NIM for ProteinMPNN
Type: Microservice Tags: NVIDIA, NIM, BioNeMo, ProteinMPNN, protein sequence design, protein engineering, graph neural network, drug discovery Related: NVIDIA-BioNeMo, NVIDIA-NIM, NIM-for-RFdiffusion, NIM-for-MSA-Search, NIM-for-AlphaFold2, NIM-for-OpenFold3, NIM-for-Boltz2, NVIDIA-Clara, NVIDIA-AI-Enterprise, TensorRT, NGC Sources: https://docs.nvidia.com/nim/bionemo/proteinmpnn/latest/overview.html, https://docs.nvidia.com/nim/bionemo/proteinmpnn/latest/index.html Last Updated: 2026-04-29
Summary
NIM for ProteinMPNN is NVIDIA’s BioNeMo NIM for predicting amino-acid sequences that fit a given protein backbone. Current NVIDIA docs describe ProteinMPNN as a graph neural network that takes a protein 3D structure in PDB format and outputs amino-acid sequences in Multi-FASTA format.
Detail
Purpose
ProteinMPNN supports protein engineering and drug-discovery workflows where researchers have a desired backbone or generated structure and need plausible sequences that are likely to fold into that structure.
Current scope
- Uses evolutionary, functional, and structural information to generate candidate amino-acid sequences.
- Accepts a protein backbone structure in PDB format.
- Outputs designed amino-acid sequences in Multi-FASTA format.
- Can be chained after NIM-for-RFdiffusion, where RFdiffusion generates a 3D protein structure and ProteinMPNN designs sequences for that structure.
- Current docs expose quickstart, endpoints, benchmarking, support matrix, and advanced logging/telemetry controls.
NVIDIA context
ProteinMPNN is a protein sequence design NIM in the BioNeMo graph. It is especially important as a companion to NIM-for-RFdiffusion and to structure-prediction validation workflows using NIM-for-AlphaFold2, NIM-for-OpenFold3, or NIM-for-Boltz2.
Connections
- NIM-for-RFdiffusion - upstream generative protein structure NIM that ProteinMPNN can follow for sequence design.
- NIM-for-MSA-Search - sequence search/MSA context for protein structure workflows.
- NIM-for-AlphaFold2, NIM-for-OpenFold3, and NIM-for-Boltz2 - structure prediction/validation NIMs adjacent to sequence design.
- NVIDIA-BioNeMo and NVIDIA-Clara - life-sciences platform context.
- NVIDIA-AI-Enterprise, TensorRT, and NGC - production deployment, inference optimization, and distribution context.
Source Excerpts
- NVIDIA docs describe ProteinMPNN as predicting amino-acid sequences for given protein backbones.
- The current docs state that ProteinMPNN can be used after RFdiffusion to determine possible amino-acid sequences for generated structures.