Drug Response Prediction using Gene Expression and Molecular Structure

This is the comprehensive source code for my Master's Thesis on Drug Response Prediction using Gene Expression and Molecular Structure

Author:

• Sun Yih-Yun (孫懿筠)

Outline

0. data
   Including both the raw and preprocessed data along with the preprocessing process in this work
   • PRISM
   • GDSC2
1. Computational_method_comparison
   Comparing the performance of different models for the task of drug response prediction (drug-blind testing: unknown compounds and known cell lines)
   • Matrix-Factorization model (MF_model)
   • Machine Learning model (ML_model)
   • NN model with SMILESVec protein representation (SMILESVec_model)
   • First-order Weisfeiler-Lehman GNN model (WL_GNN_model)
2. Combined_model_evaluation
   Constructing a combined (2-step) model for the prediction of new drugs
   • CaDRReS_CLsim: Drug response prediction on known compounds and unknown cell lines based on the similarity of cell lines (prediction of cell-blind testing set)
   • CaDRReS_CLsim_SVM: Drug response prediction on unknown compounds and unknown cell lines based on the similarity of molecular structure (prediction of disjoint testing set)
3. Model_comparison
   Comparing the performance of the combined model with Precily on external testing data
4. User-friendly_interface
   Developing a user-friendly drug response prediction tool using a Docker image
5. Discussion
   • lineage_analysis
   • dose_range
   • parameter_tuning

Environment

• System apps
  • Python 3.10.0+
• Python packages
  • requirements.txt