so that every boundary triangle would make the new tetrahedron intersect the existing complex in more than that triangle; if Algorithm 1 cannot complete step 6 for this input, the claimed minimal-graph guarantee fails for that configuration. The nullity of the resulting Hessian can then be computed directly and compared with six.

Watch

Extended reading notes

Core claim

The central discovery is a dictionary between the anisotropic network model and cellular sheaf cohomology. For a molecular graph with coordinates, each vertex carries a stalk $\mathbb{R}^3$, each edge carries a stalk $\mathbb{R}$, and the restriction map along an edge is the row vector $w_{ij}(x_j-x_i,\,y_j-y_i,\,z_j-z_i)$. The 0-th coboundary matrix $C$ is built from these signed edge maps, and its sheaf Laplacian $C^T C$ equals the ANM Hessian precisely when $w_{ij}=\gamma^{1/2}/s_{ij}$ (Theorem 3.1). Consequently, by the Hodge identification of $H^0$ with the kernel of the sheaf Laplacian, the zero modes of the Hessian are exactly the global sections of the sheaf, with the six rigid translations and rotations forming a canonical basis (Theorems 2.3 and 4.1). The paper then characterizes when the global section space has dimension exactly six: for a point cloud in general position, the 1-skeleton of any admissible homogeneous 3-complex, in particular the 3D Delaunay triangulation, induces an ANM Hessian with exactly six zero eigenvalues (Theorem 5.6, Corollary 5.7). Algorithm 1 constructs a graph with $3|V|-6$ edges that is minimal with this property, and deleting any edge raises the nullity above six.

Load-bearing premise

The load-bearing premise is that the greedy attachment step in Algorithm 1 can always find a boundary triangle such that the new tetrahedron touches the existing complex only along that triangle; the paper argues this from a separation lemma that is proved only for a single simplex, so the guarantee for a growing multi-tetrahedron complex is assumed rather than fully established.

Editorial extensions

If this is right

  • Any ANM built from the 1-skeleton of a Delaunay triangulation of a protein's $C_\alpha$ atoms in general position has exactly six trivial modes, with no cutoff distance to tune.
  • The nullity of an ANM Hessian is a sheaf-cohomology invariant, so adding or removing edges changes the number of trivial modes in a controlled way: a subgraph cannot have fewer trivial modes than the graph it sits in.
  • The greedy Algorithm 1 produces a graph with $3|V|-6$ edges that is minimal for the six-mode property; removing any edge yields at least seven zero eigenvalues.
  • Because the sheaf-to-Hessian equality holds for any nonzero edge weights, the six-mode guarantee is independent of the $\gamma^{1/2}/s_{ij}$ scaling used in practice.
  • The six zero-mode eigenvectors can be written down explicitly as translation and rotation vector fields on the atomic coordinates, without diagonalizing the Hessian (Theorem 4.1).

Reading between the lines

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

  • Beyond the paper, this dictionary suggests that the low-frequency nonzero modes of an ANM Hessian could be read as the first obstructions to gluing in sheaf cohomology, connecting vibrational softness to cohomological invariants; the paper does not develop this.
  • The minimal-graph construction may extend to non-protein point clouds, such as ligand or material configurations, where the same six-mode guarantee is desired, but Algorithm 1's greedy attachment step would need empirical testing on such distributions.
  • The Delaunay admissibility proof (Corollary 5.7) and the greedy Algorithm 1 construction are logically independent; a counterexample to the greedy attachment step would not automatically falsify the Delaunay-based six-mode claim.
  • Because the sheaf Laplacian is $C^T C$, adding edges imposes interlacing-type constraints on the nonzero spectrum; a testable extension is that Delaunay-based ANM low-frequency modes converge to complete-graph ANM modes as the point cloud is refined.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 paper introduces a cellular sheaf, the 'anisotropic sheaf,' on an undirected graph whose vertices carry position data in R3, and shows that its 0-th sheaf Laplacian equals the Hessian matrix of the anisotropic network model (ANM) when edge weights are chosen as gamma^{1/2}/s_ij (Theorem 3.1). It then identifies the kernel of this Laplacian, equivalently the space of global sections of the sheaf, with the zero-eigenvalue normal modes of the ANM, proves that this space has dimension at least six for any non-collinear configuration (Theorem 4.1), analyzes the rank of the coboundary map for complete graphs (Appendix A), and proves existence of an edge-minimal graph with exactly six zero modes (Theorem 4.5). The final section introduces 'admissible' homogeneous simplicial 3-complexes, proves that the 1-skeleton of any such complex induces a Hessian with exactly six zero eigenvalues (Theorem 5.6), derives that every 3D Delaunay triangulation of a general-position point cloud is admissible (Corollary 5.7), and gives an incremental algorithm (Algorithm 1) intended to construct a minimal graph with exactly six zero modes.

