While deep learning has revolutionized protein structure prediction, almost all experimentally characterized de novo protein designs have been generated using physically based approaches such as Rosetta. Here we describe a deep learning–based protein sequence design method, ProteinMPNN, with outstanding performance in both in silico and experimental tests. On native protein backbones, ProteinMPNN has a sequence recovery of 52.4%, compared to 32.9% for Rosetta. The amino acid sequence at different positions can be coupled between single or multiple chains, enabling application to a wide range of current protein design challenges. We demonstrate the broad utility and high accuracy of ProteinMPNN using X-ray crystallography, cryoEM and functional studies by rescuing previously failed designs, made using Rosetta or AlphaFold, of protein monomers, cyclic homo-oligomers, tetrahedral nanoparticles, and target binding proteins.

The design of proteins that bind to a specific site on the surface of a target protein using no information other than the three-dimensional structure of the target remains a challenge1–5. Here we describe a general solution to this problem that starts with a broad exploration of the vast space of possible binding modes to a selected region of a protein surface, and then intensifies the search in the vicinity of the most promising binding modes. We demonstrate the broad applicability of this approach through the de novo design of binding proteins to 12 diverse protein targets with different shapes and surface properties. Biophysical characterization shows that the binders, which are all smaller than 65 amino acids, are hyperstable and, following experimental optimization, bind their targets with nanomolar to picomolar affinities. We succeeded in solving crystal structures of five of the binder–target complexes, and all five closely match the corresponding computational design models. Experimental data on nearly half a million computational designs and hundreds of thousands of point mutants provide detailed feedback on the strengths and limitations of the method and of our current understanding of protein–protein interactions, and should guide improvements of both. Our approach enables the targeted design of binders to sites of interest on a wide variety of proteins for therapeutic and diagnostic applications.

The binding and catalytic functions of proteins are generally mediated by a small number of functional residues held in place by the overall protein structure. Here, we describe deep learning approaches for scaffolding such functional sites without needing to prespecify the fold or secondary structure of the scaffold. The first approach, “constrained hallucination,” optimizes sequences such that their predicted structures contain the desired functional site. The second approach, “inpainting,” starts from the functional site and fills in additional sequence and structure to create a viable protein scaffold in a single forward pass through a specifically trained RoseTTAFold network. We use these two methods to design candidate immunogens, receptor traps, metalloproteins, enzymes, and protein-binding proteins and validate the designs using a combination of in silico and experimental tests.

There has been considerable recent progress in designing new proteins using deep-learning methods1–9. Despite this progress, a general deep-learning framework for protein design that enables solution of a wide range of design challenges, including de novo binder design and design of higher-order symmetric architectures, has yet to be described. Diffusion models10,11 have had considerable success in image and language generative modelling but limited success when applied to protein modelling, probably due to the complexity of protein backbone geometry and sequence–structure relationships. Here we show that by fine-tuning the RoseTTAFold structure prediction network on protein structure denoising tasks, we obtain a generative model of protein backbones that achieves outstanding performance on unconditional and topology-constrained protein monomer design, protein binder design, symmetric oligomer design, enzyme active site scaffolding and symmetric motif scaffolding for therapeutic and metal-binding protein design. We demonstrate the power and generality of the method, called RoseTTAFold diffusion (RFdiffusion), by experimentally characterizing the structures and functions of hundreds of designed symmetric assemblies, metal-binding proteins and protein binders. The accuracy of RFdiffusion is confirmed by the cryogenic electron microscopy structure of a designed binder in complex with influenza haemagglutinin that is nearly identical to the design model. In a manner analogous to networks that produce images from user-specified inputs, RFdiffusion enables the design of diverse functional proteins from simple molecular specifications.

While deep learning has revolutionized protein structure prediction, almost all experimentally characterized de novo protein designs have been generated using physically based approaches such as Rosetta. Here we describe a deep learning based protein sequence design method, ProteinMPNN, with outstanding performance in both in silico and experimental tests. The amino acid sequence at different positions can be coupled between single or multiple chains, enabling application to a wide range of current protein design challenges. On native protein backbones, ProteinMPNN has a sequence recovery of 52.4%, compared to 32.9% for Rosetta. Incorporation of noise during training improves sequence recovery on protein structure models, and produces sequences which more robustly encode their structures as assessed using structure prediction algorithms. We demonstrate the broad utility and high accuracy of ProteinMPNN using X-ray crystallography, cryoEM and functional studies by rescuing previously failed designs, made using Rosetta or AlphaFold, of protein monomers, cyclic homo-oligomers, tetrahedral nanoparticles, and target binding proteins.

There has been considerable recent progress in designing new proteins using deep learning methods. Despite this progress, a general deep learning framework for protein design that enables solution of a wide range of design challenges, including de novo binder design and design of higher order symmetric architectures, has yet to be described. Diffusion models have had considerable success in image and language generative modeling but limited success when applied to protein modeling, likely due to the complexity of protein backbone geometry and sequence-structure relationships. Here we show that by fine tuning the RoseTTAFold structure prediction network on protein structure denoising tasks, we obtain a generative model of protein backbones that achieves outstanding performance on unconditional and topology-constrained protein monomer design, protein binder design, symmetric oligomer design, enzyme active site scaffolding, and symmetric motif scaffolding for therapeutic and metal-binding protein design. We demonstrate the power and generality of the method, called RoseTTAFold Diffusion (RFdiffusion), by experimentally characterizing the structures and functions of hundreds of new designs. In a manner analogous to networks which produce images from user-specified inputs, RFdiffusion enables the design of diverse, complex, functional proteins from simple molecular specifications.

Yeast surface display has proven to be an effective tool in the discovery and evolution of ligands with new or improved binding activity. Selections for binding activity are generally carried out using immobilized or fluorescently labeled soluble domains of target molecules such as recombinant ectodomain fragments. While this method typically provides ligands with high affinity and specificity for the soluble molecular target, translation to binding true membrane-bound cellular target is commonly problematic. Direct selections against mammalian cell surfaces can be carried out either exclusively or in combination with soluble target-based selections to further direct towards ligands for genuine cellular target. Using a series of fibronectin domain, affibody, and Gp2 ligands and human cell lines expressing a range of their targets, epidermal growth factor receptor and carcinoembryonic antigen, this study quantitatively identifies the elements that dictate ligand enrichment and yield. Most notably, extended flexible linkers between ligand and yeast enhances enrichment ratios from 1.4±0.8 to 62±57 for a low-affinity (>600 nM) binder on cells with high target expression and from 14±13 to 74±25 for a high-affinity binder (2 nM) on cells with medium valency. Inversion of the yeast display fusion from C-terminal display to N-terminal display still enables enrichment albeit with 40% to 97% reduced efficacy. Collectively, this study further enlightens the conditions – while highlighting new approaches – that yield successful enrichment of yeast-displayed binding ligands via panning on mammalian cells.

We explore the improvement of energy-based protein binder design using deep learning. We find that using AlphaFold2 or RoseTTAFold to assess the probability that a designed sequence adopts the designed monomer structure, and the probability that this structure binds the target as designed, increases design success rates nearly 10-fold. We find further that sequence design using ProteinMPNN rather than Rosetta considerably increases computational efficiency.