NIM for MolMIM
Type: Microservice Tags: NVIDIA, NIM, BioNeMo, MolMIM, controlled small molecule generation, drug discovery, SMILES, molecular generation Related: NVIDIA-BioNeMo, NVIDIA-NIM, NIM-for-GenMol, NIM-for-DiffDock, NIM-for-ALCHEMI-Batched-Molecular-Dynamics, NVIDIA-Clara, NVIDIA-AI-Enterprise, TensorRT, Triton-Inference-Server, NGC Sources: https://docs.nvidia.com/nim/bionemo/molmim/latest/overview.html Last Updated: 2026-04-29
Summary
NIM for MolMIM is NVIDIA’s BioNeMo NIM for controlled small molecule generation. Current NVIDIA docs describe MolMIM as a probabilistic autoencoder trained with Mutual Information Machine learning that represents variable-length SMILES strings in a fixed-length latent space and samples valid molecules from that space.
Detail
Purpose
MolMIM supports small-molecule drug discovery workflows where researchers need to embed, manipulate, decode, sample, or optimize molecules while preserving valid SMILES outputs and useful latent-space structure.
Current scope
- Learns an informative, clustered latent representation for molecules.
- Accepts SMILES-style molecular inputs and can return embeddings or hidden-state latent codes.
- Decodes hidden-state representations back into SMILES sequences.
- Samples valid molecules around a seed molecule within a scaled latent-space radius.
- Supports guided generation and molecular optimization with CMA-ES-style sampling against target properties or constraints.
- Can be chained with other BioNeMo NIMs for in silico drug discovery pipelines.
NVIDIA context
MolMIM is the controlled small-molecule generation NIM in the current NVIDIA NIM index. It sits near NIM-for-GenMol for fragment-based generation and NIM-for-DiffDock for docking/pose prediction after candidate molecules are generated.
Connections
- NVIDIA-BioNeMo - BioNeMo platform context for molecular generation and drug discovery.
- NIM-for-GenMol - fragment-based molecular generation NIM using SAFE representations.
- NIM-for-DiffDock - docking NIM for protein-ligand pose prediction after candidate generation.
- NIM-for-ALCHEMI-Batched-Molecular-Dynamics - simulation NIM for downstream molecular dynamics workflows.
- NVIDIA-Clara - broader healthcare and life-sciences umbrella.
- NVIDIA-AI-Enterprise - enterprise support and deployment context.
- TensorRT, Triton-Inference-Server, and NGC - optimization, serving, and distribution context.
Source Excerpts
- NVIDIA docs describe MolMIM as learning a clustered latent space for small molecule drug development.
- The current docs list embedding, hidden-state retrieval, decode, sample, and generate capabilities.