The Pilot 1 Unified Drug Response Predictor benchmark, also called Uno, shows how to train and use a neural network model to predict tumor dose response across multiple data sources.
Primary: Cancer biology data modeling
Secondary: Machine Learning; Bioinformatics; Computational Biology
To use the untrained model, users must be familiar with processing and feature extraction of molecular drug data, gene expression, and training of neural networks. The input to the model is preprocessed data. Users should have extended experience with preprocessing this data.
The community can use a neural network and multiple machine learning techniques to predict drug response. The general rule is that classical methods like random forests would perform better for small datasets, while neural network approaches like Uno would perform better for relatively larger datasets. The baseline for comparison can be: mean response, linear regression, or random forest regression.
The following components are in the Model and Data Clearinghouse (MoDaC):
- The Unified Drug Response Predictor (Uno) asset contains the untrained model and trained model:
- The model topology file is uno.model.json.
- The trained model is defined by combining the untrained model (uno.model.json) and model weights (uno.model.h5).
- The trained model is used in inference.
- The Pilot 1 Cancer Drug Response Prediction Dataset asset contains the processed training and test data.
Refer to this README.