miDGD

Setting up your environment

Create a fresh conda environment for this project

conda create -n midgd python=3.10

After this, activate the environment and install the requirements:

conda activate midgd
pip install -r requirements.txt

Alternatives

Use mamba for seamless installation.

mamba create -n midgd python=3.10

Then,

mamba activate midgd

mamba install -c pytorch -c nvidia pytorch torchvision torchaudio torchmetrics pytorch-cuda=11.8 scikit-learn pandas numpy matplotlib seaborn wandb tqdm

Running example code

Code for the DGD can be found in src. The DGD base code has for now been added to the dgd folder, but will be changed and imported from the Krogh group repo in the future.

An example of how to use some code has been added in the setup_test.ipynb notebook.

Running miDGD

Code for the miDGD is stored in base and used in all python script and jupyter notebook in this repository.

The main notebook to run miDGD is the tcga_midgd.ipynb and the analyses is done in the analyses.ipynb notebook.

Reference

The miDGD model is inspired and adapted from the Deep Generative Decoder (DGD) model, specifically scDGD (https://doi.org/10.1093/bioinformatics/btad497) and multiDGD (https://doi.org/10.1101/2023.08.23.554420).

The repository of the respective model is scDGD and multiDGD. The minimal version of scDGD and multiDGD serve as the building block of the miDGD model.