6870_isPLA

Spatial Proteomics: In Situ PLA data

Data Pre-processing Overview

The input to the notebooks is the output from the pipex_bigfish pipeline (a fork of pipex). This pipeline transforms CODEX-derived images and isPLA data into AnnData files (.h5ad). Each .h5ad file contains, for each cell, the mean marker intensity as well as the number of isPLA dots.

Here is a summary of the data processing steps:

Input:

• qptiff files (8-bit compressed)

PIPEX steps:

• Cell segmentation using Stardist (default settings: nuclei diameter = 20, cytoplasm expansion = 20)
• Computation of mean intensities per cell and per marker

BigFish dot detection:

• isPLA image is denoised and spots are enhanced using a Laplacian of Gaussian (LoG) filter
• Peaks are detected in the filtered image with a local maximum detection algorithm
• An intensity threshold is applied to discriminate actual spots from noisy background
• Dense regions decomposition: detects dense and bright regions with potential clustered spots, then uses gaussian simulations to correct misdetection in these regions
• Cluster detection (cluster spots in point cloud and detect relevant aggregated structures) — not used in this study
• Stores number of isPLA spots and clusters per cell

Output:

• AnnData objects with cell information (marker intensities, isPLA spots)

H&E manual alignment:

• For each tissue sample, a consecutive slice with H&E was manually aligned to the DAPI image to assist with manual annotations

Manual annotations:

• Regions with incorrect isPLA signal (edge effects, tears, over-saturation, etc.) were manually annotated and all cells within these regions are removed from the AnnData objects for the rest of the study

Tumor regions:

• Tumor and non-tumor regions were manually annotated. All cells within tumor regions are labeled as “in_tumor” for downstream analysis.

Notebook 1: Analysis of h5ad Files

The first notebook, notebooks/1_analyse_h5ad.ipynb, processes spatial proteomics data stored in AnnData (.h5ad) files. It performs the following steps:

• Loads processed data for each sample.
• Applies geometric transformations for spatial alignment.
• Removes edge effects and adds tumor region annotations if available.
• Computes positive cells for each marker using Otsu thresholding.
• (Optional) Classifies immune cells based on marker expression.
• Saves the processed data for downstream analysis.

Notebook 2: Generation of TissUUmaps Projects

The third notebook, notebooks/2_generate_tmap.ipynb, prepares interactive TissUUmaps projects for spatial exploration of the processed data. It performs the following steps:

• Loads processed .h5ad files and associated image data for each sample.
• Converts and rescales TIFF images for visualization, including biomarker and H&E images.
• Updates TissUUmaps project state with available image layers and metadata.
• Adjusts visualization settings (e.g., layer visibility, opacity, marker display).
• Writes updated .h5ad and .tmap project files for each sample.
• Optionally extracts individual biomarker images from multiplexed TIFF files.
• Copies generated TissUUmaps project files to a separate directory for sharing or deployment.

Notebook 3: Concatenation of AnnData Objects

The fourth notebook, notebooks/3_concatenate_anndata.ipynb, combines processed AnnData (.h5ad) files from all samples into a single dataset for downstream analysis. It performs the following steps:

• Searches for all processed .h5ad files in the output directories.
• Loads each file and adds a sample identifier to the metadata.
• Concatenates all AnnData objects into one unified AnnData object.
• Ensures unique observation names across samples.
• Saves the combined dataset as all_adata.h5ad for further analysis.

Notebook 4 and 5: Figures for the paper

The fifth notebook, notebooks/4_figures.ipynb and notebooks/5_figures_supp.ipynb, create the figures used in the main paper and supplementary materials, respectively.