Chemistry & ADMET · Swanson et al., Stanford
ADMET-AI
Predicts 52 small-molecule ADMET endpoints simultaneously using Chemprop D-MPNN ensembles.
What ADMET-AI is
ADMET-AI is a property prediction tool developed by Swanson and colleagues at Stanford. It calculates 52 absorption, distribution, metabolism, excretion, and toxicity (ADMET) endpoints simultaneously from small-molecule chemical structures. By computing these properties together, it enables researchers to review a candidate's complete pharmacological and safety profile at an early stage.
To assist interpretation, every predicted value is displayed alongside a reference distribution showing where the candidate falls relative to approved pharmaceuticals. The tool also supports evaluating chemical series to provide ranked comparisons across multiple analogues.
How it works
Each of the 52 evaluated endpoints is modelled using an ensemble of Chemprop directed message passing neural networks (D-MPNN). The neural network architectures operate directly on molecular graphs derived from chemical SMILES inputs.
The ensembles are trained on experimental measurements curated from the Therapeutics Data Commons. During inference, predicted values are paired with an interpretation benchmarked against profiles of approved drugs.
Key facts
- Version
- 2.0.1
- Architecture
- Chemprop D-MPNN ensemble
- Training data
- Therapeutics Data Commons
- Input modality
- Small molecule (SMILES)
- Hardware requirement
- Runs on CPU
What you can use it for
- Comprehensive liability screening
- Screen novel chemical structures simultaneously across 52 ADMET assays, including hERG cardiotoxicity, clearance, solubility, and Tox21 toxicity panels.
- Benchmarking against approved drugs
- Contextualise the pharmacokinetic and safety properties of an uncharacterised lead by comparing its predicted metrics against distributions observed in approved drugs.
- Chemical series ranking
- Evaluate a collection of analogue molecules together to generate a ranked comparative profile across all ADMET endpoints.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand CPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
ADMET-AI
Strengths and limits
Good at
- Predicts 52 absorption, distribution, metabolism, excretion, and toxicity endpoints in a single run.
- Contextualises predictions by showing where candidate values sit relative to approved drugs.
- Supports ranked comparative analyses across compound series.
- Executes on standard CPU hardware without requiring GPU acceleration.
Watch out for
- Restricted to small-molecule representations that can be supplied as SMILES strings.
- Predictions are constrained to the 52 endpoints represented in the underlying Therapeutics Data Commons training sets.
Inputs and outputs
You provide
Ligand SMILES
Canonical SMILES for the small molecule.
Reportoptional
You get back
All endpoints with predicted value and interpretation.
Licence
Commercial use permitted · MIT
The MIT licence allows use in commercial research and products.
Questions
Can I use ADMET-AI commercially?
Commercial use permitted. The MIT licence allows use in commercial research and products.
How do I run ADMET-AI without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers, typically in 3.0s, and a run costs about 6 credits. Visitors without an account get 1,000 credits a month.
How are the predicted values contextualised?
Each predicted property is reported alongside an interpretation indicating where the query molecule ranks relative to values observed across approved therapeutics.
What underlying architecture is used?
Each endpoint is modelled using an ensemble of Chemprop directed message passing neural networks (D-MPNN) trained on datasets from the Therapeutics Data Commons.
Can the model compare several compounds at once?
Yes, supplying a chemical series generates a ranked comparative evaluation across the evaluated candidates.
Written from ADMET-AI's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.