Quantify multiplexed tissue images

What this involves

Multiplexed tissue imaging records multi-channel antibody stacks and nuclear stains across single tissue sections. The workflow extracts biological structure through a strict analytical sequence: segmenting individual cell boundaries, quantifying per-cell marker expression, assigning discrete phenotypic identities, and evaluating cellular microenvironments.

Deriving reproducible conclusions depends heavily on user-guided parameters at each stage. Channel background thresholds dictate cell-type classifications, intensity floors prevent dim background cells from being arbitrarily assigned to lineages, and spatial metrics must account for image pixel dimensions. Adjacency calculations must also be matched to section scale to manage memory limits.

The platform agent runs this pipeline on its own environment using spatialproteomics, scikit-image, scipy, and AnnData. It handles data loading, background subtraction, segmentation overlays, and spatial graph construction, returning structured tables, figures, and AnnData objects directly without requiring local software or package installation.

How it goes

1. 1Image loading and channel alignmentRead multi-channel TIFF or OME-TIFF stacks using tifffile and spatialproteomics, cross-checking channel ordering against OME metadata or marker lists and logging the micrometre pixel scale.
2. 2Cell segmentationApply pre-computed facility masks or execute classical nuclear segmentation on the DAPI channel using scikit-image Otsu thresholding and watershed before expanding cell boundaries with spatialproteomics grow_cells.
3. 3Intensity quantification and background thresholdingExtract morphological measurements and mean marker intensities, apply arcsinh transformations, establish per-channel quantile background cut-offs, and filter out aberrant cell areas.
4. 4Phenotyping and subtype gatingAssign major lineages via spatialproteomics argmax classification with a minimum intensity floor to preserve unlabelled dim cells, followed by nested subtyping based on percentage-positive pixel thresholds.
5. 5Neighbourhood microenvironment clusteringCalculate local composition matrices across radius, k-nearest neighbours, or Delaunay spatial graphs in spatialproteomics, cluster microenvironments with scikit-learn KMeans, or route large fields to squidpy.
6. 6Visualisation and format exportGenerate multi-channel overlays and categorical phenotype maps, exporting the processed single-cell datasets to AnnData (.h5ad) for downstream analysis or Zarr stores for long-term archiving.

Start it here

Ask in plain language. The agent works on a Linux machine of its own — it installs and runs what the job needs, keeps every file, and reads the results back with the numbers explained. Nothing to set up, and a long run continues after you close the page.

What to watch for

• Mismatched channel order silently corrupts all downstream cell typing; always verify channel assignments against OME metadata or marker tables before processing.
• Omitting per-channel quantile background thresholds causes false positive signal to skew cell-type counts drastically.
• Running argmax cell typing without a minimum intensity threshold forces unstained and dim cells into arbitrary lineage categories rather than assigning them to unlabelled.
• Computing spatial radius neighbourhoods creates a dense cells-by-cells adjacency matrix in spatialproteomics that causes memory exhaustion on sections with more than approximately 20,000 cells.
• All spatialproteomics distance arguments use pixels rather than micrometres, requiring explicit conversion using the metadata pixel size.
• Classical watershed nuclear segmentation combined with cytoplasmic dilation frequently under-segments cells in dense lymphoid tissue.

What it runs

Questions