pith. sign in

arxiv: 2605.29329 · v1 · pith:ZCAGYRPCnew · submitted 2026-05-28 · 🧬 q-bio.QM · cs.LG

Mixing Vector Model for Copolymer Inference via Mixed Integer Linear Programming

Jianshen Zhu , Raveena Rai , Taiyo Sohkawa , Naveed Ahmed Azam , Kazuya Haraguchi , Liang Zhao , Tatsuya Akutsu This is my paper

Pith reviewed 2026-06-29 00:09 UTC · model grok-4.3

classification 🧬 q-bio.QM cs.LG
keywords copolymer inferencemixing vector modelmixed integer linear programminginverse designproperty predictionmonomer descriptorsmachine learning
0
0 comments X

The pith

The mixing vector model represents copolymers as convex combinations of monomer features, enabling MILP-based inverse design with high predictive accuracy.

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

This paper extends a mixed integer linear programming framework for designing molecules to the case of copolymers. It proposes the mixing vector model, in which a copolymer's feature vector is a weighted average of the feature vectors of its monomer components according to their mixing ratios. This simple representation does not need sequence details yet supports the use of standard machine learning predictors and keeps inverse design problems solvable as MILPs. On ten physicochemical property datasets the approach produces test R squared values above 0.7 in nine cases and above 0.9 in six cases. The resulting optimization problems stay practical even when three different monomers are allowed.

Core claim

Under the mixing vector model a copolymer feature vector is represented as a convex combination of MILP-tractable monomer descriptors weighted by the mixing ratio of the constituent monomers. Prediction functions built from this representation using neural networks, reduced quadratic regression, and random forests achieve test R squared exceeding 0.7 for nine of ten datasets and 0.9 for six. The multi-monomer inverse-design MILP instances remain tractable even for three-monomer settings, and an external consistency check confirms that re-computed property values align with the learned predictions.

What carries the argument

The mixing vector model, which encodes a copolymer as a convex combination of its monomer feature vectors weighted by mixing ratios, allowing direct use of MILP solvers for design without sequence information.

Load-bearing premise

Representing a copolymer solely as a convex combination of its constituent monomer descriptors without any sequence-class information suffices to capture the relevant structure-property relationships.

What would settle it

Finding that the property values recomputed from the inferred copolymer structures deviate substantially and systematically from the values predicted by the learned model on the same structures.

Figures

Figures reproduced from arXiv: 2605.29329 by Jianshen Zhu, Kazuya Haraguchi, Liang Zhao, Naveed Ahmed Azam, Raveena Rai, Taiyo Sohkawa, Tatsuya Akutsu.

Figure 1
Figure 1. Figure 1: Overview of the two-phase mol-infer framework. the inverse QSAR/QSPR phase, where the aim is to infer chemical graphs that exhibit specific property values. Given a set of rules, called a topological specification σ, describing the desired abstract structure of the inferred graphs, and a target range [y ∗ , y ∗ ] of the property value, Stage 4 seeks chemical graphs C ∗ satisfying σ and η(f(C ∗ )) ∈ [y ∗ , … view at source ↗
Figure 2
Figure 2. Figure 2: Examples of linear copolymers with different sequence distr [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of Phase 1 (QSAR/QSPR) under the mixing vect [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of Phase 2 (inverse QSAR/QSPR) under the mix [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustrations of the seed graphs for the instances [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Examples of constituent monomers inferred in Stage 4. (a [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Illustrations of the seed graphs for the instances [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Examples of inferred monomers for the EA dataset. (a) Instances Id1, Id2 with target range [y ∗ , y ∗ ] = [4.3, 4.5]. (b) Instances Id1, Id3 with target range [y ∗ , y ∗ ] = [3.9, 4.1]. (c) Instances Id2, Id3 with target range [y ∗ , y ∗ ] = [3.1, 3.3]. copolymer setting, namely, the ability to solve the inverse problem with guaranteed optimality relative to the learned model and the imposed structural con… view at source ↗
Figure 9
Figure 9. Figure 9: Examples of inferred monomers for the IP dataset. (a) Instances Id1, Id2 with target range [y ∗ , y ∗ ] = [5.8, 6.0]. (b) Instances Id1, Id3 with target range [y ∗ , y ∗ ] = [6.0, 6.2]. (c) Instances Id2, Id3 with target range [y ∗ , y ∗ ] = [5.6, 5.8]. therefore provide a useful test bed for examining the practical applicability of the MV model. The computational experiments demonstrate that the proposed … view at source ↗
read the original abstract

