Generative design of de novo proteins based on secondary-structure constraints using an attention-based diffusion model

We report two generative deep-learning models that predict amino acid sequences and 3D protein structures on the basis of secondary -structure design objectives via either the overall content or the per-residue structure. Both models are robust regarding imperfect inputs and offer de novo design capacity because they can discover new protein sequences not yet discovered from natural mechanisms or systems. The residue-level secondary-structure design model generally yields higher accuracy and more diverse sequences. These findings suggest unexplored opportunities for protein designs and functional outcomes within the vast amino acid sequences beyond known proteins. Our models, based on an attention-based diffusion model and trained on a dataset extracted from experimentally known 3D protein structures, offer numerous downstream applica-tions in the conditional generative design of various biological or engineering systems. Future work could include additional condi-tioning and an exploration of other functional properties of the generated proteins for various properties beyond structural objectives.

Automated summary

We present two generative deep-learning models capable of predicting amino acid sequences and 3D protein structures based on secondary structure design objectives. Both models are robust to imperfect inputs and have the capacity for de novo design, enabling the discovery of novel protein sequences not found in natural mechanisms or systems. The residue-level secondary structure design model demonstrates higher accuracy and more diverse sequences. These findings highlight untapped opportunities for protein design beyond known proteins. Our models, trained on a dataset extracted from experimentally known 3D protein structures using an attention-based diffusion model, have potential applications in the generative design of various biological or engineering systems. Further research could explore additional conditioning and other functional properties of the generated proteins beyond structural objectives.