INFLAMMATION_ATLAS: scANVI Latent Representations for Single-Cell Data

Latent representations from the scANVI model for single-cell RNA sequencing data, derived from the INFLAMMATION_ATLAS. The dataset is a 113.2 MB ZIP file published by Benjamin Frey under a CC-BY-4.0 license on figshare in May 2026. The data originates from a Nature Medicine article published in 2025.

Use Cases

• Benchmark cell-type annotation algorithms based on the scANVI latent space.
• Train downstream classifiers for inflammatory states using the provided latent features.
• Analyze batch correction performance across integrated datasets using the external split latents.
• Develop visualization tools for high-dimensional single-cell data based on the learned representations.

Strengths

• Data is directly linked to a peer-reviewed Nature Medicine article (s41591-025-04126-3), providing a strong scientific provenance.
• The 113.2 MB file size suggests a substantial amount of latent feature data.
• It is shared under a permissive CC-BY-4.0 license, allowing for broad reuse.

Limitations

• The description is a draft and explicitly states metadata is incomplete, limiting immediate understanding of the data structure.
• Column names and row counts are unknown, which hinders assessment of data scale and feature semantics.
• The dataset is labeled as an 'external split' representation, but the specific train/test/validation partitioning is not detailed.

Provenance

Source

Derived from data in the Nature Medicine article 'https://www.nature.com/articles/s41591-025-04126-3'.

Collection Method

Generated using the scANVI (single-cell ANnotation using Variational Inference) model, likely for integration and annotation.

Freshness

Last updated 2026-05-12 15:43:41