Chemistry & ADMET · Tsinghua University
GROVER
A molecular graph transformer predicting ADMET endpoints from small-molecule SMILES strings.
What GROVER is
GROVER is a self-supervised molecular graph neural network developed by researchers at Tsinghua University for chemical property prediction, with a particular focus on ADMET (absorption, distribution, metabolism, excretion, and toxicity) endpoints.
Predicting pharmacological and toxicological properties early in drug discovery helps researchers screen out unsuitable drug candidates. By employing a graph representation rather than string-based molecular descriptions, the model captures molecular structural features that correlate with diverse ADMET behaviours.
How it works
GROVER is structured as a graph transformer that operates directly on molecular graph topologies. It is pre-trained on 10 million unlabelled molecules using self-supervised objectives formulated across node, edge, and whole-graph levels.
During inference, the model takes small-molecule ligand SMILES as input, converts them into molecular graphs, and outputs tabular predictions for targeted ADMET endpoints.
Key facts
- Model architecture
- Graph transformer (GNN)
- Version
- 1.0
- Domain
- Chemistry
- Supported modality
- Small molecule
- Pre-training dataset size
- 10 million unlabelled molecules
- Input format
- Ligand SMILES
- Output format
- Tabular ADMET endpoint predictions
- Required GPU memory
- 16 GB
What you can use it for
- ADMET profiling
- Generate tabular predictions for candidate ligand absorption, distribution, metabolism, excretion, and toxicity parameters.
- Molecular property prediction
- Assess chemical properties of candidate small molecules using learned graph-level representations derived from SMILES inputs.
- Early-stage drug candidate screening
- Evaluate compound libraries against multiple pharmacokinetic and safety endpoints before chemical synthesis.
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.
Not runnable yet
This model isn't available to run right now.Run this model
GROVER
Strengths and limits
Good at
- Pre-trained on a corpus of 10 million unlabelled molecules across node, edge, and graph tasks.
- Graph structure representation outperforms string representations across many ADMET benchmark tasks.
- Produces direct multi-endpoint tabular predictions from standard SMILES strings.
Watch out for
- Requires 16 GB of GPU memory for operation.
- Input is restricted to small-molecule ligand structures and does not process biological macromolecule targets.
Inputs and outputs
You provide
Ligand SMILES
Canonical SMILES for the small molecule.
You get back
ADMET endpoint predictions.
Licence
Commercial use permitted · MIT
The MIT licence allows use in commercial research and products.
Questions
Can I use GROVER commercially?
Commercial use permitted. The MIT licence allows use in commercial research and products.
How do I run GROVER 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 on 1× A100 16GB, typically in 1.2s, and a run costs about 14 credits. Visitors without an account get 1,000 credits a month.
What inputs does GROVER require?
GROVER accepts ligand structures in standard SMILES format.
What outputs does the model generate?
The model outputs a table of predicted ADMET endpoint values.
What hardware is needed to run GROVER?
The model requires a GPU with at least 16 GB of memory.
How was the model pre-trained?
GROVER was pre-trained on 10 million unlabelled chemical structures using self-supervised objectives at the node, edge, and graph levels.
Written from GROVER's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.