Skip to content

Latest commit

 

History

History
46 lines (33 loc) · 3.28 KB

Manual.md

File metadata and controls

46 lines (33 loc) · 3.28 KB

Directory Structure

  • contactmaps/ contains the nanoHUB tool package that helped us calculate the protein contact maps
  • Dataset_Files/ contains various CSV files that were used to construct the dataset, these can be largely ignored as a pickled file of the dataset is provided
  • Dataset_Files/AlphaFold_Proteins contains all the pdb files downloaded from AlphaFold
  • Dataset_Files/Baseline_Models & Dataset_Files/Enhanced_Models contain all our trained classification and regression models in the form of .joblib files
  • Dataset_Files/Feature_Selection contains all the numpy files regarding the feature selection process
  • Dataset_Files/Neural_Networks contains the model checkpoint and its average train and validation losses
  • Dataset_Files/Protein_Graph_Data cotains all the files needed to construct the protein graphs and the protein graphs themselves
  • Dataset_Files/Training_Test_Sets contain the sets used in the training and testing phases for each model
  • Molecular_Functions_Embedding_Model_&_Files/ contains two notebooks. one that was used to create the dataset used by the embedding model, and the embedding model itself, as well as the protein embeddings calculcated, protein_embeddings.pkl, in the form of a pickled python dictionary
  • Molecular_Functions_Embedding_Model_&_Files/Dataset_Files/ follows the same structure discussed
  • Dissertation contains the Overleaf project used to create our dissertation
  • Diagrams/ contains all diagrams used in our dissertation and Streamlit app in their draw.io format
  • Interim_Report contains the Overleaf project used to create our report
  • Metrics/ contains all the metrics gathered from our trained models in the form of CSV files
  • R_Scripts/ contains the scripts that were used to calculate the amino acid and protein sequence descriptors
  • Streamlit_App/ contains anything to do with our web app
  • Dataset_Creation_&_Exploration.ipynb was the Jupyter notebook used to bring together the various CSV files to create our dataset and split it into training and test sets
  • Classification_Baseline_Models.ipynb,Regression_Baseline_Models.ipynb,Classification_Enhanced_Models.ipynb & Regression_Enhanced_Models.ipynb were the Jupyter Notebooks used to train and test our various models
  • DTIs_Classification_NN and DTIs_Regression_NN were the Jupyter notebooks used to train and test the neural networks for DTI prediction
  • amino_acid_features.py, drug_features.py, protein_features.py,extract_dtis.py, models_utils.py & utils.py contain helper functions that were used to create the various CSV files, the dataset and the models

Requirements

  • Python: 3.9.16
  • PyTorch: 1.13.0
  • PyTorch Geometric: 2.1.0
  • Cuda Version: 1.17.0
  • Packages: listed in requirements.txt
  • Tested on Windows 11

Build steps

We would suggest the creation of an anaconda virtual environment and then running:

pip3 install -r requirements.txt

pip3 install torch==1.13.0 torchvision torchaudio or conda install pytorch==1.13.0 torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia

pip3 install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric==2.1.0 -f https://data.pyg.org/whl/torch-1.13.0+cu117.html