Diffusion outperforms hallucination when using AF3-generation protein structure prediction methods

Summary

Diffusion outperforms hallucination when repurposing AlphaFold3-generation methods for protein backbone design (1). This is in contrast with the comparison between BindCraft (2) and RFDiffusion (3), which showed more equal performance.

Figures

Ref (1)

See also

• ProteinMPNN-derived inverse folding methods underdesign aromatic residues

1.

Team P, Ren M, Sun J, Guan J, Liu C, Gong C, et al. PXDesign: Fast, Modular, and Accurate De Novo Design of Protein Binders. openRxiv; 2025. Available from: https://doi.org/10.1101/2025.08.15.670450

2.

Pacesa M, Nickel L, Schellhaas C, Schmidt J, Pyatova E, Kissling L, et al. One-shot design of functional protein binders with BindCraft. Nature. 2025;646(8084):483–92. Available from: https://doi.org/10.1038/s41586-025-09429-6

3.

Watson JL, Juergens D, Bennett NR, Trippe BL, Yim J, Eisenach HE, et al. De novo design of protein structure and function with RFdiffusion. Nature. 2023;620(7976):1089–100. Available from: https://doi.org/10.1038/s41586-023-06415-8