Search

Allosteric modulation of protein function, wherein the binding of an effector to a protein triggers conformational changes at distant functional sites, plays a central part in the control of metabolism and cell signalling 1-3 . There has been considerable interest in designing allosteric systems, both to gain insight into the mechanisms underlying such 'action at a distance' modulation and to create synthetic proteins whose functions can be regulated by effectors 4-7 . However, emulating the subtle conformational changes distributed across many residues, characteristic of natural allosteric proteins, is a significant challenge 8,9 . Here, inspired by the classic Monod-Wyman-Changeux model of cooperativity 10 , we investigate the de novo design of allostery through rigid-body coupling of peptide-switchable hinge modules 11 to protein interfaces 12 that direct the formation of alternative oligomeric states. We find that this approach can be used to generate a wide variety of allosterically switchable systems, including cyclic rings that incorporate or eject subunits in response to peptide binding and dihedral cages that undergo effector-induced disassembly. Size-exclusion chromatography, mass photometry 13 and electron microscopy reveal that these designed allosteric protein assemblies closely resemble the design models in both the presence and absence of peptide effectors and can have ligand-binding cooperativity comparable to classic natural systems such as haemoglobin 14 . Our results indicate that allostery can arise from global coupling of the energetics of protein substructures without optimized side-chain-side-chain allosteric communication pathways and provide a roadmap for generating allosterically triggerable delivery systems, protein nanomachines and cellular feedback control circuitry., (© 2024. The Author(s).)

In natural proteins, structured loops have central roles in molecular recognition, signal transduction and enzyme catalysis. However, because of the intrinsic flexibility and irregularity of loop regions, organizing multiple structured loops at protein functional sites has been very difficult to achieve by de novo protein design. Here we describe a solution to this problem that designs tandem repeat proteins with structured loops (9-14 residues) buttressed by extensive hydrogen bonding interactions. Experimental characterization shows that the designs are monodisperse, highly soluble, folded and thermally stable. Crystal structures are in close agreement with the design models, with the loops structured and buttressed as designed. We demonstrate the functionality afforded by loop buttressing by designing and characterizing binders for extended peptides in which the loops form one side of an extended binding pocket. The ability to design multiple structured loops should contribute generally to efforts to design new protein functions., (© 2024. The Author(s).)

We describe an approach for designing high-affinity small molecule-binding proteins poised for downstream sensing. We use deep learning-generated pseudocycles with repeating structural units surrounding central binding pockets with widely varying shapes that depend on the geometry and number of the repeat units. We dock small molecules of interest into the most shape complementary of these pseudocycles, design the interaction surfaces for high binding affinity, and experimentally screen to identify designs with the highest affinity. We obtain binders to four diverse molecules, including the polar and flexible methotrexate and thyroxine. Taking advantage of the modular repeat structure and central binding pockets, we construct chemically induced dimerization systems and low-noise nanopore sensors by splitting designs into domains that reassemble upon ligand addition.

Deep-learning methods have revolutionized protein structure prediction and design but are presently limited to protein-only systems. We describe RoseTTAFold All-Atom (RFAA), which combines a residue-based representation of amino acids and DNA bases with an atomic representation of all other groups to model assemblies that contain proteins, nucleic acids, small molecules, metals, and covalent modifications, given their sequences and chemical structures. By fine-tuning on denoising tasks, we developed RFdiffusion All-Atom (RFdiffusionAA), which builds protein structures around small molecules. Starting from random distributions of amino acid residues surrounding target small molecules, we designed and experimentally validated, through crystallography and binding measurements, proteins that bind the cardiac disease therapeutic digoxigenin, the enzymatic cofactor heme, and the light-harvesting molecule bilin.

