Summary
De novo protein backbones designed by hallucination are more likely to be recovered by structure prediction neural networks than those designed by diffusion (such as RF-diffusion; (1)). This is particularly true of larger designs.
Figures
Ref (1)
1.
Frank C, Khoshouei A, Fuβ L, Schiwietz D, Putz D, Weber L, et al. Scalable protein design using optimization in a relaxed sequence space. Science. 2024;386(6720):439–45. Available from: https://doi.org/10.1126/science.adq1741