scDiagnostics Manuscript

Welcome

This website accompanies the manuscript “scDiagnostics: systematic assessment of cell type annotation in single-cell transcriptomics data” (Christidis et al., bioRxiv preprint).

It provides comprehensive tutorials, analysis code, and reproducible workflows demonstrating how scDiagnostics can be used to assess automated cell type annotations across different single-cell technologies and experimental conditions.

Overview

We demonstrate the utility of scDiagnostics using a simulated and three different real-world single-cell datasets:

1. Simulated single-cell data with splatter

• Synthetic data with known cell types and ground truth composition
• Illustration of common challenges of reference-based annotation transfer
• Demonstration of core diagnostic functionality in a controlled setting

2. Zeisel Mouse Brain Benchmarking

• Well-characterized mouse cortex and hippocampus scRNA-seq dataset from Zeisel et al. (2015)
• Systematic benchmarking of diagnostic sensitivity and specificity
• Stress-testing diagnostic performance against controlled gradients of label noise, class imbalance, and batch effects

3. COVID-19 PBMC scRNA-seq

• COVID-19 PBMC scRNA-seq atlas from Stephenson et al. (2021)
• Application to multi-sample multi-condition single-cell RNA sequencing dataset
• Demonstration of how the package facilitates the discovery and characterization of a disease-associated cell state in COVID-19

4. MERFISH Mouse Colitis

• Imaging-based spatially-resolved single-cell dataset of a DSS-induced mouse model of colitis from Cadinu et al. (2024)
• Demonstration of straightforward application of the diagnostic functionality to spatial transcriptomics data
• Showcase of how scDiagnostics enables the discovery and characterization of a disease-associated cell state in a mouse model of colitis

For each dataset, we predict cell type labels using four popular annotation tools:

• Azimuth — Weighted k-NN mapping
• SingleR — Correlation-based assignment
• CellTypist — Machine learning classifier
• scVI/scArches — Deep learning with VAE

Navigation

Data & Methods

Analysis environment setup, data retrival, and reproducible analysis reports with literate programming outputs:

• Setup & Installation — Install R and Python dependencies and set up the analysis environments
• Data retrieval — Download pre-processed datasets from Zenodo
• COVID-19 Preprocessing — QC, and preprocessing workflow for COVID-19 scRNA-seq data
• MERFISH Preprocessing — QC, and preprocessing workflow for MERFISH colitis spatial transcriptomics data
• Cell type annotation — Apply all four annotation methods to query data

Tutorials & Workflows

Quick start, core functionality, and common analysis workflows:

• scDiagnostics Overview — Introduction to diagnostic framework and key concepts
• Simulation Analysis — Demonstration of common challenges in reference-based annotation transfer using synthetic data and controlled perturbations
• Zeisel Brain Benchmarking — Quantitative benchmarking of scDiagnostics under stress-tested single-cell scenarios
• COVID-19 Analysis — Annotation assessment in COVID-19 scRNA-seq data
• MERFISH Analysis — Annotation assessment in MERFISH colitis spatial transcriptomics data
• Exploring Annotation Tool Diagnostics — Demonstrate complementary aspects of scDiagnostics and built-in quality metrics from major annotation tools

Citation

If you use this code or data in published research, please cite:

Christidis, A., Ghazi, A., Chawla, S., Turaga, N., Gentleman, R., & Geistlinger, L. (2026). scDiagnostics: systematic assessment of cell type annotation in single-cell transcriptomics data. Submitted.

Preprint: Available on bioRxiv

References

Datasets:

• Zeisel, A., et al. (2015). Brain structure. Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq. Science, 347(6226), 1138-1142.
• Stephenson, E., et al. (2021). Single-cell multi-omics analysis of the immune response in COVID-19. Nature Medicine, 27, 904–16.
• Cadinu, P., et al. (2024). Charting the cellular biogeography in colitis reveals fibroblast trajectories and coordinated spatial remodeling. Cell, 187(8), 2010-28.

Annotation Tools:

• Butler, A. et al. (2023). Azimuth: a Shiny app demonstrating a query-reference mapping algorithm for single-cell data. URL https://github.com/satijalab/azimuth. R package version 0.5.0.
• Aran, D., et al. (2019). SingleR: Rapid immunoglobulin and T-cell receptor annotation from single cells. Nat Immunol, 20(2), 163–72.
• Domínguez Conde, C., et al. (2022). Cross-tissue immune cell analysis reveals tissue-specific features in humans. Science, 376(6594), eabl5197.
• Lotfollahi, M., et al. (2022). Mapping single-cell data to reference atlases by transfer learning. Nat Biotechnol, 40(1), 121–30.

Repository

Code and Scripts (GitHub): github.com/ccb-hms/scDiagnosticsManuscript

Data (Zenodo):

Contact

For questions or feedback, please open an issue on GitHub.