Large-scale Quantitative Assessment of Tissue Preparation and Staining Conditions for Robust Multiplexed Imaging.

• Project Overview
• Data & Resources
• Workflow Overview
• Quick Start
• Directory Structure
• Contributors

This repository provides analysis workflows to benchmark tissue preparation and staining conditions across multiple multiplexed imaging platforms. It generates publication-ready figures, heatmaps, and statistics comparing different antigen retrieval conditions.

Key features:

• Compare marker signal intensities across conditions (Mesmer/CellXpress workflows)
• Calculate signal intensity ratios inside vs outside cell masks
• Perform manual cell type annotation (Python pipeline)
• Quantify spatial heterogeneity (Balagan analysis)

Master_metadata_Mesmer.csv is the central reference for Mesmer workflows, linking all data files across repositories. Each row represents one slide with:

• Slide_Key: Matches naming in preprocessing scripts and CSV filenames (e.g., slide1, slide2)
• FOV coordinates: Used by Stage 1 preprocessing; documented here for reference
• File paths: Relative paths to BioImage Archive assets (masks, OME-TIFFs, GeoJSONs)

Master_metadata_cellXpress.csv contains region-level metadata for CellXpress workflows, with one row per tile/region including coordinates and dimensions.

Large data files are hosted externally due to size. Download and place in the appropriate folders before running workflows.

Stage 1: Preprocessing

Stage 2: Analysis

Mesmer and CellXpress are independent segmentation platforms—choose based on your data source. Signal intensity ratios use Mesmer segmentation masks only.

Manual annotation requires Stage 1 outputs (OME-TIFFs + segmentation masks from BioImage Archive) and is used for cell phenotyping on BIDMC slides only.

flowchart TD
    subgraph stage1["Stage 1: Python Preprocessing"]
        A[("Raw QPTIFF Images<br/>(BioImage Archive)")] --> B["preprocessing/"]
        B --> C["Crop FOVs<br/>(coords from Master_metadata_Mesmer.csv)"]
        C --> D["Mesmer Segmentation"]
        D --> E["Extract Features & Intensity Ratios"]
        E --> F[("CSV Data Files<br/>(Zenodo)")]
    end

    subgraph stage2["Stage 2: R Analysis"]
        F --> G["Mesmer_dataSlide_workflow.R"]
        F --> H["cellXpress_dataSlide_workflow.R"]
        F --> I["Mesmer_SignalNoise_workflow.R"]
        G --> J[("Heatmaps & Statistics")]
        H --> J
        I --> J
    end

    subgraph stage2b["Stage 2: Python Annotation"]
        F --> K["manual_annotation/ pipeline"]
        K --> L[("Cell Type Maps & Enrichment")]
    end

Note: Most users can skip Stage 1 by downloading the pre-generated CSVs from Zenodo. Stage 1 is only needed if you want to process raw images from BioImage Archive.

Typical entry point: Download CSVs from Zenodo → place in data_mesmer/ → run Mesmer_dataSlide_workflow.R (see data_mesmer/README.md).

# Python (for preprocessing and manual annotation)
pip install -r requirements.txt
pip install deepcell  # For Mesmer segmentation

# R packages
Rscript -e 'install.packages(c("dplyr", "tidyverse", "matrixStats", "ggcorrplot", "ggpubr", "tidyr", "rstatix", "readr", "svglite", "devtools", "qs"))'
Rscript -e 'devtools::install_github("immunogenomics/presto")'

Then follow the Workflow Documentation for your analysis of interest.

.
├── preprocessing/                  # Image preprocessing
├── data_mesmer/                    # Mesmer data
├── data_cellXpress/                # CellXpress data
├── balagan_analysis/               # Spatial analysis
├── manual_annotation/              # Cell type annotation
├── Master_metadata_Mesmer.csv      # Mesmer workflow metadata
├── Master_metadata_cellXpress.csv  # CellXpress workflow metadata
├── Mesmer_dataSlide_workflow.R     # Main Mesmer workflow
├── Mesmer_SignalNoise_workflow.R   # Signal intensity ratio analysis
├── cellXpress_dataSlide_workflow.R # Main CellXpress workflow
├── helper.R                        # Shared R functions
└── requirements.txt                # Python dependencies

• Johanna Schaffenrath
• Cankun Wang
• Shaohong Feng
• Lollija Gladiseva

For questions or feedback, contact Sizun Jiang: [email protected]