RFdiffusion
De novo binder design. Composite pipeline combining RFdiffusion backbones, ProteinMPNN sequences, and AF2 multimer validation. Candidates carry real ipTM, pLDDT, and i_pAE scores. Pilot runs in roughly 15 to 30 min on caller targets.
What it is for
Pick RFdiffusion when you want general de novo binder design scored by AF2 multimer (ipTM, pLDDT, i_pAE). For antibody and nanobody scaffolds use RFantibody, for AF2-IG initial-guess scoring use PXDesign, and for hallucination-driven binders without AF2 filtering use BindCraft.
RFdiffusion (Watson et al., Nature 2023). Diffusion-based backbone generator. The Ranomics composite pipeline pairs it with ProteinMPNN sequence design and AF2 multimer scoring, so every candidate carries real ipTM, pLDDT, and i_pAE statistics from the AF2 re-prediction stage.
When it fits:
- You want general de novo binder design with AF2-grounded scoring.
- Your target is a standard protein epitope (no glycans, no PTMs).
- You want flexible binder length and topology rather than an antibody scaffold.
Inputs
You will need:
- Target structure (
.pdb/.cif). - Chain ID of the target.
- At least one hotspot residue.
Each run uses a preset that sets the scale and scope:
- Your target, ~30 min start to first results
- Real RFdiffusion run against your uploaded target PDB with AF2 multimer validation. Pick 1 to 1000 candidates. Start with a small batch (4 designs, ~30 min) to confirm your target and hotspots, then scale to 100+ once the small batch looks reasonable. Results emailed when complete; A100-80GB.
Parameters you set on the form:
- Hotspot residues
- Comma-separated target-chain residues the binder should contact during diffusion.
- Binder length (min/max)
- Residue-count window. 55 to 65 is a sane default for compact PD-L1-style targets; longer binders work for larger interfaces.
- Number of designs
- How many candidates to generate. Each passes ProteinMPNN sequence design and AF2 multimer scoring.
Typical runtime:
- pilot
- 15 to 30 min
How to read the results
Ranked candidates with ipTM, pLDDT, i_pAE, and downloadable PDBs. Aim for at least 1 in 5 with ipTM ≥ 0.65 on a tractable target before committing to a full pilot.
Where a tool reports them, the scores mean:
- ipTM
- Predicted confidence in the binder to target interface. Higher is better. Aim above roughly 0.7 on a tractable target.
- pLDDT
- Per-residue confidence in the predicted fold. Higher means the model is more sure of that part of the structure.
- i_pAE and pAE
- Predicted alignment error, at the interface (i_pAE) or across the whole structure (pAE). Lower is better.
References
Watson, J. L., Juergens, D., Bennett, N. R., et al. "De novo design of protein structure and function with RFdiffusion." Nature 620, 1089 to 1100 (2023). Composite pipeline: RFdiffusion backbones, ProteinMPNN sequences, and AF2 multimer validation.