DANNsweeps

Code for manuscript: Harris M, Mo Z, Siepel A and Garud N. The persistence and loss of hard selective sweeps amid ancient human admixture (https://www.biorxiv.org/content/10.1101/2025.10.14.682443v1).

1) Set up

To set up the Hoffman2 UCLA cluster for training the model on a A100 GPU node run steps in install_tensorflowGPU_documentation

To install all requirements activate the conda environment and install requirements found in requirements.txt

 conda activate tf_gpu_A100
 pip install -r requirements.txt

2) Data

All scripts to run simulations used for source domain and for simulation study are found in slim scripts/ directory.

3) Training

Code to train the DANN is found in the DANN/ directory.

Code to train the DANN and a CNN using simulated data only can be found in DANN/DANN_simulations/