Mapping the ensemble of protein conformations that contribute to function and can be targeted by small molecule drugs remains an outstanding challenge. Here, we explore the use of variational autoencoders for reducing the challenge of dimensionality in the protein structure ensemble generation problem. We convert high-dimensional protein structural data into a continuous, low-dimensional representation, carry out a search in this space guided by a structure quality metric, and then use RoseTTAFold guided by the sampled structural information to generate 3D structures. We use this approach to generate ensembles for the cancer relevant protein K-Ras, train the VAE on a subset of the available K-Ras crystal structures and MD simulation snapshots, and assess the extent of sampling close to crystal structures withheld from training. We find that our latent space sampling procedure rapidly generates ensembles with high structural quality and is able to sample within 1 Å of held-out crystal structures, with a consistency higher than that of MD simulation or AlphaFold2 prediction. The sampled structures sufficiently recapitulate the cryptic pockets in the held-out K-Ras structures to allow for small molecule docking.

The design of protein-protein interfaces using physics-based design methods such as Rosetta requires substantial computational resources and manual refinement by expert structural biologists. Deep learning methods promise to simplify protein-protein interface design and enable its application to a wide variety of problems by researchers from various scientific disciplines. Here, we test the ability of a deep learning method for protein sequence design, ProteinMPNN, to design two-component tetrahedral protein nanomaterials and benchmark its performance against Rosetta. ProteinMPNN had a similar success rate to Rosetta, yielding 13 new experimentally confirmed assemblies, but required orders of magnitude less computation and no manual refinement. The interfaces designed by ProteinMPNN were substantially more polar than those designed by Rosetta, which facilitated in vitro assembly of the designed nanomaterials from independently purified components. Crystal structures of several of the assemblies confirmed the accuracy of the design method at high resolution. Our results showcase the potential of deep learning-based methods to unlock the widespread application of designed protein-protein interfaces and self-assembling protein nanomaterials in biotechnology., Competing Interests: Competing interests statement:N.P.K. is a cofounder, shareholder, paid consultant, and chair of the scientific advisory board of Icosavax, Inc. The King lab has received unrelated sponsored research agreements from Pfizer and GlaxoSmithKline.

A wooden house frame consists of many different lumber pieces, but because of the regularity of these building blocks, the structure can be designed using straightforward geometrical principles. The design of multicomponent protein assemblies, in comparison, has been much more complex, largely owing to the irregular shapes of protein structures 1 . Here we describe extendable linear, curved and angled protein building blocks, as well as inter-block interactions, that conform to specified geometric standards; assemblies designed using these blocks inherit their extendability and regular interaction surfaces, enabling them to be expanded or contracted by varying the number of modules, and reinforced with secondary struts. Using X-ray crystallography and electron microscopy, we validate nanomaterial designs ranging from simple polygonal and circular oligomers that can be concentrically nested, up to large polyhedral nanocages and unbounded straight 'train track' assemblies with reconfigurable sizes and geometries that can be readily blueprinted. Because of the complexity of protein structures and sequence-structure relationships, it has not previously been possible to build up large protein assemblies by deliberate placement of protein backbones onto a blank three-dimensional canvas; the simplicity and geometric regularity of our design platform now enables construction of protein nanomaterials according to 'back of an envelope' architectural blueprints., (© 2024. The Author(s).)

Many peptide hormones form an α-helix on binding their receptors 1-4 , and sensitive methods for their detection could contribute to better clinical management of disease 5 . De novo protein design can now generate binders with high affinity and specificity to structured proteins 6,7 . However, the design of interactions between proteins and short peptides with helical propensity is an unmet challenge. Here we describe parametric generation and deep learning-based methods for designing proteins to address this challenge. We show that by extending RFdiffusion 8 to enable binder design to flexible targets, and to refining input structure models by successive noising and denoising (partial diffusion), picomolar-affinity binders can be generated to helical peptide targets by either refining designs generated with other methods, or completely de novo starting from random noise distributions without any subsequent experimental optimization. The RFdiffusion designs enable the enrichment and subsequent detection of parathyroid hormone and glucagon by mass spectrometry, and the construction of bioluminescence-based protein biosensors. The ability to design binders to conformationally variable targets, and to optimize by partial diffusion both natural and designed proteins, should be broadly useful., (© 2023. The Author(s).)

