Chemistry & ADMET · NVIDIA BioNeMo
MolMIM
MolMIM generates and optimises small molecules toward target properties using a smooth latent space.
What MolMIM is
MolMIM is a chemistry model developed by NVIDIA BioNeMo for controlled molecular generation and optimisation of small molecules. It addresses challenges in molecular design by mapping chemical structures into a well-behaved latent space.
Rather than relying on brute-force sampling and subsequent filtering, the model uses a mutual-information-machine latent space that is smooth enough to allow direct navigation toward targeted chemical properties, such as drug-likeness or lipophilicity.
How it works
The model is structured around a mutual-information-machine latent space that provides smooth transitions between chemical structures. Researchers provide a starting seed molecule in SMILES format and select an objective property, enabling the model to navigate the latent representation directly toward the target profile.
Key facts
- Version
- 1.0
- Domain
- Chemistry
- Modality
- Small molecule
- Tasks
- Molecular generation, optimization
- GPU memory
- 16 GB
What you can use it for
- Property-guided molecule optimisation
- Directly navigating the latent space starting from a seed compound to reach target values for properties like QED or logP.
- Molecular analogue generation
- Generating a designated number of novel small-molecule analogues derived from an initial seed SMILES string.
- Scored candidate prioritisation
- Producing a table of generated small molecules complete with corresponding property scores for screening.
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 Managed GPU endpoint
Runs on managed compute. Availability and startup time depend on current capacity.Run this model
MolMIM
Strengths and limits
Good at
- Provides a smooth latent space that enables directed traversal toward target properties instead of random sampling and filtering.
- Outputs structured tables containing generated molecular analogues together with their computed scores.
Watch out for
- Requires 16 GB of GPU memory.
- Optimisation targets stated in the specification are limited to predefined metrics such as QED and logP.
Inputs and outputs
You provide
Seed SMILES
Canonical SMILES for the small molecule.
Moleculesoptional
Optimise foroptional
You get back
Optimised analogues with scores.
Licence
Commercial use permitted · Apache-2.0
The Apache-2.0 licence allows use in commercial research and products.
Questions
Can I use MolMIM commercially?
Commercial use permitted. The Apache-2.0 licence allows use in commercial research and products.
How do I run MolMIM 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 8.0s, and a run costs about 56 credits. Visitors without an account get 1,000 credits a month.
What inputs are required to run MolMIM?
The model takes a seed molecule as a SMILES string, the desired number of molecules to generate, and the target property selected for optimisation.
What does the model produce as output?
MolMIM outputs a table containing generated molecular analogues alongside their property scores.
Which properties can MolMIM optimise for?
MolMIM supports property-directed optimisation for molecular properties such as QED (quantitative estimate of drug-likeness) and logP.
Written from MolMIM's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.