PRiMeFlow VCC Datasets

Preprocessed datasets, data splits and gene features used in PRiMeFlow: Capturing Complex Expression Heterogeneity in Perturbation Response Modelling (arXiv:2604.13986). PRiMeFlow is a flow-matching model of genetic and small-molecule perturbation responses operating directly in gene expression space; the data supports its pretraining and finetuning strategy and evaluation on the H1 human embryonic stem cell data from the Arc Virtual Cell Challenge. Expression data are gzipped AnnData (.h5ad.gz), about 244 GB compressed: genome-scale Perturb-seq in primary human CD4 T cells from four donors at rest and 8 h / 48 h post-stimulation (12 files, about 198 GB; Zhu, Dann et al., 2025), VCC H1 training data, Replogle 2022 K562 and RPE1, Replogle 2020, Nadig HepG2 and Jurkat, Jiang 2024, McFaline-Figueroa GxE screens and Feng 2026 iPSC CRISPRi. Also included are ESM2 perturbation-gene embeddings (Parquet), the 18,001-gene VCC expression space and CSV split files. 363.6 GB in total. The compilation is CC BY 4.0; two bundled datasets (Feng 2026 and the CD4 T Perturb-seq) remain under their original MIT licenses.

Category
Genomics & Bioinformatics
Provider
Altos Labs
License
cc-by-4.0
Hosted on
Hugging Face, bucket altoslabs/primeflow-vcc-datasets, us
Access
Free and public. No account with the data provider is needed.
Source
https://arxiv.org/abs/2604.13986

Analyze this dataset without downloading it

On Scigantic, PRiMeFlow VCC Datasets opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

Related Genomics & Bioinformatics datasets