Protein crystallization plays a central role in structural biology. Despite this, the process of crystallization remains poorly understood and highly empirical, with crystal contacts, lattice packing arrangements and space group preferences being largely unpredictable. Programming protein crystallization through precisely engineered side-chain-side-chain interactions across protein-protein interfaces is an outstanding challenge. Here we develop a general computational approach for designing three-dimensional protein crystals with prespecified lattice architectures at atomic accuracy that hierarchically constrains the overall number of degrees of freedom of the system. We design three pairs of oligomers that can be individually purified, and upon mixing, spontaneously self-assemble into >100 µm three-dimensional crystals. The structures of these crystals are nearly identical to the computational design models, closely corresponding in both overall architecture and the specific protein-protein interactions. The dimensions of the crystal unit cell can be systematically redesigned while retaining the space group symmetry and overall architecture, and the crystals are extremely porous and highly stable. Our approach enables the computational design of protein crystals with high accuracy, and the designed protein crystals, which have both structural and assembly information encoded in their primary sequences, provide a powerful platform for biological materials engineering., (© 2023. The Author(s), under exclusive licence to Springer Nature Limited.)

Controlling the biodistribution of protein- and nanoparticle-based therapeutic formulations remains challenging. In vivo library selection is an effective method for identifying constructs that exhibit desired distribution behavior; library variants can be selected based on their ability to localize to the tissue or compartment of interest despite complex physiological challenges. Here, we describe further development of an in vivo library selection platform based on self-assembling protein nanoparticles encapsulating their own mRNA genomes (synthetic nucleocapsids or synNCs). We tested two distinct libraries: a low-diversity library composed of synNC surface mutations (45 variants) and a high-diversity library composed of synNCs displaying miniproteins with binder-like properties (6.2 million variants). While we did not identify any variants from the low-diversity surface library that yielded therapeutically relevant changes in biodistribution, the high-diversity miniprotein display library yielded variants that shifted accumulation toward lungs or muscles in just two rounds of in vivo selection. Our approach should contribute to achieving specific tissue homing patterns and identifying targeting ligands for diseases of interest.

In pseudocyclic proteins, such as TIM barrels, β barrels, and some helical transmembrane channels, a single subunit is repeated in a cyclic pattern, giving rise to a central cavity that can serve as a pocket for ligand binding or enzymatic activity. Inspired by these proteins, we devised a deep-learning-based approach to broadly exploring the space of closed repeat proteins starting from only a specification of the repeat number and length. Biophysical data for 38 structurally diverse pseudocyclic designs produced in Escherichia coli are consistent with the design models, and the three crystal structures we were able to obtain are very close to the designed structures. Docking studies suggest the diversity of folds and central pockets provide effective starting points for designing small-molecule binders and enzymes., (© 2023. The Author(s).)

Predicting the effects of mutations on protein function and stability is an outstanding challenge. Here, we assess the performance of a variant of RoseTTAFold jointly trained for sequence and structure recovery, RF joint , for mutation effect prediction. Without any further training, we achieve comparable accuracy in predicting mutation effects for a diverse set of protein families using RF joint to both another zero-shot model (MSA Transformer) and a model that requires specific training on a particular protein family for mutation effect prediction (DeepSequence). Thus, although the architecture of RF joint was developed to address the protein design problem of scaffolding functional motifs, RF joint acquired an understanding of the mutational landscapes of proteins during model training that is equivalent to that of recently developed large protein language models. The ability to simultaneously reason over protein structure and sequence could enable even more precise mutation effect predictions following supervised training on the task. These results suggest that RF joint has a quite broad understanding of protein sequence-structure landscapes, and can be viewed as a joint model for protein sequence and structure which could be broadly useful for protein modeling., (© 2023 The Authors. Protein Science published by Wiley Periodicals LLC on behalf of The Protein Society.)