Significance. If the main results hold, the paper gives a clean sheaf-theoretic reformulation of ANM: the Hessian is a sheaf Laplacian, and the six trivial modes are exactly the global sections of the anisotropic sheaf. The explicit identification in Theorem 3.1 is a direct, checkable computation, and Theorem 4.1 provides an explicit basis of six global sections. The Delaunay-based result (Corollary 5.7) is potentially useful in practice, as it gives a topological guarantee that the 1-skeleton of a Delaunay triangulation produces the desired six zero modes without a cutoff-distance search. The connection to force cosheaves and Maxwell's rule is also suggestive. However, the paper's constructive claims, especially Algorithm 1, are not yet supported by the arguments given. The central sheaf-Hessian equivalence is solid, but the algorithmic contribution as stated has serious gaps.

major comments (3)
  1. [§5, Algorithm 1 and Eq. (19)] As written, Eq. (19) cannot be satisfied: v_i is a newly chosen point and is not a vertex of K_{i-1}, so the right-hand side {v_j1, v_j2, v_j3, v_i} cannot equal the intersection with K_{i-1}. If the intended condition is conv(v_j1,v_j2,v_j3,v_i) ∩ K_{i-1} = {v_j1,v_j2,v_j3}, or equivalently that the new tetrahedron meets K_{i-1} only in the 2-simplex σ, then the existence of such a 2-simplex is not established. The citation to Corollary B.6 is insufficient because that result separates a point from a single 3-simplex, not from a multi-tetrahedron complex, and K_{i-1} is not shown to be convex. More seriously, the nearest-point rule in step 5 can select an unprocessed point lying in the interior of the current tetrahedral complex (distance zero); for such a point no boundary 2-simplex satisfies the intended non-overlap condition. Thus Algorithm 1 does not provide the promised construction for a general-position point cloud, and the six-mode and minimality conclusions attributed to Algorithm 1 are unsupported.
  2. [§5, Theorem 5.6, rank-increase step] The proof of Theorem 5.6 relies on the assertion that 'each vertex in V2 \ V1 contributes at least three additional linearly independent rows to the extension matrix C(2) from matrix C(1).' This is a load-bearing rank-increase lemma, but it is only stated, not proved. While the statement is plausible and can be justified by considering the three new rows whose nonzero entries at the new vertex are linearly independent vectors (because the shared face is a nondegenerate 2-simplex), the proof as written is a one-sentence assertion rather than an argument. This gap should be fixed by stating and proving the lemma explicitly, since Theorem 5.6 depends on it at every filtration step.
  3. [§B, Proposition 5.5] In the proof that a Delaunay triangulation is admissible, the finite walk across 2-adjacent tetrahedra is terminated by the statement 'This process terminates because L is a finite simplicial complex.' Finiteness alone does not rule out a cycle in the dual graph of the component L. A rigorous termination argument is needed, for example by showing that the chosen separating face moves the walk monotonically with respect to a potential such as distance from the point x0, or by using a boundary-adjacency tree argument. Without this, the proof that a boundary face with deg_L,u(τ)=1 is reached is incomplete. This matters because Proposition 5.5 is the basis for Corollary 5.7.
minor comments (4)
  1. [§3, section heading] The heading 'Anisotropic network and Hassien matrix' contains a typo; it should read 'Hessian matrix.'
  2. [§2, Eq. (9) and surrounding text] The convention Fvi,[vj,vk] = [0 0 0] is written with a single zero row; it may be clearer to write the 1×3 zero matrix explicitly, and to state that this convention is used only for ordered pairs where the vertex is not a face of the edge.
  3. [Figure 2 caption] The caption contains a stray expression 'and = Fvj,[vi,vj]'; the equality should be written cleanly as Fvi,[vi,vj] = Fvj,[vi,vj].
  4. [§4, Theorem 4.1 proof] In the proof of linear independence of the six global sections, the phrase 'without loss of generality' is used to assume that the first three vertices are affinely independent; this is fine but should be stated explicitly before the matrix calculation, since the displayed matrix uses vertices 1, 2, 3.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the sheaf/Hessian equality is an explicit weight-chosen reformulation, not a fitted or self-referential prediction.

