• [06/11/2023] Check our recently released paper PyTrial: A Comprehensive Platform for Artificial Intelligence for Drug Development on arXiv!

PyTrial is designed for both ML researchers and medical practioners, which is featured for

• 🔑 Off-the-shelf pipelines for various clinical trial tasks proposed in the most recent top-venue papers.

• 🔥 A comprehensive implementation of many AI for drug development tasks:

  • Patient outcome prediction.
  • Trial site selection.
  • Trial outcome prediction.
  • Patient-trial matching.
  • Trial similarity search.
  • Trial data simulation.

• 📖 Unified APIs, detailed documentation, and interactive examples with preprocessed demo data for every implemented algorithm.

• 💡 Cutting-edge AI4Trial algorithms reproduced from the most recent top-venue papers.

• ✈️ Scalability to future research and development based on the PyTrial's architecture.

• Before install PyTrial, we need to install pytorch first. Please refer to install PyTorch for a version that suits your device.

• Then, it is easy to install PyTrial from github source:

pip install git+https://github.com/RyanWangZf/pytrial.git@main

The package is tested on python==3.7.

We DO NOT recommend downloading from PyPI temporarily because PyTrial is undergoing development swiftly.

In PyTrial, performing a task boils down to three steps: load data -> define model -> fit and predict.

To minimize the efforts learning to use PyTrial, we keep a consistent user interface for all tasks all models, i.e.,

model.fit(train_data, val_data)

model.predict(test_data)

model.save_model(save_dir)

model.load_model(load_dir)

hence all tasks are defined the input and output. All we need to do is to prepare for the input following the protocol.

We provide the following tutorials to help users get started with our PyTrial. After go through all these chapters, you will become the expert in AI for clinical trials and are ready to explore the frontier of this field.

The full documentation is at PyTrial-docs.

• Intro 1: Overview of PyTrial
• Intro 2: PyTrial API & Pipeline
• Intro 3: Basic Patient Data Class
• Intro 4: Basic Trial Data Class

• Task 1: Individual Patient Outcome Prediction
• Task 2: Clinical Trial Site Selection
• Task 3: Trial Outcome Prediction
• Task 4: Patient-Trial Matching
• Task 5: Trial Similarity Search
• Task 6: Trial Patient Records Simulation

• Misc 1: Load Preprocessed Demo Data
• Misc 2: Prepare Oncology Trial Patient Data
• Misc 3: Pretrained BERT Model
• Misc 4: ICD9 & 10 Knowledge Graph
• Misc 5: Drug Knowledge Graph

If you use PyTrial in a scientific publication, we would appreciate citations to:

@article{wang2023pytrial,
  title={PyTrial: A Comprehensive Platform for Artificial Intelligence for Drug Development},
  author={Wang, Zifeng and Theodorou, Brandon and Fu, Tianfan and Xiao, Cao and Sun, Jimeng},
  journal={arXiv preprint arXiv:2306.04018},
  year={2023},
}