Metagenomes encode an enormous diversity of proteins, reflecting a multiplicity of functions and activities 1,2 . Exploration of this vast sequence space has been limited to a comparative analysis against reference microbial genomes and protein families derived from those genomes. Here, to examine the scale of yet untapped functional diversity beyond what is currently possible through the lens of reference genomes, we develop a computational approach to generate reference-free protein families from the sequence space in metagenomes. We analyse 26,931 metagenomes and identify 1.17 billion protein sequences longer than 35 amino acids with no similarity to any sequences from 102,491 reference genomes or the Pfam database 3 . Using massively parallel graph-based clustering, we group these proteins into 106,198 novel sequence clusters with more than 100 members, doubling the number of protein families obtained from the reference genomes clustered using the same approach. We annotate these families on the basis of their taxonomic, habitat, geographical and gene neighbourhood distributions and, where sufficient sequence diversity is available, predict protein three-dimensional models, revealing novel structures. Overall, our results uncover an enormously diverse functional space, highlighting the importance of further exploring the microbial functional dark matter., (© 2023. The Author(s).)

There has been considerable recent progress in designing new proteins using deep-learning methods 1-9 . 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 models 10,11 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., (© 2023. The Author(s).)

With the recent success in calculating protein structures from amino acid sequences using artificial intelligence-based algorithms, an important next step is to decipher how dynamics is encoded by the primary protein sequence so as to better predict function. Such dynamics information is critical for protein design, where strategies could then focus not only on sequences that fold into particular structures that perform a given task, but would also include low-lying excited protein states that could influence the function of the designed protein. Herein, we illustrate the importance of dynamics in modulating the function of C34, a designed α/β protein that captures β-strands of target ligands and is a member of a family of proteins designed to sequester β-strands and β hairpins of aggregation-prone molecules that lead to a variety of pathologies. Using a strategy to "see" regions of apo C34 that are invisible to NMR spectroscopy as a result of pervasive conformational exchange, as well as a mutagenesis approach whereby C34 molecules are stabilized into a single conformer, we determine the structures of the predominant conformations that are sampled by C34 and show that these attenuate the affinity for cognate peptide. Subsequently, the observed motion is exploited to develop an allosterically regulated peptide binder whose binding affinity can be controlled through the addition of a second molecule. Our study emphasizes the unique role that NMR can play in directing the design process and in the construction of new molecules with more complex functionality.

Recording transcriptional histories of a cell would enable deeper understanding of cellular developmental trajectories and responses to external perturbations. Here we describe an engineered protein fiber that incorporates diverse fluorescent marks during its growth to store a ticker tape-like history. An embedded HaloTag reporter incorporates user-supplied dyes, leading to colored stripes that map the growth of each individual fiber to wall clock time. A co-expressed eGFP tag driven by a promoter of interest records a history of transcriptional activation. High-resolution multi-spectral imaging on fixed samples reads the cellular histories, and interpolation of eGFP marks relative to HaloTag timestamps provides accurate absolute timing. We demonstrate recordings of doxycycline-induced transcription in HEK cells and cFos promoter activation in cultured neurons, with a single-cell absolute accuracy of 30-40 minutes over a 12-hour recording. The protein-based ticker tape design we present here could be generalized to achieve massively parallel single-cell recordings of diverse physiological modalities., (© 2023. The Author(s), under exclusive licence to Springer Nature America, Inc.)

As a result of evolutionary selection, the subunits of naturally occurring protein assemblies often fit together with substantial shape complementarity to generate architectures optimal for function in a manner not achievable by current design approaches. We describe a "top-down" reinforcement learning-based design approach that solves this problem using Monte Carlo tree search to sample protein conformers in the context of an overall architecture and specified functional constraints. Cryo-electron microscopy structures of the designed disk-shaped nanopores and ultracompact icosahedra are very close to the computational models. The icosohedra enable very-high-density display of immunogens and signaling molecules, which potentiates vaccine response and angiogenesis induction. Our approach enables the top-down design of complex protein nanomaterials with desired system properties and demonstrates the power of reinforcement learning in protein design.