full rationale

The paper's central identity (Theorem 3.1) is an explicit coordinate identification: with the anisotropic sheaf's restriction maps fixed as w_ij times the displacement vector and w_ij = gamma^{1/2}/s^o_ij, the sheaf Laplacian L_F = C^T C equals the ANM Hessian block-by-block. The paper states this as an equality after choosing weights, not as a prediction from fitted data; the equality is a reformulation of the ANM Hessian in sheaf language, so it does not reduce a derived conclusion to its own input. No parameter is fitted to any subset of data and then called a prediction. The >=6 global-section bound (Theorem 4.1) and the exact-6 results (Theorem 5.6, Corollary 5.7) are proved by rank computations and induction from non-degeneracy/general-position assumptions, with Hodge theory imported from Hansen and Ghrist as external support. The only flagged weakness, Algorithm 1 step 6's existence assertion relying on Corollary B.6 for a multi-tetrahedron complex, is a correctness gap and a possible missing proof, not a circularity: it does not assume the conclusion it is meant to establish. Self-citations are background references for ANM settings and are not load-bearing. Therefore no circular step is exhibited.

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

The central claims rest on standard sheaf-theoretic facts, the standard ANM model, and general position; two additional assertions (rank-increase per added vertex and Algorithm 1's face-existence) are load-bearing but not fully proven. No parameters are fitted to data in this paper.

assumptions (6)
  • standard math Cellular sheaf Hodge theorem: for a finite cellular sheaf, ker(delta_0) is the 0-th cohomology and global section space.
    Invoked as Theorem 2.3, citing Hansen and Ghrist [36]; foundational for identifying zero modes with global sections.
  • domain assumption ANM potential and weight choice w_ij = gamma^{1/2}/s_ij.
    Section 3, Eq. (8) and Theorem 3.1; inherited from the ANM literature [3]. The equality H_ANM = L_F depends on this weight assignment.
  • domain assumption Point cloud in R^3 is in general position and has at least three affinely independent coordinates.
    Used in Theorem 4.1, Theorem 5.6, Corollary 5.7, and Algorithm 1; degeneracies (coplanar or collinear points) break the exact six-mode count.
  • domain assumption The underlying graph is the 1-skeleton of a homogeneous simplicial 3-complex whose tetrahedra are connected via shared 2-faces (admissible).
    Definition 5.4; the proof of Theorem 5.6 uses 2-face connectivity to propagate the rank argument through the filtration.
  • ad hoc to paper In Theorem 5.6, each newly added vertex contributes at least three linearly independent new rows to the coboundary matrix.
    Asserted in the proof of Theorem 5.6 without full justification; load-bearing for rank(C(2))=3|V2|-6.
  • ad hoc to paper Algorithm 1 can always find a 2-simplex satisfying Eq. (19) at each step.
    Paragraph after Algorithm 1; existence is asserted using Corollary B.6 and the minimal-distance property, but not proven for a multi-tetrahedron complex.
invented entities (2)
  • anisotropic sheaf
    purpose: Models the ANM spring network as a cellular sheaf over a graph; its Laplacian equals the ANM Hessian.
    New definition in Section 3; a mathematical reformulation with no direct falsifiable handle outside the paper.
  • admissible simplicial complex
    purpose: Defines the class of triangulated 3-complexes whose 1-skeleton is guaranteed to produce six zero modes.
    New definition (Def. 5.4) used to state Theorem 5.6; its biological validity is not independently testable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Physics-informed Sheaf Model." pith.science (2026). https://pith.science/paper/W45XSX7Y

@misc{pith2026250106197,
  author       = {Pith},
  title        = {Pith review of: A Physics-informed Sheaf Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W45XSX7Y}},
  note         = {Machine review of arXiv:2501.06197}
}
read the original abstract