A novel two-phase molecule inference framework, mol-infer, has recently been developed to infer chemical graphs with prescribed abstract structures and desired property values through mixed integer linear programming (MILP) under the two-layered model, with guaranteed optimality and exactness relative to the given learned prediction function and structural constraints. In this study, we extend this framework to copolymers by introducing a simple feature representation, called the mixing vector (MV) model. In the proposed model, a copolymer feature vector is represented as a convex combination of MILP-tractable monomer descriptors weighted by the mixing ratio of the constituent monomers. This representation does not require explicit sequence-class information and is therefore naturally compatible with MILP-based inverse design. Under this model, we construct prediction functions for several copolymer property datasets using artificial neural networks, reduced quadratic multiple linear regression, and random forests. The proposed representation achieves practically useful predictive performance across multiple physicochemical property datasets; in particular, the best test R^2 score exceeds 0.7 for nine of the ten datasets and exceeds 0.9 for six datasets. We also formulate a multi-monomer inverse-design problem under the MV representation with a prescribed mixing ratio and show that the resulting MILP instances remain tractable, even for three-monomer settings. Finally, we perform an external consistency check by re-evaluating the inferred candidates and comparing the re-computed property values with those predicted by the learned model. Overall, the proposed framework gives a tractable first step toward model-level exact inverse design of copolymers under the two-layered model.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The manuscript proposes the mixing vector (MV) model for copolymer feature representation as a convex combination of monomer descriptors by mixing ratio. This enables MILP-based inverse design in the two-layered mol-infer framework without needing sequence-class information. Prediction functions built with artificial neural networks, reduced quadratic multiple linear regression, and random forests yield test R² exceeding 0.7 for nine of ten datasets and 0.9 for six. The multi-monomer inverse-design MILPs are tractable for up to three monomers, and an external consistency check is conducted by re-evaluating inferred candidates.

Significance. If the reported results hold, the work offers a tractable approach to model-level exact inverse design for copolymers, extending prior MILP frameworks. The high predictive performance on multiple datasets, demonstration of MILP tractability even in three-monomer cases, and the external consistency check are notable strengths that support practical utility. This could facilitate inverse design in polymer chemistry where composition dominates the properties of interest.

major comments (2)
  1. [Results section (predictive performance)] Details on data splits, feature construction for the monomer descriptors, hyperparameter choices for the ANN, RQMLR, and RF models, and any post-hoc data exclusions are not provided. These are essential to substantiate the test R² claims and assess potential issues like overfitting or selection bias.
  2. [MV model definition] The sufficiency of the composition-only MV representation for the physicochemical properties is assumed without explicit validation against sequence-aware alternatives or discussion of whether the ten datasets exhibit sequence-dependent behaviors. This assumption underpins both the predictive scores and the inverse-design applicability.
minor comments (3)
  1. [Abstract] The abstract could specify the total number of datasets and the range of properties considered for better context.
  2. [Methods] Clarify the exact formulation of the reduced quadratic multiple linear regression and how it differs from standard quadratic regression.
  3. [Inverse design section] Provide more details on the MILP formulation size (e.g., number of variables/constraints) for the three-monomer cases to support the tractability claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and positive recommendation for minor revision. The feedback highlights important areas for improving clarity and reproducibility. We address each major comment below and will incorporate revisions accordingly.

read point-by-point responses

  1. Referee: [Results section (predictive performance)] Details on data splits, feature construction for the monomer descriptors, hyperparameter choices for the ANN, RQMLR, and RF models, and any post-hoc data exclusions are not provided. These are essential to substantiate the test R² claims and assess potential issues like overfitting or selection bias.