General approaches for designing sequence-specific peptide-binding proteins would have wide utility in proteomics and synthetic biology. However, designing peptide-binding proteins is challenging, as most peptides do not have defined structures in isolation, and hydrogen bonds must be made to the buried polar groups in the peptide backbone 1-3 . Here, inspired by natural and re-engineered protein-peptide systems 4-11 , we set out to design proteins made out of repeating units that bind peptides with repeating sequences, with a one-to-one correspondence between the repeat units of the protein and those of the peptide. We use geometric hashing to identify protein backbones and peptide-docking arrangements that are compatible with bidentate hydrogen bonds between the side chains of the protein and the peptide backbone 12 . The remainder of the protein sequence is then optimized for folding and peptide binding. We design repeat proteins to bind to six different tripeptide-repeat sequences in polyproline II conformations. The proteins are hyperstable and bind to four to six tandem repeats of their tripeptide targets with nanomolar to picomolar affinities in vitro and in living cells. Crystal structures reveal repeating interactions between protein and peptide interactions as designed, including ladders of hydrogen bonds from protein side chains to peptide backbones. By redesigning the binding interfaces of individual repeat units, specificity can be achieved for non-repeating peptide sequences and for disordered regions of native proteins., (© 2023. The Author(s).)

Small beta barrel proteins are attractive targets for computational design because of their considerable functional diversity despite their very small size (<70 amino acids). However, there are considerable challenges to designing such structures, and there has been little success thus far. Because of the small size, the hydrophobic core stabilizing the fold is necessarily very small, and the conformational strain of barrel closure can oppose folding; also intermolecular aggregation through free beta strand edges can compete with proper monomer folding. Here, we explore the de novo design of small beta barrel topologies using both Rosetta energy-based methods and deep learning approaches to design four small beta barrel folds: Src homology 3 (SH3) and oligonucleotide/oligosaccharide-binding (OB) topologies found in nature and five and six up-and-down-stranded barrels rarely if ever seen in nature. Both approaches yielded successful designs with high thermal stability and experimentally determined structures with less than 2.4 Å rmsd from the designed models. Using deep learning for backbone generation and Rosetta for sequence design yielded higher design success rates and increased structural diversity than Rosetta alone. The ability to design a large and structurally diverse set of small beta barrel proteins greatly increases the protein shape space available for designing binders to protein targets of interest.

A challenge for design of protein-small-molecule recognition is that incorporation of cavities with size, shape, and composition suitable for specific recognition can considerably destabilize protein monomers. This challenge can be overcome through binding pockets formed at homo-oligomeric interfaces between folded monomers. Interfaces surrounding the central homo-oligomer symmetry axes necessarily have the same symmetry and so may not be well suited to binding asymmetric molecules. To enable general recognition of arbitrary asymmetric substrates and small molecules, we developed an approach to designing asymmetric interfaces at off-axis sites on homo-oligomers, analogous to those found in native homo-oligomeric proteins such as glutamine synthetase. We symmetrically dock curved helical repeat proteins such that they form pockets at the asymmetric interface of the oligomer with sizes ranging from several angstroms, appropriate for binding a single ion, to up to more than 20 Å across. Of the 133 proteins tested, 84 had soluble expression in E. coli , 47 had correct oligomeric states in solution, 35 had small-angle X-ray scattering (SAXS) data largely consistent with design models, and 8 had negative-stain electron microscopy (nsEM) 2D class averages showing the structures coming together as designed. Both an X-ray crystal structure and a cryogenic electron microscopy (cryoEM) structure are close to the computational design models. The nature of these proteins as homo-oligomers allows them to be readily built into higher-order structures such as nanocages, and the asymmetric pockets of these structures open rich possibilities for small-molecule binder design free from the constraints associated with monomer destabilization.

The de novo design of three protein chains that associate to form a heterotrimer (but not any of the possible two-chain heterodimers) and that can drive the assembly of higher-order branching structures is an important challenge for protein design. We designed helical heterotrimers with specificity conferred by buried hydrogen bond networks and large aromatic residues to enhance shape complementary packing. We obtained ten designs for which all three chains cooperatively assembled into heterotrimers with few or no other species present. Crystal structures of a helical bundle heterotrimer and extended versions, with helical repeat proteins fused to individual subunits, showed all three chains assembling in the designed orientation. We used these heterotrimers as building blocks to construct larger cyclic oligomers, which were structurally validated by electron microscopy. Our three-way junction designs provide new routes to complex protein nanostructures and enable the scaffolding of three distinct ligands for modulation of cell signaling., (© 2022. The Author(s).)

