This is the code collection for Hongzhe Li's Lab rotation project
Two versions of source code are provided: 1. Jupyter notebooks 2. Py with suffixes func and exe files. To reproduce the results, just rerun all cells in all Jupyter notebooks or execute the command in all py with suffix exe files. Py with suffix func files contain all functions used in Jupyter notebooks and py with suffix exe files contain all execution commands used in Jupyter notebooks.
There are four Jupyter notebooks: no_adver_final.ipynb, adver_final.ipynb, general_scvi.ipynb and benckmark.ipynb. no_adver_final.ipynb contains the grid search part for batch representation method without adversarial training plan and corresponding UMAPs. general_scvi.ipynb contains the grid search part for the traditional Scvi-tools and corresponding UMAPs. adver_final.ipynb contains the grid search part for batch representation method with adversarial training plan, the results for heterogeneity research and corresponding UMAPs. Benchmark.ipynb contains the overall summary and plots of the results.
csv folder contains all results from grid search and heterogeneity research.