/Emden

Predicting the effect of mutations on clinical drug response using graph representation and transformer encoders

Primary LanguageJupyter NotebookMIT LicenseMIT

Emden

Predicting the Effect of Mutations on Clinical Drug Response using Graph Representation and Transformer Encoders

Environment

  • Python == 3.9
  • Pytorch == 1.12
  • Scikit-learn == 1.1

Evaluate model

We provide the test dataset used in this study, you can use test_data.pt in pytorch format or testset.csv in .csv format to evaluate our method. The trained model weights can be downloaded from OneDrive url.

python evaluation.py

Prepare dataset

Requiements

  • HH-suite for generating HHblits files from protein sequences (with the file suffix of .hhm)
  • Alphafold2 for generating PDB files from protein sequences (with the file suffix of .pdb)
  • DSSP for generating DSSP files from pdb files (with the file suffix of .dssp)

Prepare data in pytorch format

cd /src
python prepareData.py

Online predict

Emden Web Server

Note: Due to the hardware limitation, the server version of Emden do not contain features generated by Alphafold2(dssp). We strongly recommend that you use this locally installed version.