    Authors: We agree that these methodological details are essential for reproducibility and to allow assessment of the reported performance. In the revised manuscript, we will expand the Results and/or Methods sections to include: (i) the data splitting strategy (including ratios and whether random or stratified), (ii) the specific monomer descriptors employed and their construction process, (iii) the hyperparameter selection procedure and final values for the ANN, RQMLR, and RF models, and (iv) explicit confirmation that no post-hoc data exclusions were applied beyond standard preprocessing. These additions will directly address concerns regarding potential overfitting or selection bias. revision: yes

  2. Referee: [MV model definition] The sufficiency of the composition-only MV representation for the physicochemical properties is assumed without explicit validation against sequence-aware alternatives or discussion of whether the ten datasets exhibit sequence-dependent behaviors. This assumption underpins both the predictive scores and the inverse-design applicability.

    Authors: The MV model is deliberately formulated as a composition-only representation to ensure compatibility with MILP-based inverse design without requiring sequence-class information, which is frequently unavailable for copolymers. The ten datasets used are standard copolymer property collections where composition is the dominant variable, and the achieved predictive performance supports applicability in this regime. We will add a clarifying paragraph in the manuscript discussing the scope of the MV model, explicitly noting its suitability when composition dominates properties and that sequence-dependent cases would require alternative representations. However, a direct empirical comparison to sequence-aware models is not feasible here, as the datasets lack sequence annotations; such validation would constitute a separate study. revision: partial

Circularity Check

0 steps flagged

No circularity: MV representation and MILP extension are independent of fitted outputs

full rationale

The paper introduces the MV model as a new convex-combination representation explicitly chosen for MILP compatibility, trains standard ML regressors on external datasets, and states that inverse-design optimality holds only relative to the learned functions. No equation reduces a claimed result to its own fitted parameters by construction, no uniqueness theorem is imported from self-citation, and the mol-infer reference is used only as the base framework being extended rather than as load-bearing justification for the new claims. Performance numbers are reported on held-out test data, satisfying the self-contained benchmark criterion.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that a convex combination of monomer descriptors suffices for copolymer property modeling and on the learned parameters of the ML predictors; no free parameters are explicitly introduced beyond those implicit in the ML training.

axioms (1)
  • domain assumption Copolymer properties can be modeled as a convex combination of monomer descriptors weighted by mixing ratio without sequence information
    This is the core modeling choice stated in the abstract for the MV representation.
invented entities (1)
  • Mixing vector (MV) model no independent evidence
    purpose: To enable MILP-compatible feature representation for copolymers
    New representation introduced to bridge monomer descriptors and copolymer inference

pith-pipeline@v0.9.1-grok · 5838 in / 1384 out tokens · 30527 ms · 2026-06-29T00:09:44.985332+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

83 extracted references · 1 canonical work pages

  1. [1]

    Akutsu and H

    T. Akutsu and H. Nagamochi. A mixed integer linear programming fo rmulation to artificial neural networks. In Proceedings of the 2nd International Conference on Informa tion Science and Systems , pages 215–220, 2019

    2019

  2. [2]

    ´Asgeirsson, C

    V. ´Asgeirsson, C. A. Bauer, and S. Grimme. Quantum chemical calculat ion of electron ioniza- tion mass spectra for general organic and inorganic molecules. Chemical Science, 8:4879–4895, 2017

    2017

  3. [3]

    N. A. Azam, R. Chiewvanichakorn, F. Zhang, A. Shurbevski, H. N agamochi, and T. Akutsu. A novel method for the inverse QSAR/QSPR based on artificial neura l networks and mixed inte- ger linear programming with guaranteed admissibility. In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies - BIOINFORMAT...

