spCellEval


Benchmarking Spatial Proteomics
Cell Phenotyping Methods

A comprehensive, large-scale evaluation of leading computational methods for identifying cell types in complex tissue images.

Last updated: May 2026

About the Benchmark

spCellEval is a comprehensive benchmarking platform designed to evaluate and compare computational methods for cell phenotyping in spatial proteomics and tissue imaging data.

Why Benchmark?

With the rapid growth of spatial omics, objective comparison of cell phenotyping methods is crucial for robust biological discovery.

What We Evaluate

We assess accuracy, robustness, scalability, and interpretability across diverse datasets and tissue types.

Who Is It For?

Researchers, computational biologists, and tool developers seeking unbiased insights into method performance.

Methods Evaluated

20

We evaluated a wide range of popular and novel methods for cell phenotyping.

Datasets Evaluated

13

Our benchmark includes datasets from various cancer types and healthy tissues to ensure robustness.

Metric Categories

4

Performance was measured across accuracy, robustness, scalability, and interpretability

Performance Overview

Average scores and rankings across all datasets

Full Results

Detailed metric matrices and per-class F1 heatmaps have moved to a dedicated Results page.

Insights & Recommendations

Key findings from the benchmark analysis

Optimal Method Selection

Supervised methods can recover cell types the best but require labelled data which can be challenging to obtain.

Reliable Fallback Option

Prior-knowledge based methods perform well when labeled data is scarce and can be used when quick results are needed. However, they require careful tuning and may not generalize well to all datasets.

Specialized Solution

For large datasets, clustering and/or visual gating with scimap can help identify key regions of interest before applying supervised methods.

Ready to Dive Deeper?

Access the full manuscript, browse the source code, and explore the complete datasets.

View on GitHub

Contribute Your Dataset or Method

Interested in adding your cell phenotyping method or sharing a new spatial proteomics dataset? We welcome community contributions to expand and improve the benchmark.

We review all submissions and will work with you to integrate your contribution into the platform.