GDEC

Transfer learning for clustering single-cell RNA-seq data crossing-species and batch, case on Uterine fibroids

Code description

The files starting with util is the processing code for each dataset.
The files starting with DECtransfer are the training codes for each dataset.
dglNet.py is the util for construct the GCN.
dglNet_batch_gene.py is used for constructing embedding features of genes.
DECpredictValid.py is used for predict the result.

Environment

For details environment.yml

Data

The data used in paper can be obtained on https://drive.google.com/file/d/1o0JOBVzTp7GdgbCi3YCKU4mWrh8fDiPg/view?usp=drive_link

Pipline

1. Constructing GCN Network Nodes and Relationships Using dglNet.py .
2. Constructing GCN embedding features with dglNet_batch_gene.py .
3. Create a folder corresponding to the dataset. For example, we should create the file named "mouse"and "mouse_gcn" folder to save the data for model training on the brain dataset.
4. Constructing features for train using the file starting with util.
5. Training the model using the file starting with DECtransfer.