Chemistry & ADMET · NVIDIA / BioNeMo
MegaMolBART
Generates SMILES strings and predicts forward reaction products for chemical design and synthesis planning.
What MegaMolBART is
MegaMolBART is a computational chemistry model developed by NVIDIA and BioNeMo to support small-molecule research. It provides autoregressive generation of chemical structures represented as Simplified Molecular Input Line Entry System (SMILES) strings.
The model is designed to support both molecular discovery and chemical synthesis planning. By carrying out unconditional molecular generation alongside forward reaction product prediction, it assists researchers in designing novel chemical entities and anticipating reaction outcomes.
How it works
MegaMolBART processes ligand SMILES strings autoregressively to generate molecular structures and predict forward reaction products. The model was trained on small-molecule datasets sourced from ZINC and ChEMBL.
Key facts
- Version
- 1.0
- Domain
- Chemistry
- Supported tasks
- Molecular generation, reaction prediction
- Modality
- Small molecule
- Training data
- ZINC, ChEMBL
- Hardware requirement
- 16 GB GPU memory
What you can use it for
- Unconditional molecular generation
- Generating novel small-molecule candidates as SMILES strings for exploratory chemical library design.
- Forward reaction prediction
- Predicting the expected molecular products generated from specified chemical reaction inputs.
- Synthesis planning
- Evaluating potential reaction steps and outcomes to help design synthetic pathways for target compounds.
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
MegaMolBART
Strengths and limits
Good at
- Addresses both de novo molecular generation and forward reaction prediction within a single framework.
- Trained on large, standard chemical repositories including ZINC and ChEMBL.
Watch out for
- Requires a GPU with at least 16 GB of memory to run.
- Input is restricted to small-molecule representations provided in SMILES format.
Inputs and outputs
You provide
Ligand SMILESoptional
Canonical SMILES for the small molecule.
You get back
Sampled molecules.
Licence
Commercial use permitted · Apache-2.0
The Apache-2.0 licence allows use in commercial research and products.
Questions
Can I use MegaMolBART commercially?
Commercial use permitted. The Apache-2.0 licence allows use in commercial research and products.
How do I run MegaMolBART 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 5.0s, and a run costs about 42 credits. Visitors without an account get 1,000 credits a month.
What input does MegaMolBART take?
The model accepts small-molecule ligand structures represented as SMILES strings.
What output format does MegaMolBART produce?
It returns sampled molecules presented in a tabular format.
What training sets were used for MegaMolBART?
MegaMolBART was trained on compound data from the ZINC and ChEMBL databases.
How does the model assist in chemical synthesis?
It models forward reaction product prediction, allowing chemists to predict what products will form from specific reactants.
Written from MegaMolBART's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.