Benchmarking Recent Single-Cell Synthetic Data Generation Methods on the OneK1K Dataset
Abstract
Synthetic single-cell RNA-seq data generation offers a promising route for privacy-aware biomedical data sharing, but donor-linked datasets require careful balancing of biological utility and privacy risk. In this work, we benchmark synthetic generation methods for the CAMDA 2026 ELSA Health Privacy Challenge using the donor-split OneK1K dataset, comparing scGFT, scLDM, and Cell2Sentence across statistical fidelity, downstream biological preservation, and membership inference vulnerability. Because full-transcriptome generation is computationally demanding, we focus generative modeling on highly variable genes and evaluate alternative strategies for reconstructing non-HVG expression. Our results show that non-HVG handling is a central determinant of privacy: directly copying non-HVG values leads to severe leakage, whereas Poisson-based reconstruction substantially reduces membership inference risk. We further find that scGFT and scLDM better preserve biological structure than an independent Poisson baseline, while maintaining near-random privacy profiles under appropriate configurations. These findings highlight the importance of evaluating synthetic single-cell data beyond marginal similarity alone, emphasizing robustness, interpretability, and privacy-conscious design choices throughout preprocessing, generation, and post-processing.
Related
- AVID: A 5T fMRI Dataset for Benchmarking Auditory-induced Visual Mental Imagery Decoding
- A time-resolved image-environment dataset for tatsoi growth monitoring and intelligent phytoprotection benchmarking
- OncoSpot enables benchmarking of cancer spatial transcriptomics methods across tumor evolutionary modes
- Benchmarking normalisation methods for differential binding analysis in CUT&RUN
- Evaluating in-context learning with prompting regimes for transformer-based synthetic health tabular data generation
BibTeX
@misc{giledereli2026benchmarking,
title = {{Benchmarking Recent Single-Cell Synthetic Data Generation Methods on the OneK1K Dataset}},
author = {Buse Giledereli and Basar Temiz and Amirreza Sattarzadeh Khanehbargh},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.31454},
eprint = {2610.31454},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.31454},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Benchmarking Recent Single-Cell Synthetic Data Generation Methods on the OneK1K Dataset
AU - Buse Giledereli
AU - Basar Temiz
AU - Amirreza Sattarzadeh Khanehbargh
PY - 2026
DA - 2026/10/22
PB - arXiv
DO - 10.48550/arxiv.2610.31454
UR - https://arxiv.org/abs/2610.31454
AB - Synthetic single-cell RNA-seq data generation offers a promising route for privacy-aware biomedical data sharing, but donor-linked datasets require careful balancing of biological utility and privacy risk. In this work, we benchmark synthetic generation methods for the CAMDA 2026 ELSA Health Privacy Challenge using the donor-split OneK1K dataset, comparing scGFT, scLDM, and Cell2Sentence across statistical fidelity, downstream biological preservation, and membership inference vulnerability. Because full-transcriptome generation is computationally demanding, we focus generative modeling on highly variable genes and evaluate alternative strategies for reconstructing non-HVG expression. Our results show that non-HVG handling is a central determinant of privacy: directly copying non-HVG values leads to severe leakage, whereas Poisson-based reconstruction substantially reduces membership inference risk. We further find that scGFT and scLDM better preserve biological structure than an independent Poisson baseline, while maintaining near-random privacy profiles under appropriate configurations. These findings highlight the importance of evaluating synthetic single-cell data beyond marginal similarity alone, emphasizing robustness, interpretability, and privacy-conscious design choices throughout preprocessing, generation, and post-processing.
ER -