Use RFAntibody Online
Commercially Available RFAntibody No-Code Web Server
RFAntibody: De Novo Design of Antibodies and Nanobodies with a Fine-Tuned RFdiffusion
RFAntibody refers to the workflow from the preprint "Atomically accurate de novo design of antibodies with RFdiffusion", which uses a fine-tuned RFdiffusion network to design antibody variable domains (VHHs and scFvs) that bind user-specified epitopes. The work comes from a team centered at the Department of Biochemistry and Institute for Protein Design at the University of Washington, with senior author David Baker, together with collaborators at several other institutions.
Antibodies are central to therapeutics and diagnostics, yet the authors state that no existing method could design novel antibodies entirely in silico against a chosen epitope. Standard RFdiffusion could not design antibodies de novo, and AlphaFold2 routinely fails at antibody-antigen prediction, so a dedicated approach was needed.
By fine-tuning RFdiffusion mainly on antibody complex structures and screening the output experimentally, the authors produced binders to several disease-relevant targets.
Feature | Details |
|---|---|
Formats designed | VHH (single-domain) antibodies and scFvs |
Inputs | Target structure, epitope hotspot residues, antibody framework template |
Sequence design | ProteinMPNN for CDR sequences |
Filtering | Fine-tuned RoseTTAFold2 binding prediction |
Venue | bioRxiv preprint (version 2, February 2025) |
How RFAntibody Designs Antibodies Around a Target Epitope
Framework as template: The antibody framework is supplied as a template of pairwise distances and dihedral angles, so only the CDR loops and the antibody-target docking are designed. The VHH framework used was h-NbBcII10FGLA.
Epitope targeting: Hotspot residues on the target direct where the CDR loops contact the antigen.
CDR sequence design: ProteinMPNN designs the CDR sequences.
In silico filtering: A fine-tuned RoseTTAFold2 model predicts binding and is used to select designs.
Experimental screening: Designs are screened by yeast surface display, or by E. coli expression with SPR, and affinity can be improved with OrthoRep.
Experimental Validation: VHH and scFv Binders to Disease Targets
About 9,000 designs per target were screened by yeast display for RSV sites I and III, SARS-CoV-2 RBD and influenza hemagglutinin, and 95 designs per target by E. coli and SPR for TcdB, IL-7R-alpha and influenza hemagglutinin.
Target | Initial affinity or result |
|---|---|
Influenza HA | 78 nM (best) |
RSV site III | 1.4 uM |
SARS-CoV-2 RBD | 5.5 uM |
TcdB | About 262 nM; neutralized toxicity in CSPG4 knockout cells with EC50 of 460 nM |
Structural and affinity-maturation results
A cryo-EM structure of an influenza VHH at 3.0 A showed a backbone RMSD of 1.45 A and a CDR3 RMSD of 0.84 A to the design. OrthoRep affinity maturation improved affinities by about two orders of magnitude, bringing some binders to single-digit nanomolar. The authors also designed scFvs against TcdB and a Phox2b peptide-MHC complex by combining designed heavy and light chain CDRs.
What is Tamarind Bio?
Tamarind Bio is a no-code bioinformatics platform built to give life scientists and researchers access to powerful computational tools. Many cutting-edge machine learning models are hard to deploy and use. Tamarind provides an intuitive, web-based environment that removes the complexity of high-performance computing, software dependencies and command-line interfaces.
The platform is designed for biologists, chemists and other researchers who may not have a background in programming or cloud infrastructure but want to run models on their own data. Key features include:
A user-friendly graphical interface for setting up and launching experiments
A robust API for integration into existing research pipelines
An automated system for managing and scaling computational resources
Tamarind treats information and data security as a top priority, as detailed in its Trust Center and Terms of Service.
Accelerating Discovery with RFAntibody on Tamarind Bio
Antiviral binders: The authors designed VHHs against RSV, SARS-CoV-2 RBD and influenza hemagglutinin.
Toxin neutralization: A designed VHH against C. difficile toxin B (TcdB) neutralized toxicity in cell assays.
Receptor targeting: Designs were tested against IL-7R-alpha.
Peptide-MHC recognition: scFvs were designed against a Phox2b peptide-MHC complex.
How to Use RFAntibody on Tamarind Bio
Open the tool: Log in to tamarind.bio and open RFAntibody.
Provide the target: Upload the target structure the method requires (format as exposed in the interface).
Define the epitope: Specify hotspot residues for the epitope you want to bind.
Choose the framework: Select the antibody framework, such as a VHH or scFv framework, if exposed.
Run the design job: Submit the run to generate many candidate designs.
Download and filter: Download the designs and prioritize those with strong predicted binding.
Test experimentally: Screen top designs, for example by yeast display or SPR, as in the paper.
Things to Keep in Mind
Initial designs have modest affinity, often in the nanomolar to micromolar range, and may need affinity maturation.
Only one of four tested VHHs bound the fully glycosylated influenza HA trimer, and binding was partial.
The paper screened thousands of designs per target, so expect to generate and filter many candidates.
The record reviewed is a preprint and is not peer-reviewed.
Source: Bennett, Watson, Ragotte, Borst, See et al. (senior author D. Baker), "Atomically accurate de novo design of antibodies with RFdiffusion", bioRxiv preprint v2, 2025, DOI 10.1101/2024.03.14.585103, https://www.biorxiv.org/content/10.1101/2024.03.14.585103v2