Hyperion multiplex IMC analysis notebooks and example data for Bartemes et al. Increased interaction between B cells and CD3+ T cells is a hallmark of non-progression in human papilloma virus-associated oropharyngeal squamous cell carcinoma
01-xgboost_classification.ipynb
- Input: Individual ROI cell marker quantification tables with cell-type annotations (exported from QuPath) for each image.
- Performs Box Cox normalization of marker signal.
- Trains XGBoost classifier for cell type prediction.
- Outputs of this notebook must be loaded back into QuPath for nearest neighbor calculations between cell types.
02-prep_allClassData.ipynb
- Input: Individual ROI quantification files including nearest cell-type neighbors.
- Merges quantification files from all images into a single data frame for input into spatial analyses.
03-spatial_analysis.ipynb
- Input: Merged quantification table.
- Performs spatial analysis, including spatial interaction and pairwise distance associations.
04-niche_analysis.ipynb
- Input: Merged quantification table.
- Performs niche analysis, identifying nine spatially-associated cell niches via k-means clustering.
05-niche_survival.Rmd
- Performs survival analysis, testing patient outcomes with respect to niche composition of corresponding ROIs.
- Requires survival data, full dataset available upon request.
Sample data from four ROIs is provided in example_quant_HPV_OPSCC.zip. One thousand rows of marker quantification data for each ROI is provided, including cell annotations. Full Hyperion dataset is available upon request to the authors.
First author: Kathleen Bartemes, Ph.D. Corresponding author: Kathryn M. Van Abel, M.D.
Bioinformatic analyses performed by Brenna Novotny, M.S. and Raymond Moore, M.S.