    2020

  4. [4]

    N. A. Azam, J. Zhu, K. Haraguchi, L. Zhao, H. Nagamochi, and T. Akutsu. Molecular design based on artificial neural networks, integer programming and grid neighbor search. In 2021 IEEE International Conference on Bioinformatics and Biome dicine (BIBM) , pages 360–363. IEEE, 2021

    2021

  5. [5]

    N. A. Azam, J. Zhu, Y. Sun, Y. Shi, A. Shurbevski, L. Zhao, H. Na gamochi, and T. Akutsu. A novel method for inference of acyclic chemical compounds with bou nded branch-height based on artificial neural networks and integer programming. Algorithms for Molecular Biology , 16:1–39, 2021

    2021

  6. [6]

    Y. Bai, L. Wilbraham, B. J. Slater, M. A. Zwijnenburg, R. S. Sprick , and A. I. Cooper. Accelerated discovery of organic polymer photocatalysts for hyd rogen evolution from water through the integration of experiment and theory. Journal of the American Chemical Society , 141(22):9063–9071, 06 2019

    2019

  7. [7]

    W. Bort, D. Mazitov, D. Horvath, F. Bonachera, A. Lin, G. Marc ou, I. Baskin, T. Madzhidov, and A. Varnek. Inverse QSAR: Reversing descriptor-driven pred iction pipeline using attention- based conditional variational autoencoder. Journal of Chemical Information and Modeling , 62(22):5471–5484, 11 2022

    2022

  8. [8]

    Brierley-Croft, P

    S. Brierley-Croft, P. D. Olmsted, P. J. Hine, R. J. Mandle, A. Cha plin, J. Grasmeder, and J. Mattsson. Polymer informatics method for fast and accurate p rediction of the glass tran- sition temperature from chemical structure. Macromolecules, 58(13):6407–6417, 07 2025

    2025

  9. [9]

    H. Cai, H. Zhang, D. Zhao, J. Wu, and L. Wang. FP-GNN: a versat ile deep learning architec- ture for enhanced molecular property prediction. Briefings in Bioinformatics , 23(6):bbac408, 09 2022

    2022

  10. [10]

    Cheng, Y

    Y. Cheng, Y. Gong, Y. Liu, B. Song, and Q. Zou. Molecular design in drug discovery: a comprehensive review of deep generative models. Briefings in Bioinformatics , 22(6):bbab344, 08 2021

    2021

  11. [11]

    Cherkasov, E

    A. Cherkasov, E. N. Muratov, D. Fourches, A. Varnek, I. I. Baskin, M. Cronin, J. Dearden, P. Gramatica, Y. C. Martin, R. Todeschini, et al. QSAR modeling: wher e have you been? where are you going to? Journal of Medicinal Chemistry , 57(12):4977–5010, 2014

    2014

  12. [12]

    J. G. Coldstream, P. J. Camp, D. J. Phillips, and P. J. Dowding. Gr adient copolymers versus block copolymers: self-assembly in solution and surface adsorption . Soft Matter, 18:6538–6549, 2022

    2022

  13. [13]

    E. F. Connor, I. Lees, and D. Maclean. Polymers as drugs—adv ances in therapeutic appli- cations of polymer binding agents. Journal of Polymer Science Part A: Polymer Chemistry , 55(18):3146–3157, 2017

    2017

  14. [14]

    IBM ILOG CPLEX Optimization Studio, 2025

    2025

  15. [15]

    A. Das, T. Ringu, S. Ghosh, and N. Pramanik. A comprehensive r eview on recent advances in preparation, physicochemical characterization, and bioenginee ring applications of biopoly- mers. Polymer Bulletin , 80(7):7247–7312, 2023

    2023

  16. [16]

    P. J. Flory. Principles of polymer chemistry . Cornell university press, 1953. 2LMM copolymer mv v5: May 29, 2026 25

    1953

  17. [17]

    Gao and D

    C. Gao and D. Yan. Hyperbranched polymers: from synthesis t o applications. Progress in Polymer Science , 29(3):183–275, 2004

    2004

  18. [18]

    Fast and uncertainty-aware directional message passing for non-equilibrium molecules.arXiv preprint arXiv:2011.14115,

    J. Gasteiger, S. Giri, J. T. Margraf, and S. G¨ unnemann. Fast and uncertainty-aware directional message passing for non-equilibrium molecules. arXiv:2011.14115, 20 22

    work page arXiv 2011

  19. [19]

    Grimme, C

    S. Grimme, C. Bannwarth, and P. Shushkov. A robust and accu rate tight-binding quantum chemical method for structures, vibrational frequencies, and n oncovalent interactions of large molecular systems parametrized for all spd-block elements (z = 1–8 6). Journal of Chemical Theory and Computation , 13(5):1989–2009, 05 2017

