QuantCell: machine learning based cell annotation integrating qualitative and quantitative imaging profiles

QuantCell boosts cell annotation in mouse bone marrow from 33.1% to 90.2% at 5% FDR using PhenoCycler imaging, while reaching 96.5% accuracy.

• Why it matters: Spatial omics can measure many RNA and protein markers in tissues, but assigning cell identities is difficult when tissues are complex or contain rare cell types.
• What they did: QuantCell is a machine learning framework that integrates quantitative imaging with qualitative profiles, evaluates multiple models, and applies a user-defined false discovery rate.
• The result: In PhenoCycler mouse bone marrow data, it increases annotated coverage to 90.2% at 5% FDR and attains 96.5% accuracy, supporting detection of rare populations.

AI summary of the abstract · may contain errors · not medical advice · read the paper.