While native scaffolds offer a large diversity of shapes and topologies for enzyme engineering, their often unpredictable behavior in response to sequence modification makes de novo generated scaffolds an exciting alternative. Here we explore the customization of the backbone and sequence of a de novo designed eight stranded β-barrel protein to create catalysts for a retro-aldolase model reaction. We show that active and specific catalysts can be designed in this fold and use directed evolution to further optimize activity and stereoselectivity. Our results support previous suggestions that different folds have different inherent amenability to evolution and this property could account, in part, for the distribution of natural enzymes among different folds., (© 2022 The Authors. Protein Science published by Wiley Periodicals LLC on behalf of The Protein Society.)

Although 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, that has outstanding performance in both in silico and experimental tests. On native protein backbones, ProteinMPNN has a sequence recovery of 52.4% compared with 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, cryo-electron microscopy, and functional studies by rescuing previously failed designs, which were made using Rosetta or AlphaFold, of protein monomers, cyclic homo-oligomers, tetrahedral nanoparticles, and target-binding proteins.

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.

Mutations in human proteins lead to diseases. The structure of these proteins can help understand the mechanism of such diseases and develop therapeutics against them. With improved deep learning techniques, such as RoseTTAFold and AlphaFold, we can predict the structure of proteins even in the absence of structural homologs. We modeled and extracted the domains from 553 disease-associated human proteins without known protein structures or close homologs in the Protein Databank. We noticed that the model quality was higher and the Root mean square deviation (RMSD) lower between AlphaFold and RoseTTAFold models for domains that could be assigned to CATH families as compared to those which could only be assigned to Pfam families of unknown structure or could not be assigned to either. We predicted ligand-binding sites, protein-protein interfaces and conserved residues in these predicted structures. We then explored whether the disease-associated missense mutations were in the proximity of these predicted functional sites, whether they destabilized the protein structure based on ddG calculations or whether they were predicted to be pathogenic. We could explain 80% of these disease-associated mutations based on proximity to functional sites, structural destabilization or pathogenicity. When compared to polymorphisms, a larger percentage of disease-associated missense mutations were buried, closer to predicted functional sites, predicted as destabilizing and pathogenic. Usage of models from the two state-of-the-art techniques provide better confidence in our predictions, and we explain 93 additional mutations based on RoseTTAFold models which could not be explained based solely on AlphaFold models., (© The Author(s) 2022. Published by Oxford University Press.)

The ability to design efficient enzymes from scratch would have a profound effect on chemistry, biotechnology and medicine. Rapid progress in protein engineering over the past decade makes us optimistic that this ambition is within reach. The development of artificial enzymes containing metal cofactors and noncanonical organocatalytic groups shows how protein structure can be optimized to harness the reactivity of nonproteinogenic elements. In parallel, computational methods have been used to design protein catalysts for diverse reactions on the basis of fundamental principles of transition state stabilization. Although the activities of designed catalysts have been quite low, extensive laboratory evolution has been used to generate efficient enzymes. Structural analysis of these systems has revealed the high degree of precision that will be needed to design catalysts with greater activity. To this end, emerging protein design methods, including deep learning, hold particular promise for improving model accuracy. Here we take stock of key developments in the field and highlight new opportunities for innovation that should allow us to transition beyond the current state of the art and enable the robust design of biocatalysts to address societal needs., (© 2022. Springer Nature Limited.)

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 challenge 1-5 . 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., (© 2022. This is a U.S. government work and not under copyright protection in the U.S.; foreign copyright protection may apply.)

