Supporting data and code for the paper: "Low-frequency ERK and Akt activity dynamics are predictive of stochastic cell division events"
![](https://github.com/03bennej/predicting-cell-division/raw/main/figures/workflow/workflow.png)
The following instructions can be run on a *nix machine to reproduce our work:
-
Download and install Python. Our results were produced using Python 3.11.5, and the specific versions of packages in
requirements.txt
may require it, but you may try with another Python version... -
Fork the project repo and navigate to it on your local machine. Typing
make help
gives a complete list of make commands to be run consecutively. -
Type
make all
to run all make commands consecutively, though this will take some time. Instead you can run step-by-step:- Type
make venv
to create a virtual environment and download all required python packages. - Type
make processed
to process the raw data for classification. InMakefile
, editsplit_seed
for a different train/test split, ortruncate_seed
for a different sampling of truncated time points. - Cross validated performance analysis can be run for a number of individual methods, e.g.
make lstm
analyses the LSTM method. Inspect the Makefile, or typemake help
for more commands. - Type
make models
to train final ensemble models. - Type
make interpretation
to run the interpretation algorithm. - Type
make test
to test models on test sets.
- Type
-
Results are saved in the
results/
folder. We provide notebooks that provide visualizations and further analysis innotebooks/
.