Research map / Graph learning / Graph self-supervised learning
Molecular graph pretraining
Pretrain molecular graphs with motifs and fragments for property prediction and design.
101 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in Graph self-supervised learning, keyword cluster “molecular · generation · molecules”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 2 of 2 tracked benchmarks.
25 papers in 2023–24 and 40 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 7.3% to 6.8%.
Open questions the data points to
- HIV (ROC-AUC (%)): best went from 79.5 (up to 2024) to 81.8 in 2025 — MIPT: Multilevel Informed Prompt Tuning for Robust Molecular Property Prediction (ICML 2025)
- QM9 (MAE): best went from 0.02 (up to 2024) to 0.007 in 2025 — MOL-Mamba: Enhancing Molecular Representation with Structural & Electronic Insights (AAAI 2025 · 6 citations)
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2018 · Junction Tree Variational Autoencoder for Molecular Graph Generation (ICML 2018 · 830 citations)
- 2019 · NeVAE: A Deep Generative Model for Molecular Graphs (AAAI 2019 · 143 citations)
- 2020 · Self-Supervised Graph Transformer on Large-Scale Molecular Data (NeurIPS 2020 · 418 citations)
- 2021 · MIMOSA: Multi-constraint Molecule Sampling for Molecule Optimization (AAAI 2021 · 45 citations)
- 2022 · Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2022 · 158 citations)
- 2023 · Mole-BERT: Rethinking Pre-training Graph Neural Networks for Molecules (ICLR 2023 · 135 citations)
- 2024 · Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 2024 · 7 citations)
- 2025 · gRNAde: Geometric Deep Learning for 3D RNA inverse design (ICLR 2025 · 24 citations)
- 2026 · Graph VQ-Transformer (GVT): Fast and Accurate Molecular Generation via High-Fidelity Discrete Latents (AAAI 2026 · 1 citation)
Earliest papers here
- Junction Tree Variational Autoencoder for Molecular Graph Generation (ICML 2018 · 830 citations)
- Constrained Graph Variational Autoencoders for Molecule Design (NeurIPS 2018 · 330 citations)
- Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation (NeurIPS 2018 · 243 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
HIV · ROC-AUC (%) (higher is better)
4 papers, 2021–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 79.5 | Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021 · 44 citations) |
| 2025 | 81.8 | MIPT: Multilevel Informed Prompt Tuning for Robust Molecular Property Prediction (ICML 2025) |
QM9 · MAE (lower is better)
4 papers, 2023–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 0.02 | Energy-Motivated Equivariant Pretraining for 3D Molecular Graphs (AAAI 2023 · 26 citations) |
| 2025 | 0.007 | MOL-Mamba: Enhancing Molecular Representation with Structural & Electronic Insights (AAAI 2025 · 6 citations) |
What has been tried
79 papers here report 856 measured changes (a component added, removed or swapped, compared with a baseline): helped 68%, no clear effect 6%, hurt 21%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use multi-state input with 3 states instead of single-state input | Multi-state split · native sequence recovery | 0.481 -> 0.531 | Helped | gRNAde: Geometric Deep Learning for 3D RNA inverse design (ICLR 2025 · 24 citations) |
| generate explicit hydrogens | 3D molecule generation with explicit hydrogens, pretraining · Valid | 98.58% -> 77.33% | Hurt | BindGPT: A Scalable Framework for 3D Molecular Design via Language Modeling and Reinforcement Learning (AAAI 2025 · 11 citations) |
| Add DeepSets conformer aggregation | MARCEL tasks (9 tasks, 6 base models) · test error | Helped | Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 2024 · 7 citations) | |
