Description
RFD4-Proteina is a sophisticated flow-matching model designed for biomolecular design, developed in collaboration between NVIDIA and the Baker Lab at the Institute for Protein Design, University of Washington. This model excels in generating protein sequences and all-atom structures simultaneously, conditioned on specific molecular contexts such as binding targets, enzyme motifs, ligands, or selected structural features. It is particularly useful for researchers and scientists in the field of protein engineering and molecular biology.
Read More
The model operates by using a joint sequence-structure generation approach with a matched flow model and autoencoder. It accepts inputs in the form of annotated CIF structures or YAML condition specifications and outputs designed molecular structures in CIF format, complete with generated residue identities and coordinates. The model's weights are stored as EMA weights, with a total of 435,622,465 parameters across the paired bundle, available in FP32 format.
RFD4-Proteina is available for use through the Hugging Face platform, where users can access the model's code, documentation, and training data. The model is compatible with Linux x86_64 systems, requiring Python 3.12 and a CUDA 13-compatible NVIDIA GPU. It is released under the BSD 3-Clause License, with optional third-party assets having separate terms.
The model's capabilities are demonstrated through various benchmarks, showing its efficiency in designing protein binders, small-molecule binders, atomic motif enzymes, and DNA binders. These benchmarks highlight RFD4-Proteina's superior performance compared to other methods in generating structurally distinct computational passers per GPU-hour.
RFD4-Proteina is an invaluable tool for computational biomolecular design, offering researchers the ability to generate and evaluate protein designs with high precision. Its integration with Hugging Face's platform ensures ease of access and use, making it a preferred choice for scientists in the field.
- CategoryAI Models & LLMs
- Pricing ModelFree
- DeploymentCloud
- PlatformsHugging Face
RFD4-Proteina Highlights
Flow-matching model for biomolecular design
Joint sequence-structure generation
Conditioned on molecular contexts
Outputs in CIF format
435,622,465 parameters
Compatible with NVIDIA GPUs
BSD 3-Clause License
Benchmarked for efficiency
Getting Started with RFD4-Proteina
Access model: Visit Hugging Face
Authenticate: Log in to Hugging Face
Set up environment: Install dependencies
Integrate via API: Use Proteina.from_pretrained
Optimise: Adjust settings for specific tasks
RFD4-Proteina's Use Cases
- Protein Design
- Enzyme Motif Design
- Ligand Binding
- DNA Binding
- Molecular Context Preservation
