Python packages:
h5py >= 2.7.0
numpy >= 1.14.2
pandas == 0.22.0
scipy >= 0.19.1
pyfasta >= 0.5.2
torch >= 1.0.0
0_preproc_dataset
for process dataset.
1_train
for init, train and test models.
1_train/utils.py
contains model architecture.
2_explain
for explain models.
2_explain/explainer.py
contains model explainer.
3_application
for predicting genomic tracks.
main
examples for run model in each species.
Analysis_plotting
analysis and plotting function.
We provided single cell labels for eight species in Dataset.
Examples:
python 1_train/1_hyperopt_BCE_best.py ./Dataset.Dmel_train_test.h5
python 1_train/1_hyperopt_BCE_best.py ./Dataset.Dmel_train_test.h5 --mode test
python 2_explain/1_run_explain.py ./Dataset.Dmel_train_test.h5
Details:
./Dataset.Dmel_train_test.h5
: example of Dataset.h5 file.
./1_train/1_hyperopt_BCE_best.py
: for init, train and test models.
--mode
: mode choice for train, test, test_all_gene.
2_explain/1_run_explain.py
: for explain models.
--help
: print help info.
Nvwa is now more like in-house scripts for reproducing our work in Deep learning of cross-species single-cell landscapes identifies conserved regulatory programs underlying cell types, if you find any problem running Nvwa code, please contant us https://github.com/JiaqiLiZju/Nvwa.