    1989

  20. [20]

    GitHub - grimme-lab/xtb: Semiempirical Extended Tight-Binding P rogram Package, 2024

    2024

  21. [21]

    R. Ido, N. A. Azam, J. Zhu, H. Nagamochi, and T. Akutsu. A dyn amic programming algorithm for generating chemical isomers based on frequency vectors. Scientific Reports , 15(1):22214, 2025

    2025

  22. [22]

    R. Ido, S. Cao, J. Zhu, N. A. Azam, K. Haraguchi, L. Zhao, H. N agamochi, and T. Akutsu. A method for inferring polymers based on linear regression and integ er programming. IEEE/ACM Transactions on Computational Biology and Bioinf ormatics, 21(6):1623–1632, 2024

    2024

  23. [23]

    Ikebata, K

    H. Ikebata, K. Hongo, T. Isomura, R. Maezono, and R. Yoshid a. Bayesian molecular design with a chemical language model. Journal of Computer-aided Molecular Design , 31:379–391, 2017

    2017

  24. [24]

    R. Ito, N. A. Azam, C. Wang, A. Shurbevski, H. Nagamochi, and T. Akutsu. A novel method for the inverse QSAR/QSPR to monocyclic chemical compounds base d on artificial neural networks and integer programming. In Advances in Computer Vision and Computational Biology: Proceedings from IPCV’20, HIMS’20, BIOCOMP’20, a nd BIOENG’20, pages 641–

  25. [25]

    E. A. Jackson and M. A. Hillmyer. Nanoporous membranes derive d from block copolymers: From drug delivery to water filtration. ACS Nano , 4(7):3548–3553, 07 2010

    2010

  26. [26]

    A. Jain, R. Gurnani, A. Rajan, H. J. Qi, and R. Ramprasad. A phy sics-enforced neural network to predict polymer melt viscosity. npj Computational Materials , 11(1):42, 2025

    2025

  27. [27]

    H. Kaneko. Molecular descriptors, structure generation, an d inverse QSAR/QSPR based on SELFIES. ACS Omega, 8(24):21781–21786, 06 2023

    2023

  28. [28]

    A. Khan, L. K. Kian, M. Jawaid, A. A. P. Khan, M. M. Alotaibi, A. M. Asiri, and H. M. Mar- wani. Preparation of styrene-butadiene rubber (SBR) composite incorporated with collagen- functionalized graphene oxide for green tire application. Gels, 8(3), 2022

    2022

  29. [29]

    S. B. Kharchenko, R. M. Kannan, J. J. Cernohous, and S. Ven kataramani. Role of architecture on the conformation, rheology, and orientation behavior of linear, star, and hyperbranched polymer melts. 1. synthesis and molecular characterization. Macromolecules, 36(2):399–406, 01 2003

    2003

  30. [30]

    Kumar, S

    L. Kumar, S. Singh, A. Horechyy, A. Fery, and B. Nandan. Bloc k copolymer template-directed catalytic systems: Recent progress and perspectives. Membranes, 11(5), 2021. 2LMM copolymer mv v5: May 29, 2026 26

    2021

  31. [31]

    M. D. Lefebvre, M. Olvera de la Cruz, and K. R. Shull. Phase segr egation in gradient copoly- mer melts. Macromolecules, 37(3):1118–1123, 02 2004

    2004

  32. [32]

    X. Li, J. Huang, Y. Chen, F. Zhu, Y. Wang, W. Wei, and Y. Feng. P olymer-based electronic packaging molding compounds, specifically thermal performance imp rovement: An overview. ACS Applied Polymer Materials , 6(24):14948–14969, 12 2024

    2024

  33. [33]

    J. W. Lim. Polymer materials for optoelectronics and energy app lications. Materials, 17(15), 2024

    2024

  34. [34]

    Y.-C. Lo, S. E. Rensi, W. Torng, and R. B. Altman. Machine learnin g in chemoinformatics and drug discovery. Drug Discovery Today , 23(8):1538–1546, 2018

    2018

  35. [35]

