Research map / Probabilistic and scientific modeling / Molecules and proteins
Generative latent models for single-cell data
Models cellular perturbations and hierarchies with VAEs and concept bottlenecks.
41 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Molecules and proteins, keyword cluster “latent · space · representations”
State History Tried Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
8 papers in 2023–24 and 24 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 1.6% to 2.6%.
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
- 2016 · Dirichlet Process Mixture Model for Correcting Technical Variation in Single-Cell Gene Expression Data (ICML 2016 · 106 citations)
- 2018 · Syntax-Directed Variational Autoencoder for Structured Data (ICLR 2018 · 232 citations)
- 2020 · RNA Secondary Structure Prediction By Learning Unrolled Algorithms (ICLR 2020 · 36 citations)
- 2021 · PANTHER: Pathway Augmented Nonnegative Tensor Factorization for HighER-order Feature Learning (AAAI 2021 · 8 citations)
- 2022 · Probabilistic Transformer: Modelling Ambiguities and Distributions for RNA Folding and Molecule Design (NeurIPS 2022 · 4 citations)
- 2023 · Modelling Cellular Perturbations with the Sparse Additive Mechanism Shift Variational Autoencoder (NeurIPS 2023 · 8 citations)
- 2024 · Constraint Latent Space Matters: An Anti-anomalous Waveform Transformation Solution from Photoplethysmography to Arterial Blood Pressure (AAAI 2024 · 4 citations)
- 2025 · SUICA: Learning Super-high Dimensional Sparse Implicit Neural Representations for Spatial Transcriptomics (ICML 2025 · 1 citation)
- 2026 · What Makes a Representation Good for Single-Cell Perturbation Prediction? (ICML 2026 · 0 citations)
Earliest papers here
- Dirichlet Process Mixture Model for Correcting Technical Variation in Single-Cell Gene Expression Data (ICML 2016 · 106 citations)
- Syntax-Directed Variational Autoencoder for Structured Data (ICLR 2018 · 232 citations)
- Constrained Generation of Semantically Valid Graphs via Regularizing Variational Autoencoders (NeurIPS 2018 · 79 citations)
What has been tried
24 papers here report 227 measured changes (a component added, removed or swapped, compared with a baseline): helped 65%, no clear effect 9%, hurt 25%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add CAM to the VQ-VAE K-nearest-neighbor baseline | VitalDB, MR 10% · RMSE | 0.019 -> 0.017 | Helped | Constraint Latent Space Matters: An Anti-anomalous Waveform Transformation Solution from Photoplethysmography to Arterial Blood Pressure (AAAI 2024 · 4 citations) |
| Add KL loss (rather than removing it) | DS1 node deletion (setting1) · ELBO | -6725.49 -> -6711.47 | Helped | PhyloGen: Language Model-Enhanced Phylogenetic Inference via Graph Structure Generation (NeurIPS 2024 · 2 citations) |
| Add AE to Vanilla INR | Embryo E16.5 · MSE | 2.35 -> 1.60 | Helped | SUICA: Learning Super-high Dimensional Sparse Implicit Neural Representations for Spatial Transcriptomics (ICML 2025 · 1 citation) |
| replace the structured-space GP with a latent-space GP | discrete molecular BO benchmark · best objective score | 0.217+/-0.02 -> 0.202+/-0.00 | Hurt | Return of the Latent Space COWBOYS: Re-thinking the use of VAEs for Bayesian Optimisation of Structured Spaces (ICML 2025 · 0 citations) |
| Use fuzzy θ, Stiefel basis, and equivariance loss instead of vanilla codon embeddings | E. coli · accuracy | 0.580 -> 0.605 | Helped | Equi-mRNA: Protein Translation Equivariant Encoding for mRNA Language Models (NeurIPS 2025 · 0 citations) |
| replace DeepSDF decoder with symmetric decoder | Neural Heaviside SDF test dataset · MSEHv | 0.000277 ± 1.07e-05 -> 0.00128 ± 0.000201 | Hurt | Feature-Mapping Topology Optimization with Neural Heaviside Signed Distance Functions (ICML 2025) |
| add feature gating | C. elegans Small, supervised · Test RF | 0.286 (0.05) -> 0.226 (0.02) | Helped | Reconstructing Cell Lineage Trees from Phenotypic Features with Metric Learning (ICML 2025) |
| add latent imputation before peptide prediction | Nine-species · amino acid precision | 0.741 -> 0.797 | Helped | Latent Imputation before Prediction: A New Computational Paradigm for De Novo Peptide Sequencing (ICML 2025) |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 32 papers; from OpenAlex and the papers' affiliations).
- Skolkovo Institute of Science and Technology · 2 papers
- National University of Singapore · 2 papers
Use it
Download BibTeX of 31 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
- Neuro-symbolic probabilistic layers (120 papers)
- Sparse graphical and tensor models (97 papers)
- Probabilistic embeddings and relational inference (93 papers)
- Equivariant and disentangled latent representations (90 papers)
Elsewhere in the topic
- Generative protein structure design (142 papers)
- Prediction-powered and conformal inference (129 papers)
- Statistical analysis of ML practice (122 papers)
- Structure-based 3D molecule generation (107 papers)
- Flow matching generative modeling (97 papers)
- Machine learning for physical simulation (96 papers)
- Multimodal trajectory prediction (95 papers)
- Counterfactual and statistical explanations (81 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).