Skip to content

haiyangLab/Subtype-Former

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Subtype-Former

Subtype-Former is a deep learning cancer subtyping method based on multi-omics input. The input of Subtype-Former includes copy number, mRNA, miRNA, and DNA methylation, and the output is the corresponding cancer subtype for each sample. Meanwhile, Subtype-Former can automatically determine the number of subtypes when the subtype number is not specified. Subtype-Former runs fast and can work for pan-cancer tasks with a large amount of data.

# We can use the following command to finish the subtyping process (We take BRCA as an example): 
python SubtypeFormer.py -m SubtypeFormer -c BRCA   
# the Clustering output file are stored in ./results/subtype_1/BRCA.SubtypeFormer  

Subtype-Former's Consensus clustering module is used as follows:

python SubtypeFormer.py -m cc -c BRCA
# record the corresponding class label for each sample and the output file is ./results/cc/BRCA.cc

Subtype-Former also supports the calculation of NMI and ARI:

python SubtypeFormer.py -m SubtypeFormer -c BRCA -d dataset_2
python SubtypeFormer.py -m nmiari -c BRCA
# record the corresponding class label for each sample and the output file is ./results/nmiari/BRCA.SubtypeFormer

Finally, we give a method to calculate all the results of Subtype-Former easily:

python run_all.py
# record the corresponding class label for each sample and the output file is ./results

Subtype-Former is based on the Python program language. The version of python is 3.8. We also use some open-source libraries: scikit-learn 0.23.2, pandas 1.1.3, numpy 1.19.2, pytorch 1.8.0. Subtype has been working on the Windows platform correctly. We used the NVIDIA GTX 2060 (8G) and Intel 10400F for the training.

About

a deep learning cancer subtyping method based on multi-omics input.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published