    M. A. R. Meier and C. Barner-Kowollik. A new class of materials: Se quence-defined macro- molecules and their emerging applications. Advanced Materials, 31(26):1806027, 2019

    2019

  36. [36]

    L. A. Miccio and G. A. Schwartz. From chemical structure to qu antitative polymer properties prediction through convolutional neural networks. Polymer, 193:122341, 2020

    2020

  37. [37]

    Miyao, H

    T. Miyao, H. Kaneko, and K. Funatsu. Inverse QSPR/QSAR ana lysis for chemical structure generation (from y to x). Journal of Chemical Information and Modeling , 56(2):286–299, 2016

    2016

  38. [38]

    D. A. Olson, L. Chen, and M. A. Hillmyer. Templating nanoporous p olymers with ordered block copolymers. Chemistry of Materials , 20(3):869–890, 02 2008

    2008

  39. [39]

    R. A. Patel, C. H. Borca, and M. A. Webb. Featurization strate gies for polymer sequence or composition design by machine learning. Mol. Syst. Des. Eng. , 7:661–676, 2022

    2022

  40. [40]

    M. Reis, F. Gusev, N. G. Taylor, S. H. Chung, M. D. Verber, Y. Z . Lee, O. Isayev, and F. A. Leibfarth. Machine-learning-guided discovery of 19F MRI age nts enabled by automated copolymer synthesis. Journal of the American Chemical Society , 143(42):17677–17689, 10 2021

    2021

  41. [41]

    Rodriguez, C

    F. Rodriguez, C. Cohen, C. K. Ober, and L. Archer. Principles of polymer systems . CRC press, 2014

    2014

  42. [42]

    Rupakheti, A

    C. Rupakheti, A. Virshup, W. Yang, and D. N. Beratan. Strate gy to discover diverse optimal molecules in the small molecule universe. Journal of Chemical Information and Modeling , 55(3):529–537, 2015

    2015

  43. [43]

    J. L. Self, A. J. Zervoudakis, X. Peng, W. R. Lenart, C. W. Mac osko, and C. J. Ellison. Linear, graft, and beyond: Multiblock copolymers as next-genera tion compatibilizers. JACS Au, 2(2):310–321, 02 2022

    2022

  44. [44]

    Y. Shi, J. Zhu, N. A. Azam, K. Haraguchi, L. Zhao, H. Nagamoch i, and T. Akutsu. An inverse QSAR method based on a two-layered model and integer programmin g. International Journal of Molecular Sciences , 22(6):2847, 2021

    2021

  45. [45]

    Shino and H

    Y. Shino and H. Kaneko. Improving molecular design with direct inv erse analysis of QSAR/QSPR model. Molecular Informatics, 44(1):e202400227, 2025

    2025

  46. [46]

    Sinclair, X

    A. Sinclair, X. Zhou, S. Tangpong, D. S. Bajwa, M. Quadir, and L . Jiang. High-performance styrene-butadiene rubber nanocomposites reinforced by surfa ce-modified cellulose nanofibers. ACS Omega, 4(8):13189–13199, 08 2019. 2LMM copolymer mv v5: May 29, 2026 27

    2019

  47. [47]

    M. I. Skvortsova, I. I. Baskin, O. L. Slovokhotova, V. A. Paly ulin, and N. S. Zefirov. Inverse problem in QSAR/QSPR studies for the case of topological indexes ch aracterizing molecular shape (Kier indices). Journal of Chemical Information and Computer Sciences , 33(4):630–634, 1993

    1993

  48. [48]

    M. P. Stoykovich, H. Kang, K. C. Daoulas, G. Liu, C.-C. Liu, J. J. de Pablo, M. M¨ uller, and P. F. Nealey. Directed self-assembly of block copolymers for nanolit hography: Fabrication of isolated features and essential integrated circuit geometries. ACS Nano , 1(3):168–175, 10 2007

    2007

  49. [49]

    N. Q. Su and X. Xu. Insights into direct methods for predictions of ionization potential and electron affinity in density functional theory. The Journal of Physical Chemistry Letters , 10(11):2692–2699, 06 2019

    2019

  50. [50]