Normal mode analysis (NMA) provides a mathematical framework for exploring the intrinsic global dynamics of molecules through the definition of an energy function, where normal modes correspond to the eigenvectors of the Hessian matrix derived from the second derivatives of this function. The energy required to 'trigger' each normal mode is proportional to the square of its eigenvalue, with six zero-eigenvalue modes representing universal translation and rotation, common to all molecular systems. In contrast, modes associated with small non-zero eigenvalues are more easily excited by external forces and are thus closely related to molecular functions. Inspired by the anisotropic network model (ANM), this work establishes a novel connection between normal mode analysis and sheaf theory by introducing a cellular sheaf structure, termed the anisotropic sheaf, defined on undirected, simple graphs, and identifying the conventional Hessian matrix as the sheaf Laplacian. By interpreting the global section space of the anisotropic sheaf as the kernel of the Laplacian matrix, we demonstrate a one-to-one correspondence between the zero-eigenvalue-related normal modes and a basis for the global section space. We further analyze the dimension of this global section space, representing the space of harmonic signals, under conditions typically considered in normal mode analysis. Additionally, we propose a systematic method to streamline the Delaunay triangulation-based construction for more efficient graph generation while preserving the ideal number of normal modes with zero eigenvalues in ANM analysis.

Figures

Figures reproduced from arXiv: 2501.06197 by the authors.

Figure 1
Figure 1. Illustration of the eigenmodes with zero and non-zero eigenvalues for chain A of the protein with ID 1URP [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. An illustration of two global sections of an anisotropic sheaf defined on a graph [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Illustration of six linearly independent eigenmodes with zero eigenvalues for the protein with ID 103M from [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Cellular Sheaves on Higher-Dimensional Structures

    math.AT 2025-05 conditional novelty 4.0 of 10

    A collection of explicit constructions for cellular sheaves on simplicial complexes of dimension two and higher, mixing anisotropic network models with algebraic sheaves of ideals and modules.

Reference graph

Works this paper leans on

71 extracted references · 47 canonical work pages · cited by 1 Pith paper

  1. [1]

    Alvarez-Garcia and X

    D. Alvarez-Garcia and X. Barril. Relationship between protein flexibility and binding: Lessons for structure-based drug design. Journal of Chemical Theory and Computation, 10(6):2608–2614, 2014

    work page 2014

  2. [2]

    Artin, G

    M. Artin, G. Cornell, C. Chai, J. Silverman, C. Chinburg, G. Faltings, B. Gross, F. McGuiness, J. Milne, M. Rosen, et al. Arithmetic Geometry. Springer New York, 2012

    work page 2012

  3. [3]

    A. R. Atilgan, S. R. Durell, R. L. Jernigan, M. C. Demirel, O. Keskin, and I. Bahar. Anisotropy of fluctuation dynamics of proteins with an elastic network model. Biophysical journal, 80(1):505–515, 2001

    work page 2001

  4. [4]

    Bahar, A

    I. Bahar, A. R. Atilgan, M. C. Demirel, and B. Erman. Vibrational dynamics of folded proteins: significance of slow and fast motions in relation to function and stability. Physical review letters, 80(12):2733–2736, 1998

    work page 1998

  5. [5]

    Bahar, A

    I. Bahar, A. R. Atilgan, and B. Erman. Direct evaluation of thermal fluctuations in proteins using a single-parameter harmonic potential. Folding and Design, 2(3):173–181, 1997

    1997

  6. [6]

    Barbero, C

    F. Barbero, C. Bodnar, H. S. de Ocáriz Borde, and P. Lio. Sheaf attention networks. In NeurIPS 2022 Workshop on Symmetry and Geometry in Neural Representations, 2022

    2022

  7. [7]

    Battiloro, Z

    C. Battiloro, Z. Wang, H. Riess, P. Di Lorenzo, and A. Ribeiro. Tangent bundle filters and neural networks: From manifolds to cellular sheaves and back. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2023

    2023

  8. [8]

    H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig, I. N. Shindyalov, and P. E. Bourne. The protein data bank. Nucleic acids research, 28(1):35–242, 2000

    work page 2000

Show all 71 references
  1. [9]

    Björkman and S

    A. Björkman and S. L. Mowbray. Multiple open forms of ribose-binding protein trace the path of its conformational change. Journal of Molecular Biology, 279(3):651–664, 1998

    1998

  2. [10]

    Bodnar, F

    C. Bodnar, F. Di Giovanni, B. Chamberlain, P. Lio, and M. Bronstein. Neural sheaf diffusion: A topological perspective on heterophily and oversmoothing in GNNs. Advances in Neural Information Processing Systems, 35:18527–18541, 2022

    2022

  3. [11]

