A Bond-Level Sequence Framework for Molecular Representation Learning with Structural Constraints.
The 3 matches · 1 of them tie a paragraph to a whole file, not to given lines: a weak match, whose lines are not tinted
- [1] § 3. Materials and Methods › 3.2. Materials › 3.2.2. Implementation Details ↔ utils/config.py, the whole file · a weak match · score 0.57 · masking ratios, hyperparameters, schedules, triplet, seeds, layer
- [2] § 3. Materials and Methods › 3.1. Methodology › 3.1.1. Preliminary ↔ datasets/dataset.py, lines 32–75 · score 0.56 · formal charges, atomic, aromaticity, conjugation, ring, vector
- [3] § 3. Materials and Methods › 3.1. Methodology › 3.1.1. Preliminary ↔ scripts/preprocess.py, lines 46–62 · score 0.56 · formal charges, atomic, aromaticity, conjugation, ring, vector
Paper
Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC
The paper is loaded when this pane is shown.
The authors' code
Python · 152 lines · 4.3 KB · no license · 1 match
config.py at commit 5b5028e, no license · at the source
Overview
- National Engineering Research Center of Industrial Crystallization Technology, School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China; (H.F.); (H.Q.); (X.H.); (D.Z.)
- Engineering Research Center of Green Refining Process, Ministry of Education, Tianjin University, Tianjin 300072, China
Abstract
Molecular property prediction is a fundamental task in drug discovery and materials design. While graph neural networks (GNNs) and SMILES-based Transformers have made significant strides, the former are often limited by local message-passing bottlenecks such as over-squashing, while the latter frequently lack explicit topological constraints and suffer from severe vocabulary imbalance. In this work, we revisit the granularity of molecular modeling and propose a representation learning framework built upon bond-level sequences. Our framework models molecules as sequences of directed bond tokens and introduces a structure-aware hybrid attention mechanism. By imposing hard topological constraints on a subset of attention heads to reinforce local connectivity while preserving global receptive fields in the remaining heads, the design is intended to separate short-range chemical bonding from long-range contextual dependencies. For pre-training, we implemented a multi-scale consistency learning paradigm, which utilizes an atom-centric group masking strategy to induce a hierarchical loss of local structural information and employs contrastive and triplet losses to ensure identity consistency across varying scales of structural degradation. Furthermore, by incorporating macro-scale physicochemical descriptors (e.g., LogP, TPSA) as global anchors, we examined how the inclusion of global attribute bias can provide weak physicochemical priors during pre-training, while its effect during downstream fine-tuning remains task-dependent. Experimental results demonstrate that our lightweight model, with approximately 3.5 million parameters, exhibits a dataset-dependent performance profile across MoleculeNet benchmarks and shows promising behavior on selected topology-sensitive tasks, particularly MUV. Ablation studies further analyze the contribution of bond-level connectivity, the stage-dependent dynamics of global attribute bias, structured masking, and pre-training configurations. Ultimately, this work provides an alternative representation design for molecular modeling, offering a parameter-efficient option for future molecular learning systems alongside traditional SMILES-based and graph-based formulations.
Reproduced under the paper's license (CC BY), from the paper cited above.
Repository
Its files are read in the Code ↔ Paper reader above, with 3 matches between paragraphs and lines of code.
Haoran-Fan/bondformer
5b5028ee4f3b0df07ee10bc5bb9cd174b59f8366, 11 March 2026Availability: 1 check, the latest on 27 September 2026: the link answers
- 27 September 2026: the link answers
21 files, not copied: shown from their source
OSCR keeps no copy of these files: this repository has no license that allows it. The reader above shows each one from its source, fetched by your browser at commit 5b5028e, when its fingerprint is the one OSCR verified. How this works.