    Tanaka, J

    K. Tanaka, J. Zhu, N. A. Azam, K. Haraguchi, L. Zhao, H. Naga mochi, and T. Akutsu. An inverse QSAR method based on decision tree and integer program ming. In Intelligent Computing Theories and Application: 17th International Co nference, ICIC 2021, Shenzhen, China, August 12–15, 2021, Proceedings, Part II , pages 628–644. Springer, 2021

    2021

  51. [51]

    L. Tao, J. Byrnes, V. Varshney, and Y. Li. Machine learning str ategies for the structure- property relationship of copolymers. iScience, 25(7):104585, 2022

    2022

  52. [52]

    I. V. Tetko and O. Engkvist. From big data to artificial intelligenc e: chemoinformatics meets new challenges. Journal of Cheminformatics , 12:1–3, 2020

    2020

  53. [53]

    Trucillo

    P. Trucillo. Biomaterials for drug delivery and human applications. Materials, 17(2), 2024

    2024

  54. [54]

    Vogel and J

    G. Vogel and J. M. Weber. Inverse design of copolymers includin g stoichiometry and chain architecture. Chemical Science, 16(3):1161–1178, 2025

    2025

  55. [55]

    Wilbraham, R

    L. Wilbraham, R. S. Sprick, K. E. Jelfs, and M. A. Zwijnenburg. M apping binary copolymer property space with neural networks. Chemical Science, 10:4973–4984, 2019

    2019

  56. [56]

    T. Yue, L. Tao, V. Varshney, and Y. Li. Benchmarking study of deep generative models for inverse polymer design. Digital Discovery , 4:910–926, 2025

    2025

  57. [57]

    Y. Zhai, C. Li, and L. Gao. Degradable block copolymer-derived n anoporous membranes and their applications. Giant, 16:100183, 2023

    2023

  58. [58]

    Zhang, J

    F. Zhang, J. Zhu, R. Chiewvanichakorn, A. Shurbevski, H. Nag amochi, and T. Akutsu. A new approach to the design of acyclic chemical compounds using ske leton trees and integer linear programming. Applied Intelligence, 52(15):17058–17072, 2022

    2022

  59. [59]

    Zhang, Y

    S. Zhang, Y. Liu, and L. Xie. A universal framework for accura te and efficient geometric deep learning of molecular systems. Scientific Reports , 13(1):19171, 2023

    2023

  60. [60]

    Zhang, J.-C

    X. Zhang, J.-C. Daigle, and K. Zaghib. Comprehensive review of p olymer architecture for all-solid-state lithium rechargeable batteries. Materials, 13(11), 2020

    2020

  61. [61]

    Y. Zhao, R. J. Mulder, S. Houshyar, and T. C. Le. A review on th e application of molecular descriptors and machine learning in polymer design. Polymer Chemistry , 14:3325–3346, 2023. 2LMM copolymer mv v5: May 29, 2026 28

    2023

  62. [62]

    J. Zhu. Novel Methods for Chemical Compound Inference Based on Mach ine Learning and Mixed Integer Linear Programming . PhD thesis, Kyoto University, 9 2023

    2023

  63. [63]

    J. Zhu, N. A. Azam, S. Cao, R. Ido, K. Haraguchi, L. Zhao, H. N agamochi, and T. Akutsu. Quadratic descriptors and reduction methods in a two-layered mod el for compound inference. Frontiers in Genetics , 15:1483490, 2025

    2025

  64. [64]

    J. Zhu, N. A. Azam, K. Haraguchi, L. Zhao, and T. Akutsu. Com bining graph neural net- works and mixed integer linear programming for molecular inference u nder the two-layered model. In Proceedings of the 2025 9th International Conference on Com putational Biology and Bioinformatics, ICCBB ’25, pages 1–7, New York, NY, USA, 2026. Association for C omputi...

