Single-cell parameter inference for Ca²⁺ pathway model

This repository contains code for Bayesian parameter inference for Ca²⁺ pathway model using No-U-Turn Sampler (NUTS) and post-sampling analyses.

Package requirement

The project requires Python 3.6 or newer. The following packages are also required:

• PyStan 2.19
• NumPy
• SciPy
• statsmodels
• pandas
• scikit-learn
• ArviZ
• Matplotlib 3.4 or newer
• seaborn
• Jupyter notebook

Project content

Code demos

• stan_calcium_mdoel_analysis.ipynb: post-sampling analysis for a single cell.
• cell_chain_analysis.ipynb: analysis for a cell chain.

Modules

• stan_helpers.py: classes for sampling parameters for single cells and analyzing sampled cells, as well as helper functions.
• calcium_models.py: ordinary differential equations (ODEs) of each Ca²⁺ model variant, which are used when simulating Ca²⁺ trajectories during post-sampling analysis, as well as parameter names in the models.

Executable scripts

All Python scripts for generating data and results presented in our manuscript in preparation are in the root folder.

If a script has a main() function, it is to be executed in command line. See help using python [script_name].py -h.

If a script has no main() function, it is divided into code cells by # %%. In Visual Studio Code or PyCharm (or any editor/IDE that supports this feature), each code cell can be executed separately.

Data

• stan_models: folder for Stan model specifications. Each subfolder contains a model. Some models have associated prior specifications. If a model is compiled by compile_stan_model.py, the compiled model will also be saved in the same subfolder.
• canorm_trajectories.csv: Ca²⁺ response trajectories
• vol_adjusted_genes.csv: MERFISH data (comma delimited)
• vol_adjusted_genes_transpose.txt: MERFISH data (tab delimited, transposed)
• cell_lists: folder for cell lists generated according to different criteria.
• cell_chain_example: folder used by code demos. The samples subfolder stores posterior distribution and related NUTS information of 100 cells. The cell_chain_analysis.ipynb notebook will also create a subfolder called multi-sample-analysis for its results.
• stan_run_meta.json: metadata of sampled cell chains.
• stan_run_comparison_meta.json: metadata for comparing sampled cell chains.
• soptsc.mat: saved variables from SoptSC, which includes a cell-cell similarity matrix.

Miscellaneous

• deprecated: archive folder for outdated scripts.