- datasets/
build_vocab.py — Python, 111 lines, shown from its source - datasets/
dataset.py — Python, 413 lines, 1 match, shown from its source - datasets/
tokenizer.py — Python, 109 lines, shown from its source - models/
__init__.py — Python, 10 lines, shown from its source - models/
bert.py — Python, 42 lines, shown from its source - models/
embedding.py — Python, 10 lines, shown from its source - models/
heads.py — Python, 77 lines, shown from its source - models/
transformer.py — Python, 162 lines, shown from its source - scripts/
finetune_cls.py — Python, 336 lines, shown from its source - scripts/
master_preprocess.py — Python, 75 lines, shown from its source - scripts/
preprocess.py — Python, 161 lines, 1 match, shown from its source - scripts/
pretrain.py — Python, 345 lines, shown from its source - tools/
build_adj_cache.py — Python, 83 lines, shown from its source - tools/
prepare_ogb_split.py — Python, 130 lines, shown from its source - trainer/
finetune_cls_trainer.py — Python, 187 lines, shown from its source - trainer/
finetune_reg_trainer.py — Python, 184 lines, shown from its source - trainer/
losses.py — Python, 55 lines, shown from its source - trainer/
metrics.py — Python, 43 lines, shown from its source - trainer/
pretrain_trainer.py — Python, 326 lines, shown from its source - utils/
config.py — Python, 152 lines, 1 match, shown from its source - readme.md — Text, 93 lines, shown from its source
The paper's code and data availability statement is in the Data section.
Tracing map
Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.
What the map holds:
- 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
- 20 scripts, each with its path and the digest of its content;
- 3 matches between paragraphs of the paper and lines of the code (method lexical-v1);
- neither the text of the paper nor the code itself.
Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.
Data
No dataset and no data link were found in the paper.
Data Availability Statement
The source code, environment specifications, and a self-contained demonstration dataset supporting the findings of this study are publicly available in the official repository to ensure reproducibility and open access: [https://
Reproduced under the paper's license (CC BY), from the paper cited above.
Versions
The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.
Version 1, 27 September 2026: the first record
Recorded: type, language, journal, volume, issue, pages, dates, 7 authors, 5 keywords, 1 funder, 8 references.
Cite
This paper
Fan, H., Qi, H., Huang, X., Zhu, D., Wang, N., Wang, T., & Hao, H. (2026). A Bond-Level Sequence Framework for Molecular Representation Learning with Structural Constraints. Molecules (Basel, Switzerland), 31(11), 1972. https://
BibTeX
@article{fan2026bond,
author = {Fan, Haoran and Qi, Haoqiang and Huang, Xin and Zhu, Dongyang and Wang, Na and Wang, Ting and Hao, Hongxun},
title = {{A Bond-Level Sequence Framework for Molecular Representation Learning with Structural Constraints}},
journal = {Molecules (Basel, Switzerland)},
year = {2026},
month = jun,
volume = {31},
number = {11},
pages = {1972},
publisher = {Multidisciplinary Digital Publishing Institute (MDPI)},
issn = {1420-3049},
doi = {10.3390/
url = {https://
pmid = {42280272},
pmcid = {PMC13257498}
}
RIS
TY - JOUR
AU - Fan, Haoran
AU - Qi, Haoqiang
AU - Huang, Xin
AU - Zhu, Dongyang
AU - Wang, Na
AU - Wang, Ting
AU - Hao, Hongxun
TI - A Bond-Level Sequence Framework for Molecular Representation Learning with Structural Constraints
T2 - Molecules (Basel, Switzerland)
J2 - Molecules
PY - 2026
DA - 2026/
VL - 31
IS - 11
SP - 1972
SN - 1420-3049
PB - Multidisciplinary Digital Publishing Institute (MDPI)
DO - 10.3390/
UR - https://
LA - en
ER -
CSL-JSON
{
"id": "10.3390/
"type": "article-journal",
"title": "A Bond-Level Sequence Framework for Molecular Representation Learning with Structural Constraints",
"container-title": "Molecules (Basel, Switzerland)",
"author": [
{
"family": "Fan",
"given": "Haoran"
},
{
"family": "Qi",
"given": "Haoqiang"
},
{
"family": "Huang",
"given": "Xin"
},
{
"family": "Zhu",
"given": "Dongyang"
},
{
"family": "Wang",
"given": "Na"
},
{
"family": "Wang",
"given": "Ting"
},
{
"family": "Hao",
"given": "Hongxun"
}
],
"container-title-short":
"volume": "31",
"issue": "11",
"page": "1972",
"DOI": "10.3390/
"PMID": "42280272",
"PMCID": "PMC13257498",
"ISSN": "1420-3049",
"publisher": "Multidisciplinary Digital Publishing Institute (MDPI)",
"URL": "https://
"language": "en",
"issued": {
"date-parts": [
[
2026,
6,
5
]
]
}
}
The tracing map gets a citation of its own once an author has validated it and it has a DOI.