| Remove all self-supervised losses | BBBP · ROC-AUC (%) | 75.0 (±0.2) -> 72.5 (±0.4) | Hurt | MOL-Mamba: Enhancing Molecular Representation with Structural & Electronic Insights (AAAI 2025 · 6 citations) |
| Add cell morphology features | ChEMBL2K · AUC | 80.7 ± 0.6 -> 81.3 ± 0.6 | Helped | Learning Molecular Representation in a Cell (ICLR 2025 · 6 citations) |
| Add bond vectors by retaining the equivariant updating layer | JARVIS formation energy · Test MAE | 30.2 -> 28.4 | Helped | Complete and Efficient Graph Transformers for Crystal Material Property Prediction (ICLR 2024 · 6 citations) |
| add the proposed robust graph construction instead of multi-edge graph construction | JARVIS Bandgap (MBJ) · MAE | 0.28 -> 0.13 | Helped | Conformal Crystal Graph Transformer with Robust Encoding of Periodic Invariance (AAAI 2024 · 5 citations) |
| add M-GNN pre-training | BBBP · ROC-AUC | 92.2 -> 94.8 | Helped | Bi-level Contrastive Learning for Knowledge-Enhanced Molecule Representations (AAAI 2025 · 4 citations) |
| Add the contrastive objective to SMILES reconstruction | ChEMBL23 anchor-supermutant evaluation · Spearman's ρ | 0.514 ± 0.53 -> 0.878 ± 0.20 | Helped | SALSA: Semantically-Aware Latent Space Autoencoder (AAAI 2024 · 3 citations) |
| Remove 3rd-order interaction | PCQM4Mv2 validation-3D · Cross-entropy | 1.199 -> 1.270 | Hurt | Triplet Interaction Improves Graph Transformers: Accurate Molecular Graph Learning with Triplet Graph Transformers (ICML 2024 · 3 citations) |
| Use the proposed geometric diversity descriptor instead of SOAP | QM9 mu · MAE | 1322 -> 1092 | Helped | Empowering Active Learning for 3D Molecular Graphs with Geometric Graph Isomorphism (NeurIPS 2024 · 3 citations) |
| Replace the original graph with a fixed Cayley graph | molhiv, molbbbp, and molbace · ROC-AUC | GraphConv: molhiv 68.24 ± 1.77 -> 67.91 ± 0.75; molbbbp 64.11 ± 4.50 -> 61.60 ± 4.48; molbace 63.18 ± 4.56 -> 56.94 ± 7. | Mixed | Position: Graph Learning Will Lose Relevance Due To Poor Benchmarks (ICML 2025 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 30 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- QM9 · 8 of 30 papers
- BBBP · 4 of 30 papers
- BACE · 4 of 30 papers
- PCQM4Mv2 · 4 of 30 papers
- SIDER · 3 of 30 papers
- ClinTox · 3 of 30 papers
- Tox21 · 3 of 30 papers
- ToxCast · 3 of 30 papers
Baselines
- MolCLR · 5 of 30 papers
- Uni-Mol · 5 of 30 papers
- SchNet · 4 of 30 papers
- DMPNN · 4 of 30 papers
- GROVER · 4 of 30 papers
- GCN · 3 of 30 papers
- GIN · 3 of 30 papers
- AttentiveFP · 3 of 30 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 65 papers; from OpenAlex and the papers' affiliations).
- University of Science and Technology of China · 4 papers
- Massachusetts Institute of Technology · 4 papers
- Tsinghua University · 3 papers
- Korea Advanced Institute of Science and Technology · 3 papers
- Tencent · 2 papers
- Pohang University of Science and Technology · 2 papers
- Westlake University · 2 papers
- Cornell University · 2 papers
- University of Illinois, Urbana Champaign · 2 papers
- Rensselaer Polytechnic Institute · 2 papers
Use it
Download BibTeX of 40 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Diffusion-based graph generation (138 papers)
Elsewhere in the topic
- Semi-supervised node representation learning (214 papers)
- Heterophily and imbalance on graphs (178 papers)
- LLMs and graph foundation models (117 papers)
- Graph anomaly detection (101 papers)
- Graph domain adaptation (82 papers)
- Brain and biological graph learning (78 papers)
- Multi-view graph clustering (70 papers)
- Graph contrastive learning analysis (61 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).