    2025

  65. [65]

    J. Zhu, N. A. Azam, K. Haraguchi, L. Zhao, H. Nagamochi, and T . Akutsu. An inverse QSAR method based on linear regression and integer programming. Frontiers in Bioscience- Landmark, 27(6):188, 2022

    2022

  66. [66]

    J. Zhu, M. Takekida, N. A. Azam, K. Haraguchi, L. Zhao, and T. Akutsu. Toward environment-sensitive molecular inference via mixed integer linear pr ogramming. ACS Omega, 10(40):46467–46481, 10 2025

    2025

  67. [67]

    J. Zhu, C. Wang, A. Shurbevski, H. Nagamochi, and T. Akutsu. A novel method for inference of chemical compounds of cycle index two with desired properties ba sed on artificial neural networks and integer programming. Algorithms, 13(5):124, 2020. 2LMM copolymer mv v5: May 29, 2026 29 Appendix A Preliminary We give some notions and terminologies on graphs in...

    2020

  68. [68]

    interior

    For any subset V ′ ⊆ V (G), the graph G − V ′ is obtained by removing all vertices in V ′ along with any edges incident to them. An edge uv incident to a leaf-vertex v is called a leaf-edge. We denote the sets of leaf-vertices and leaf-edges in G by Vleaf (G) and Eleaf (G), respectively. For a graph G (possibly rooted), a sequence of graphs Gi,i ∈ Z+ is d...

    2026

  69. [69]

    [22], we treat the two connecting-edges as a single edge e∗ 1 to simplify the representation of the polymer, as illustrated in Figure A11(b)

    Following Ido et al. [22], we treat the two connecting-edges as a single edge e∗ 1 to simplify the representation of the polymer, as illustrated in Figure A11(b). The resulting graph is called the monomer representation of the polymer, and edge e∗ 1 is also called a link-edge. In what follows, we represent polymers by their monomer representations C. The ...

    2026

  70. [70]

    dcp 1(C): the number |V (H)| − |VH| of non-hydrogen atoms in C

  71. [71]

    dcp 2(C): the number |V int(C)| of interior-vertices in C

  72. [72]

    This descriptor is only for the case of polymers

    dcp 3(C): the number |Elnk(C)| of link-edges in C. This descriptor is only for the case of polymers

  73. [73]

    dcp 4(C): the average ms(C) of mass ∗ over all atoms in C; i.e., ms(C) ≜ 1 |V (H)| ∑ v∈V (H) mass∗ (α (v))

  74. [74]

    dcp i(C), i = 4 + d,d ∈ [1, 4]: the number dg H d(C) of non-hydrogen vertices v ∈ V (H) \VH of degree deg ⟨C⟩(v) = d in the hydrogen-suppressed chemical graph ⟨C⟩

  75. [75]

    dcp i(C), i = 8 + d,d ∈ [1, 4]: the number dg int d (C) of interior-vertices of interior-degree degCint(v) = d in the interior Cint = (V int(C),E int(C)) of C

  76. [76]

    dcp i(C),i = 12 +m,m ∈ [2, 3]: the number bd int m (C) of interior-edges with bond multiplicity m in C; i.e., bd int m (C) ≜ |{e ∈ Eint(C) |β (e) = m}|

  77. [77]

    dcp i(C),i = 14 + [a]int, a ∈ Λ int(Dπ ): the frequency na int a (C) = |Va(C) ∩V int(C)|of chemical element a in the set V int(C) of interior-vertices in C

  78. [78]

    dcp i(C),i = 14 + |Λ int(Dπ )|+ [a]ex, a ∈ Λ ex(Dπ ): the frequency na ex a (C) = |Va(C) ∩ V ex(C)| of chemical element a in the set V ex(C) of exterior-vertices in C

  79. [79]

    2LMM copolymer mv v5: May 29, 2026 35

    dcp i(C), i = 14 + |Λ int(Dπ )|+ |Λ ex(Dπ )|+ [γ], γ ∈ Γ int(Dπ ): the frequency ec γ (C) of edge- configuration γ in the set Eint(C) of interior-edges in C. 2LMM copolymer mv v5: May 29, 2026 35

    2026

  80. [80]

    This descriptor is only for the case of polymers

    dcp i(C), i = 14 + |Λ int(Dπ )|+ |Λ ex(Dπ )|+ |Γ int(Dπ )|+ [γ], γ ∈ Γ lnk(Dπ ): the frequency ecγ (C) of edge-configuration γ in the set Elnk(C) of link-edges in C. This descriptor is only for the case of polymers

Showing first 80 references.