Natural molecular machines contain protein components that undergo motion relative to each other. Designing such mechanically constrained nanoscale protein architectures with internal degrees of freedom is an outstanding challenge for computational protein design. Here we explore the de novo construction of protein machinery from designed axle and rotor components with internal cyclic or dihedral symmetry. We find that the axle-rotor systems assemble in vitro and in vivo as designed. Using cryo-electron microscopy, we find that these systems populate conformationally variable relative orientations reflecting the symmetry of the coupled components and the computationally designed interface energy landscape. These mechanical systems with internal degrees of freedom are a step toward the design of genetically encodable nanomachines.

Assembly of biomolecules at solid–water interfaces requires molecules to traverse complex orientation-dependent energy landscapes through processes that are poorly understood, largely due to the dearth of in situ single-molecule measurements and statistical analyses of the rotational dynamics that define directional selection. Emerging capabilities in high-speed atomic force microscopy and machine learning have allowed us to directly determine the orientational energy landscape and observe and quantify the rotational dynamics for protein nanorods on the surface of muscovite mica under a variety of conditions. Comparisons with kinetic Monte Carlo simulations show that the transition rates between adjacent orientation-specific energetic minima can largely be understood through traditional models of in-plane Brownian rotation across a biased energy landscape, with resulting transition rates that are exponential in the energy barriers between states. However, transitions between more distant angular states are decoupled from barrier height, with jump-size distributions showing a power law decay that is characteristic of a nonclassical Levy-flight random walk, indicating that large jumps are enabled by alternative modes of motion via activated states. The findings provide insights into the dynamics of biomolecules at solid–liquid interfaces that lead to self-assembly, epitaxial matching, and other orientationally anisotropic outcomes and define a general procedure for exploring such dynamics with implications for hybrid biomolecular–inorganic materials design.

Engineered proteins generally must possess a stable structure in order to achieve their designed function. Stable designs, however, are astronomically rare within the space of all possible amino acid sequences. As a consequence, many designs must be tested computationally and experimentally in order to find stable ones, which is expensive in terms of time and resources. Here we use a high-throughput, low-fidelity assay to experimentally evaluate the stability of approximately 200,000 novel proteins. These include a wide range of sequence perturbations, providing a baseline for future work in the field. We build a neural network model that predicts protein stability given only sequences of amino acids, and compare its performance to the assayed values. We also report another network model that is able to generate the amino acid sequences of novel stable proteins given requested secondary sequences. Finally, we show that the predictive model-despite weaknesses including a noisy data set-can be used to substantially increase the stability of both expert-designed and model-generated proteins., Competing Interests: JMS and AZ are employed by Two Six Technologies, which has filed a patent on a portion of the technology described in this manuscript. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

The combination of computational design and directed evolution could offer a general strategy to create enzymes with new functions. So far, this approach has delivered enzymes for a handful of model reactions. Here we show that new catalytic mechanisms can be engineered into proteins to accelerate more challenging chemical transformations. Evolutionary optimization of a primitive design afforded an efficient and enantioselective enzyme (BH32.14) for the Morita-Baylis-Hillman (MBH) reaction. BH32.14 is suitable for preparative-scale transformations, accepts a broad range of aldehyde and enone coupling partners and is able to promote selective monofunctionalizations of dialdehydes. Crystallographic, biochemical and computational studies reveal that BH32.14 operates via a sophisticated catalytic mechanism comprising a His23 nucleophile paired with a judiciously positioned Arg124. This catalytic arginine shuttles between conformational states to stabilize multiple oxyanion intermediates and serves as a genetically encoded surrogate of privileged bidentate hydrogen-bonding catalysts (for example, thioureas). This study demonstrates that elaborate catalytic devices can be built from scratch to promote demanding multi-step processes not observed in nature., (© 2021. The Author(s), under exclusive licence to Springer Nature Limited.)

Asymmetric multiprotein complexes that undergo subunit exchange play central roles in biology but present a challenge for design because the components must not only contain interfaces that enable reversible association but also be stable and well behaved in isolation. We use implicit negative design to generate β sheet-mediated heterodimers that can be assembled into a wide variety of complexes. The designs are stable, folded, and soluble in isolation and rapidly assemble upon mixing, and crystal structures are close to the computational models. We construct linearly arranged hetero-oligomers with up to six different components, branched hetero-oligomers, closed C4-symmetric two-component rings, and hetero-oligomers assembled on a cyclic homo-oligomeric central hub and demonstrate that such complexes can readily reconfigure through subunit exchange. Our approach provides a general route to designing asymmetric reconfigurable protein systems.

