Novae: Spatial Transcriptomics and Protein Samples for Model Training

MICS-Lab released a full dataset for the Novae project on December 10, 2025. The collection includes spatial transcriptomics samples used to train the Novae model, protein samples referenced in the associated article, and some Visium and Visium HD samples. It also contains synthetic data samples.

Use Cases

• Training image-based spatial transcriptomics models based on the samples used for the Novae model.
• Analyzing protein expression patterns based on the protein samples mentioned in the article.
• Benchmarking spatial analysis algorithms based on the included Visium and Visium HD samples.
• Generating or validating synthetic biological data based on the synthetic data samples.

Strengths

• Includes the specific spatial transcriptomics samples used to train the published Novae model.
• Contains protein samples directly referenced in the associated research article.
• Provides a mix of real data (Visium/Visium HD) and synthetic data samples.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• The description notes the model was not trained on the Visium/Visium HD samples, indicating their role may be supplementary.

Provenance

Source

MICS-Lab via Hugging Face.

Collection Method

Likely compiled from experimental spatial transcriptomics and protein assays, plus generated synthetic data.

Time Range

null

Freshness

Last updated 2025-12-10 16:12:14; freshness should be verified.

Geography

null