Auton Survival is a Python package designed for advanced survival analysis, catering to researchers and practitioners working with censored time-to-event data. It offers a comprehensive suite of tools for regression, counterfactual estimation, phenotyping, and evaluation, enabling rapid experimentation and deeper insights into time-dependent outcomes.

How It Works

The package leverages deep learning models, including Deep Survival Machines (DSM) and Deep Cox Mixtures (DCM), to handle complex survival data. It provides flexible APIs for data preprocessing, model training, and prediction, supporting various survival regression techniques and offering specialized modules for unsupervised, supervised, and counterfactual phenotyping to identify patient subgroups with distinct survival patterns or treatment responses.

Quick Start & Requirements

• Install via pip install -r requirements.txt after cloning the repository.
• Requires Python 3.5+ and PyTorch 1.1+. Scikit-survival is needed for standard metric evaluation.
• Official documentation, demo notebooks, and a white paper are available.

Highlighted Details

• Implements multiple deep learning survival models: DeepCoxPH, DeepCoxMixtures, DeepSurvivalMachines, and DeepCoxMixturesHeterogenousEffects.
• Offers diverse phenotyping methods: Intersectional, Unsupervised (PCA+GMM), Supervised, Counterfactual, and Virtual Twins.
• Provides comprehensive evaluation metrics: Brier Score, Integrated Brier Score, ROC AUC, Concordance Index, Time at Risk, RMST, and Risk at Time.
• Includes utilities for dataset loading (SUPPORT, FRAMINGHAM, PBC) and preprocessing (imputation, scaling).

Maintenance & Community

The project is hosted on GitHub and welcomes contributions, bug reports, and pull requests.

Licensing & Compatibility

MIT License. Permissive for commercial use and closed-source linking.

Limitations & Caveats

The package requires PyTorch 1.1+, which is an older version. Compatibility with newer PyTorch versions may require updates.