    Braithwaite, I

    L. Braithwaite, I. Duta, and P. Liò. Heterogeneous sheaf neural networks. arXiv preprint arXiv:2409.08036, 2024

    2024 arXiv

  4. [12]

    G. E. Bredon. Sheaf theory, volume 170. Springer Science & Business Media, 2012

    2012

  5. [13]

    B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S. a. Swaminathan, and M. Karplus. CHARMM: a program for macromolecular energy, minimization, and dynamics calculations. Journal of computational chemistry, 4(2):187–217, 1983

    1983

  6. [14]

    Bu and D

    Z. Bu and D. J. Callaway. Proteins move! Protein dynamics and long-range allostery in cell signaling. Advances in protein chemistry and structural biology, 83:163–221, 2011

    2011

  7. [15]

    F. H. Caralt, G. Bernardez, I. Duta, E. Alarcon, and P. Lio. Joint diffusion processes as an inductive bias in sheaf neural networks. In ICML 2024 Workshop on Geometry-grounded Representation Learning and Generative Modeling, 2024

    2024

  8. [16]

    Cooperband

    Z. Cooperband. Cellular Cosheaves, Graphic Statics, and Mechanics. PhD thesis, University of Pennsylvania, 2024

    2024

  9. [17]

    Cooperband and R

    Z. Cooperband and R. Ghrist. Towards homological methods in graphic statics. Journal of the International Association for Shell and Spatial Structures, 64(4):266–277, 2023

    2023

  10. [18]

    Cooperband, R

    Z. Cooperband, R. Ghrist, and J. Hansen. A cosheaf theory of reciprocal figures: Planar and higher genus graphic statics. arXiv preprint arXiv:2311.12946, 2023. 18 A PREPRINT - JANUARY 14, 2025

    2023 arXiv

  11. [19]

    Cooperband, M

    Z. Cooperband, M. Lopez, and B. Schulze. Equivariant cosheaves and finite group representations in graphic statics. arXiv preprint arXiv:2401.09392, 2024

    2024 arXiv

  12. [20]

    Cui and I

    Q. Cui and I. Bahar. Normal Mode Analysis: Theory and Applications to Biological and Chemical Systems . Chapman & Hall/CRC Mathematical Biology Series. CRC Press, 2005

    2005

  13. [21]

    Curry, R

    J. Curry, R. Ghrist, and V . Nanda. Discrete morse theory for computing cellular sheaf cohomology.Foundations of Computational Mathematics, 16:875–897, 2016

    2016

  14. [22]

    J. M. Curry. Sheaves, cosheaves and applications. PhD thesis, University of Pennsylvania, 2014

    2014

  15. [23]

    J. M. Curry. Topological data analysis and cosheaves. Japan Journal of Industrial and Applied Mathematics, 32:333–371, 2015

    2015

  16. [24]

    Delaunay, S

    B. Delaunay, S. Vide, A. Lamémoire, and V . De Georges. Bulletin de l’Academie des Sciences de l’URSS.Classe des sciences mathématiques et naturelles, 6:793–800, 1934

    1934

  17. [25]

    Doruker, A

    P. Doruker, A. R. Atilgan, and I. Bahar. Dynamics of proteins predicted by molecular dynamics simulations and analytical approaches: Application to α-amylase inhibitor. Proteins: Structure, Function, and Bioinformatics, 40(3):512–524, 2000

    2000

  18. [26]

    O. K. Dudko, G. Hummer, and A. Szabo. Intrinsic rates and activation free energies from single-molecule pulling experiments. Physical review letters, 96(10):108101, 2006

    2006

  19. [27]

    I. Duta, G. Cassarà, F. Silvestri, and P. Liò. Sheaf hypergraph networks. Advances in Neural Information Processing Systems, 36, 2024

    2024

  20. [28]

    G. Ewald. Combinatorial Convexity and Algebraic Geometry. Graduate Texts in Mathematics. Springer New York, 1996

    1996

  21. [29]

    Eyal, L.-W

    E. Eyal, L.-W. Yang, and I. Bahar. Anisotropic network model: systematic evaluation and a new web interface. Bioinformatics, 22(21):2619–2627, 2006

    2006

  22. [30]

    Fischer, R

    M. Fischer, R. G. Coleman, J. S. Fraser, and B. K. Shoichet. Incorporation of protein flexibility and conformational energy penalties in docking screens to improve ligand discovery. Nature Chemistry, 6:575–583, 2014

