interpretableAI_DRP

Overview

The code for running each model is divided into individual sub-folders. Two types of model execution can be done: 1) run a pretrained model with specified hyperparameters; 2) run a model from scratch with specified hyperparameters. The former execution can be done by running the run_pretrained.sh script and the latter can be done by running run_model_with_hyp.sh script.

Hyperparameter tuning has been performed on the validation set and the best set of hyperparameters for each validation strategy (leave-ccls-out/LCO, leave-drugs-out/LDO, leave-pairs-out /LPO) and each pathway collection (KEGG, PID, Reactome) are provided in sub-folders named best_hyp.

All pathway-based models (PathDNN, ConsDeepSignaling, HiDRA, PathDSP) are re-implementations of the original models, with a very small component of code being adaptations (direct usage) of the original code provided by the authors of these pathway-based models. References for such adaptations are included in the comments of the code.

References

• PathDNN Link to PathDNN paper Deng, L. et al. Pathway-guided deep neural network toward interpretable and predictive modeling of drug sensitivity. J. Chem. Inf. Model. 60, 4497–4505 (2020)

• ConsDeepSignaling (CDS) Link to ConsDeepSignaling paper Zhang, H., Chen, Y. & Li, F. Predicting Anticancer Drug Response With Deep Learning Constrained by Signaling Pathways. Front. Bioinforma. 1, (2021)

• HiDRA Link to HiDRA paper Jin, I. & Nam, H. HiDRA: Hierarchical Network for Drug Response Prediction with Attention. Journal of Chemical Information and Modeling vol. 61 3858–3867 (2021)

• PathDSP Link to PathDSP paper Tang, Y. C. & Gottlieb, A. Explainable drug sensitivity prediction through cancer pathway enrichment. Sci. Rep. 11, (2021)

Input Data

The input data for the models can be found at Zenodo.

Environment Requirement

• python -3.9.7
• pytorch -1.11.0
• pandas -1.3.4
• numpy -1.20.3
• scipy -1.7.1
• scikit-learn -0.24.2

Usage

• Download the input data folder from Zenodo and place the input data folder under the same directory as the models folder (found in this GitHub repository). You could place the input data folder else where, but make sure the dataroot argument in the bash scripts is modified accordingly. DO NOT MODIFY the input data folder.
• To run any desired model, open the terminal and navigate into the corresponding model folder (e.g. cd models/PathDNN/)
• Depending on what task you wish to run, modify the arguments (dataroot, outroot, etc.) in run_pretrained.sh or run_model_with_hyp.sh accordingly.
• Activate the Python environment you intend to use in the terminal.
• If you wish to run a pretrained model with specified hyperparameters, run bash run_pretrained.sh in the terminal; if you wish to run a model from scratch using specified hyperparameters, run bash run_model_with_hyp.sh