Similar papers
The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.
- [1] doi:10.34133/csbj.0036 [code]
- HYG-mol: An Interpretable Multimodal Hypergraph Framework for Molecular Property Prediction.Journal: Computational and structural biotechnology journalIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular, 2 references
- [2] doi:10.1093/bioinformatics/btag153 [code]
- MAISNet: a multi-species integrated graph neural network for acetylcholinesterase inhibitor screening.Journal: Bioinformatics (Oxford, England)In common: RDKit, PyTorch, scikit-learn, 2 other tools, 1 reference
- [3] doi:10.1021/acs.jcim.6c01299 [code]
- Targeting BCL-2 through Deep Learning-Based Drug Repurposing: A Multimodal Approach Combining Diffusion-Based Generative Modeling, Neural Relational Inference, and In Vitro Validation.Journal: Journal of chemical information and modelingIn common: RDKit, PyTorch, pandas, 1 other tool, cellular / molecular, 1 reference
- [4] doi:10.1038/s42003-026-10957-8 [code]
- Brain defence by the extracellular matrix protein Cochlin.Journal: Communications biologyIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular
- [5] doi:10.34133/csbj.0184 [code]
- Cross-Species Multitask Learning with Molecular and ADME Descriptors for Liver Microsomal Metabolic Stability.Journal: Computational and structural biotechnology journalIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular
- [6] doi:10.1002/pro.70695 [code]
- MIF-MAPMS: Enhancing identification of myelin autoantigenic peptides in multiple sclerosis through multimodal information fusion.Journal: Protein science : a publication of the Protein SocietyIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular
- [7] doi:10.3390/ijms27156614 [code]
- Candidalysin Inhibits &
lt;i& gt;Porphyromonas gingivalis& lt;/ i& gt; Lipoprotein-Induced IL-1β Production in BV-2 Microglia via Hydrophobic Microbial Interactions. Journal: International journal of molecular sciencesIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular - [8] doi:10.1038/s41586-026-10670-w [code]
- Zero-shot design of drug-binding proteins via neural iterative selection-expansion.Journal: NatureIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular
- [9] doi:10.1038/s41598-026-53415-5 [code]
- Computational design and immunoinformatics validation of a T cell multi-epitope vaccine targeting glioblastoma stem cells.Journal: Scientific reportsIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular
- [10] doi:10.1038/s41586-026-10391-0 [code]
- Cell-type-targeted mitochondrial transplantation rescues cell degeneration.Journal: NatureIn common: RDKit, PyTorch, scikit-learn, 2 other tools, cellular / molecular
Contribute
The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.
Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.
Claim this paper
Correct its record
Say what each link of this record is, remove the ones that are not the paper's, add the ones that are missing. The correction becomes a new version of the record, in its Versions section.
Validate its tracing map
You validate the map as this page shows it: 1 repository of the authors' code, each at its verified commit and with its license, 20 scripts, and 3 matches between paragraphs and code (see the Code and Map sections). It then receives a DOI on Zenodo, with you (your ORCID iD) and OSCR as its creators; the code itself is not deposited.
The map's fingerprint: sha256:0136455e1360d4b0…
Add the badge to its README
The badge links the code to this page. Copy one of these into the README of the paper's code: only you decide where it goes, and nothing is changed for you.
Markdown
[, paste the snippet at the top, then “Commit changes…” and, to review it first, “Create a new branch and start a pull request”. You open the pull request; OSCR asks for no permission.
Request its removal
To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).
Discussion, reproductions, activity
Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.
Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.
Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.