    2014

  23. [31]

    J. S. Fraser, M. W. Clarkson, S. C. Degnan, R. Erion, D. Kern, and T. Alber. Hidden alternative structures of proline isomerase essential for catalysis. Nature, 462(7273):669–673, 2009

    2009

  24. [32]

    Ghrist and H

    R. Ghrist and H. Riess. Cellular sheaves of lattices and the Tarski Laplacian. Homology, Homotopy and Applications, 24(1):325–345, 2022

    2022

  25. [33]

    R. W. Ghrist. Elementary applied topology, volume 1. Createspace Seattle, 2014

    2014

  26. [34]

    N. Go, T. Noguti, and T. Nishikawa. Dynamics of a small globular protein in terms of low-frequency vibrational modes. Proceedings of the National Academy of Sciences, 80(12):3696–3700, 1983

    1983

  27. [35]

    Hansen and T

    J. Hansen and T. Gebhart. Sheaf neural networks. arXiv preprint arXiv:2012.06333, 2020

    2012 arXiv

  28. [36]

    Hansen and R

    J. Hansen and R. Ghrist. Toward a spectral theory of cellular sheaves. Journal of Applied and Computational Topology, 3(4):315–358, 2019

    2019

  29. [37]

    Hansen and R

    J. Hansen and R. Ghrist. Opinion dynamics on discourse sheaves. SIAM Journal on Applied Mathematics , 81(5):2033–2060, 2021

    2021

  30. [38]

    Hartshorne

    R. Hartshorne. Algebraic geometry, volume 52. Springer Science & Business Media, 2013

    2013

  31. [39]

    Y . He, C. Bodnar, and P. Lio. Sheaf-based positional encodings for graph neural networks. In NeurIPS 2023 Workshop on Symmetry and Geometry in Neural Representations, 2023

    2023

  32. [40]

    K. Hinsen. Analysis of domain motions by approximate normal mode calculations. Proteins: Structure, Function, and Bioinformatics, 33(3):417–429, 1998

    1998

  33. [41]

    K. Hinsen. Structural flexibility in proteins: impact of the crystal environment. Bioinformatics, 24(4):521–528, 2008

    2008

  34. [42]

    Huang, W

    Y . Huang, W. Lu, J. Robinson, Y . Yang, M. Zhang, S. Jegelka, and P. Li. On the stability of expressive positional encodings for graphs. In The Twelfth International Conference on Learning Representations, 2024

    2024

  35. [43]

    Humphrey, A

    W. Humphrey, A. Dalke, and K. Schulten. VMD – visual molecular dynamics. Journal of Molecular Graphics, 14(1):33–38, 1996

    1996

  36. [44]

    D. J. Jacobs, A. J. Rader, L. A. Kuhn, and M. F. Thorpe. Protein flexibility predictions using graph theory. Proteins: Structure, Function, and Bioinformatics, 44(2):150–165, 2001. 19 A PREPRINT - JANUARY 14, 2025

    2001

  37. [45]

    Kashiwara and P

    M. Kashiwara and P. Schapira. Persistent homology and microlocal sheaf theory. Journal of Applied and Computational Topology, 2(1):83–113, 2018

    2018

  38. [46]

    D. A. Kondrashov, A. W. Van Wynsberghe, R. M. Bannen, Q. Cui, and G. N. Phillips. Protein structural variation in computational models and crystallographic data. Structure, 15(2):169–177, 2007

    2007

  39. [47]

    Levitt, C

    M. Levitt, C. Sander, and P. S. Stern. Protein normal-mode dynamics: trypsin inhibitor, crambin, ribonuclease and lysozyme. Journal of molecular biology, 181(3):423–447, 1985

    1985

  40. [48]

    Li and Q

    G. Li and Q. Cui. A coarse-grained normal mode approach for macromolecules: an efficient implementation and application to Ca2+-ATPase. Biophysical Journal, 83(5):2457–2474, 2002

    2002

  41. [49]

    J. A. Marsh and S. A. Teichmann. Protein flexibility facilitates quaternary structure assembly and evolution. PLoS biology, 12(5):e1001870, 2014

    2014

  42. [50]

    J. A. McCammon, B. R. Gelin, and M. Karplus. Dynamics of folded proteins. nature, 267(5612):585–590, 1977

    1977

  43. [51]

