Overview
ADMET-AI is an ADMET prediction platform whose repository states that it uses Chemprop models trained on ADMET datasets from Therapeutics Data Commons (TDC). It can make predictions on new molecules via the command line, a Python API, or a web server; a live server is hosted at admet.ai.greenstonebio.com, operated separately from the code repository.
The project was published with the paper 'ADMET-AI: A machine learning ADMET platform for evaluation of large-scale chemical libraries' (Bioinformatics, 2024, doi:10.1093/bioinformatics/btae416), and the package is distributed on PyPI as admet_ai. The repository LICENSE.txt is MIT (Copyright (c) 2026 Kyle Swanson).
The platform builds on Chemprop, which is cataloged separately; ADMET-AI is the named model set, code and service around that method. The hosted site has its own terms of service, and uploaded molecules are subject to those terms. Predicted ADMET values are screening signals whose applicability depends on the training domain; they do not replace experimental ADME or toxicity measurements.
Key Features
- ADMET predictions from Chemprop models trained on ADMET datasets from Therapeutics Data Commons.
- Command-line, Python API and web-server usage modes for single molecules and large libraries.
- Reference implementation published with a Bioinformatics paper (2024) and packaged on PyPI as admet_ai.
Use Cases
- Prioritize large chemical libraries by predicted ADMET properties before synthesis or testing.
- Run local ADMET screening on proprietary molecules without uploading them to a hosted service.
How to Use
Install from PyPI (admet_ai) or from the repository, then run predictions via the command line or Python API; a hosted web server is available at admet.ai.greenstonebio.com under that site's own terms. Predictions are computational estimates and do not replace experimental measurements. See: https://github.com/swansonk14/admet_ai
Sources and verification
This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.
Repository source reviewed for this catalog version:
Read the editorial and verification policy