RASP-MIBI:
This repository contains the code for the article, "Multiplexed Imaging Analysis of the Tumor-Immune Microenvironment Reveals Predictors of Outcome in Triple-Negative Breast Cancer."
Published in Nature Communications Biology here.
This code was developed in the following settings.
- MacOS 10.14.6
- Intel Core i5
- python (3.7.3)
- numpy (1.16.4)
- pandas (0.24.2)
- opencv-python (3.4.2.16)
- Pillow (6.0.0)
- scikit-image (0.15.0)
- scipy (1.4.1)
- matplotlib (3.1.0)
- lifelines (0.24.0)
- seaborn (0.10.1)
- statsmodels (0.11.1)
- pysurvival (0.1.2)
- scikit-image (0.15.0)
- shap (0.37.0)
To install the required packages, you can download the required packages individually using:
pip3 install package_name
or, alternatively, use the requirements.txt file:
pip3 install -r requirements.txt
Download the repository as a whole to run the demos. Install time should be less than 10 minutes.
The computational pipeline built for this study is included as rasp_mibi_pipeline.pynb. It can be run to reproduce results or played with to see how results change. It can also be launched interactively with Colab or Binder.
Images can be downloaded from: https://www.angelolab.com/mibi-data.
Only step 2 requires these images directly--all other parts of analysis can be run without it, as intermediate data is provided.
Note: this step requires the original image dataset in order to be run. If the reviewer chooses to download the images, please edit the paths in the code.
python3 calculate_cell_prevalence.py
Purpose: calculate the proportion of cells of each cell type in each patient's image.
Output: a CSV file in the intermediate_data/ folder indicating the prevalence of each cell type in each patient's image. \
python3 calculate_protein_expression.py
Purpose: Calculate protein expression in each cell of each patient's image and assign positivity to each cell based on a threshold.
Output: CSVs of expression levels in the intermediate_data/protein_expression/ folder and positivity assignments in intermediate_data/created_protein_positivity/
python3 immune_composition.py
Purpose: determine whether immune composition is associated with recurrence or survival.
Ouput: two results CSVs in results/.
python3 protein_expression.py
Purpose: determine whether the expression of functional proteins is associated with recurrence or survival.
Output: two results CSVs in results/.
python3 calculate_coexpression.py
Purpose: Calculate instances of co-expression between proteins.
Output: intermediate_data/created_coexpression_matrices/. The reader can compare this output to coexpression_matrices/ to ensure reproducibility.
Estimated time: 10 minutes.
python3 protein_coexpression.py
Purpose: determine whether protein co-expression patterns are predictors of recurrence and survival.
Output: two KM curves with log-rank test p-value in results/.
python3 voronoi_interactions.py
Purpose: calculate cell-to-cell interactions using Voronoi diagrams.
Output: interaction matrices in intermediate_data/created_interaction_matrices. The reader can compare this output to interaction_matrices/ to ensure reproducibility.
Estimated time: 40 minutes.
python3 functional_protein_interactions.py
Purpose: determine whether interactions involving functional proteins are predictors of recurrence and survival.
Output: two KM curves with log-rank test p-value in results/.
python3 immunoregulatory_protein_interactions.py
Purpose: determine whether interactions involving immunoregulatory proteins are predictors of recurrence and survival.
Output: two KM curves with log-rank test p-value in results/.
python3 multivariate_cox_regression.py
Purpose: perform multivariate Cox regression.
Demo Note: the reader should change the type of cluster to be examined based on the options given in the comments.
Output: a model summary printed to output.
python3 random_survival_forest.py
Purpose: build a random survival forest to evaluate importance and measure model accuracy.
Output: an importance plot in results/ and a concordance index printed to output.
Estimated time: 2 minutes.
Multiplexed Imaging Analysis of the Tumor-Immune Microenvironment Reveals Predictors of Outcome in Triple-Negative Breast Cancer.
Aalok Patwa, Rikiya Yamashita, Jin Long, Michael Angelo, Leeat Keren, Daniel Rubin