    J. R. Munkres. Elements of algebraic topology. CRC Press, 2018

    2018

  44. [52]

    D. D. Nguyen, Z. Cang, and G.-W. Wei. A review of mathematical representations of biomolecular data. Physical Chemistry Chemical Physics, 22(8):4343–4367, 2020

    2020

  45. [53]

    D. D. Nguyen, Z. Cang, K. Wu, M. Wang, Y . Cao, and G.-W. Wei. Mathematical deep learning for pose and binding affinity prediction and ranking in D3R grand challenges. Journal of computer-aided molecular design, 33:71–82, 2019

    2019

  46. [54]

    D. D. Nguyen, K. Gao, M. Wang, and G.-W. Wei. MathDl: mathematical deep learning for D3R grand challenge

  47. [55]

    Journal of computer-aided molecular design, 34:131–147, 2020

    2020

  48. [56]

    Opron, K

    K. Opron, K. L. Xia, and G. W. Wei. Fast and anisotropic flexibility-rigidity index for protein flexibility and fluctuation analysis. Journal of Chemical Physics, 140:234105, 2014

    2014

  49. [57]

    J.-K. Park, R. Jernigan, and Z. Wu. Coarse grained normal mode analysis vs. refined gaussian network model for protein residue-level structural fluctuations. Bulletin of mathematical biology, 75:124–160, 2013

    2013

  50. [58]

    H. Pei, B. Wei, K. C.-C. Chang, Y . Lei, and B. Yang. Geom-GCN: Geometric graph convolutional networks. In International Conference on Learning Representations, 2020

    2020

  51. [59]

    Robinson

    M. Robinson. Topological signal processing. Mathematical Engineering. Springer Berlin Heidelberg, 2014

    2014

  52. [60]

    R. D. Smith. Correlations between bound N-alkyl isocyanide orientations and pathways for ligand binding in recombinant myoglobins. Rice University, 1999

    1999

  53. [61]

    Tama and Y .-H

    F. Tama and Y .-H. Sanejouand. Conformational change of proteins arising from normal mode calculations.Protein engineering, 14(1):1–6, 2001

    2001

  54. [62]

    Tasumi, H

    M. Tasumi, H. Takeuchi, S. Ataka, A. Dwivedi, and S. Krimm. Normal vibrations of proteins: Glucagon. Biopolymers: Original Research on Biomolecules, 21(3):711–714, 1982

    1982

  55. [63]

    J. F. Thompson, B. K. Soni, and N. P. Weatherill. Handbook of grid generation. CRC press, 1998

    1998

  56. [64]

    Wei and G.-W

    X. Wei and G.-W. Wei. Persistent sheaf laplacians. Foundations of Data Science, 2024

    2024

  57. [65]

    R. O. Wells and O. RAYMOND. Differential and complex geometry: origins, abstractions and embeddings . Springer, 2017

    2017

  58. [66]

    F. Xia, D. Tong, L. Yang, D. Wang, S. C. Hoi, P. Koehl, and L. Lu. Identifying essential pairwise interactions in elastic network model using the alpha shape theory. Journal of Computational Chemistry, 35(15):1111–1121, 2014

    2014

  59. [67]

    K. Xia. Multiscale virtual particle based elastic network model (mvp-enm) for normal mode analysis of large-sized biomolecules. Physical Chemistry Chemical Physics, 20(1):658–669, 2018

    2018

  60. [68]

    K. Xia, K. Opron, and G.-W. Wei. Multiscale gaussian network model (mGNM) and multiscale anisotropic network model (mANM). The Journal of chemical physics, 143(20), 2015

    2015

  61. [69]

    L. W. Yang and C. P. Chng. Coarse-grained models reveal functional dynamics–I. elastic network models–theories, comparisons and perspectives. Bioinformatics and Biology Insights, 2:25 – 45, 2008

    2008

  62. [70]

    Z. Ye, K. S. Liu, T. Ma, J. Gao, and C. Chen. Curvature graph network. In International conference on learning representations, 2019

    2019

  63. [71]

    Zhou and H

    W. Zhou and H. Yan. Alpha shape and delaunay triangulation in studies of protein-related interactions. Briefings in bioinformatics, 15(1):54–64, 2014. 20 A PREPRINT - JANUARY 14, 2025 A Anisotropic Sheaves on Complete Graphs To prove that any anisotropic sheaf F : (Kn, ≤) → Ve...

    2014

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.