Velociraptor Machine Learning Quantifies Similarity to Known Cell Types and Matches Cells Across Flow and Imaging Cytometry Platforms.

Velociraptor accurately registers and identifies cell types across diverse flow and imaging cytometry platforms with median F1-measure of 0.81 and correlation coefficient of 0.99.

• Why it matters: Integrating data from flow and imaging cytometry enhances cellular analysis by combining high-throughput profiling with spatial context, addressing a key gap in current methods.
• What they did: Velociraptor employs a graph-based Marker Enrichment Modeling to generate per-cell phenotype labels and calculates similarities to known cell types across multiple cytometry datasets, including IMC, cycIHC, CyTOF, and SFC, in several challenging comparisons.
• The result: The workflow effectively registers cells, reveals novel cell subsets such as CD15+ tumor cells, and accurately identifies macrophages, demonstrating its utility for analyzing heterogeneous samples and rare cell populations.