LUAD-Spatial

Multimodal spatial omics analysis of human LUAD by Visium and PhenoCycler

DOIT_LUAD3B.py

• A Python script integrates spatial omics data by 10x Genomics Visium and Akoya Biosciences PhenoCycler.
  • Supposes Visium (not CytAssist) and PhenoCycler-Open (not Fusion) in the example of this repository.
• Imvokes the following DOIT-xxxx scripts (1 to 8).
  • Tested with Python 3.12.
  • Requires packages such as numpy, pandas, cv2, tifffile, depending on environment.
  • Internally invokes R scripts.
    • Tested with R 4.3.2 and Seurat v4.4.0 (and will be compatible with v5).
• Results are stored in "FFPE_LUAD_3_B" directory in the current working directory.

% path_to_the_src_directory/src/DOIT_LUAD3B.py

1. DOIT_GetDATA_LUAD3B.py

• Downloads and expands the datafile (LUAD3B_dat.tgz) in "dat" directory in the current working directory.

2. DOIT_Visium_LUAD3B.py

• Generates mask images of Visium tissue region using H&E stain images in the output of SpaceRanger ("dat/Visium_FFPE_LUAD_3_B/spatial").
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-Visium-MASK/FFPE_LUAD_3_B-Visium.html".

H&E stain image	Fiducial marker removal	Binarization	Contour extraction	Mask image

3. DOIT_PhenoCycler_LUAD3B.py

• Generate mask images of PhenoCycler tissue region using fluorescence intensity in the QPTIFF data by CODEX Processor ("dat/PhenoCycler_LUAD_3_B/Experiment_LUAD_3_B.qptiff")
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-PhenoCycler-MASK/FFPE_LUAD_3_B-PhenoCycler.html".

Fluorescence image	Grayscaling	Binarization	Contour extraction	Mask image

4. DOIT_Normalize_LUAD3B.py

• Normalizes Visium expression levels by SCTransform normalization (version 1) provided by Seurat.
  • Internally invokes an R script (exec_preprocess.r).
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-Visium-NORM/FFPE_LUAD_3_B-Visium-NORM.html".

5. DOIT_Align_LUAD3B.py

• Aligns mask images of Visium (red) and PhenoCycler (blue) respectively obtained in step 2 and 3 above.
  • Mask images are scaled, translated, and rotated (affine transformations) to maximize their overlap.
  • Overlap is evaluated by their intersection over union (IoU), i.e., IoU = (Visium ∩ PhenoCycler) / (Visium ∪ PhenoCycler).
  • By exhaustive grid search in the neighborhood of the solution obtained above, IoU is optimized to generate the final result.
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-ALIGN/FFPE_LUAD_3_B-Visium-PhenoCycler-ALIGN.html".

	Scaling	Translation	Rotation	Optimization

FFPE_LUAD_3_B_transition_wh_200.mp4

6. DOIT_Eval_LUAD3B.py

• Evalutes correlation between Visium gene expression level and PhenoCycler antibody fluorescence intensity.
  • In circular regions corresponding to Visium spots average of fluorescence intensities are calculated.
  • Spatial and quantitative correlations are evaluated.
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-EVAL/FFPE_LUAD_3_B-EVAL.html".

Example of comparison

• MS4A1 gene expression by Visium and CD20 antibody fluorescence intensity by PhenoCycler.

Spatial correlation	Quantitative correlation

7. DOIT_Clustering_LUAD3B.py

• Carries out clustering of Visium spots by Seurat.
  • Internally invokes an R script (exec_clustering.r).
  • Carries out dimensionality reduction and extracts spatially variable features.
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-Visium-CLST/FFPE_LUAD_3_B-Visium-CLST.html".

8. DOIT_GetROI_LUAD3B.py

• Extracts an aligned/integrated region of interest (ROI) image based on mutual coordinate mapping.
• Results are browsable from "FFPE_LUAD_3_B/FFPE_LUAD_3_B-ROI/FFPE_LUAD_3_B-ROI.html".

Mutual coordinate mapping

PhenoCycler image coordinate (Visium on PhenoCycler)	Visium image coordinate (PhenoCycler on Visium)

Example of projection of a ROI in PhenoCycler image onto Visium image

A region in the PhenoCycler coordinate	Projection onto the Visium coordinate

Example of an integrated ROI (1)

• Cell (nuclear) positions estimated by QuPath/StarDist.
  • Colors show the estimated cell types.
  • Cell number of each type in the ROI is profiled.
  • Positions of Visium spots are indicated by circles.

• Heatmap of fluorescence intensity by PhenoCycler (superimposed).

• Heatmaps of fluorescence intensity of antibodies by PhenoCycler.

DAPI	E-cadherin	CD4	CD8	CD20	CD68

• Heatmaps of expression level of features/genes by Visium.

H&E stain	CDH1	CD4	CD8A	MS4A1	CD68

Example of an integrated ROI (2)

• Cell (nuclear) positions estimated by QuPath/StarDist.
  • Colors show the estimated cell types.
  • Cell number of each type in the ROI is profiled.
  • Positions of Visium spots are indicated by circles.

• Heatmap of fluorescence intensity by PhenoCycler (superimposed).

• Heatmaps of fluorescence intensity of antibodies by PhenoCycler.

DAPI	E-cadherin	CD4	CD8	CD20	CD68

• Heatmaps of expression level of features/genes by Visium.

H&E stain	CDH1	CD4	CD8A	MS4A1	CD68