Natural aldolase enzymes and created retro-aldolase protein catalysts often catalyze both aldol and retro-aldol reactions depending on the concentrations of the reactants and the products. Here, we report that the directionality of protein catalysts can be altered by replacing one amino acid. The protein catalyst derived from a scaffold of a previously reported retro-aldolase catalyst, catalyzed aldol reactions more efficiently than the previously reported retro-aldolase catalyst. The retro-aldolase catalyst efficiently catalyzed the retro-aldol reaction but was less efficient in catalyzing the aldol reaction. The results indicate that protein catalysts with varying levels of directionality in usually reversibly catalyzed aldol and retro-aldol reactions can be generated from the same protein scaffold., (© 2021 Wiley-VCH GmbH.)

The established approach to unsupervised protein contact prediction estimates coevolving positions using undirected graphical models. This approach trains a Potts model on a Multiple Sequence Alignment. Increasingly large Transformers are being pretrained on unlabeled, unaligned protein sequence databases and showing competitive performance on protein contact prediction. We argue that attention is a principled model of protein interactions, grounded in real properties of protein family data. We introduce an energy-based attention layer, factored attention, which, in a certain limit, recovers a Potts model, and use it to contrast Potts and Transformers. We show that the Transformer leverages hierarchical signal in protein family databases not captured by single-layer models. This raises the exciting possibility for the development of powerful structured models of protein family databases.

The trRosetta (transform-restrained Rosetta) server is a web-based platform for fast and accurate protein structure prediction, powered by deep learning and Rosetta. With the input of a protein's amino acid sequence, a deep neural network is first used to predict the inter-residue geometries, including distance and orientations. The predicted geometries are then transformed as restraints to guide the structure prediction on the basis of direct energy minimization, which is implemented under the framework of Rosetta. The trRosetta server distinguishes itself from other similar structure prediction servers in terms of rapid and accurate de novo structure prediction. As an illustration, trRosetta was applied to two Pfam families with unknown structures, for which the predicted de novo models were estimated to have high accuracy. Nevertheless, to take advantage of homology modeling, homologous templates are used as additional inputs to the network automatically. In general, it takes ~1 h to predict the final structure for a typical protein with ~300 amino acids, using a maximum of 10 CPU cores in parallel in our cluster system. To enable large-scale structure modeling, a downloadable package of trRosetta with open-source codes is available as well. A detailed guidance for using the package is also available in this protocol. The server and the package are available at https://yanglab.nankai.edu.cn/trRosetta/ and https://yanglab.nankai.edu.cn/trRosetta/download/ , respectively., (© 2021. The Author(s), under exclusive licence to Springer Nature Limited.)

The trRosetta structure prediction method employs deep learning to generate predicted residue-residue distance and orientation distributions from which 3D models are built. We sought to improve the method by incorporating as inputs (in addition to sequence information) both language model embeddings and template information weighted by sequence similarity to the target. We also developed a refinement pipeline that recombines models generated by template-free and template utilizing versions of trRosetta guided by the DeepAccNet accuracy predictor. Both benchmark tests and CASP results show that the new pipeline is a considerable improvement over the original trRosetta, and it is faster and requires less computing resources, completing the entire modeling process in a median < 3 h in CASP14. Our human group improved results with this pipeline primarily by identifying additional homologous sequences for input into the network. We also used the DeepAccNet accuracy predictor to guide Rosetta high-resolution refinement for submissions in the regular and refinement categories; although performance was quite good on a CASP relative scale, the overall improvements were rather modest in part due to missing inter-domain or inter-chain contacts., (© 2021 The Authors. Proteins: Structure, Function, and Bioinformatics published by Wiley Periodicals LLC.)

For CASP14, we developed deep learning-based methods for predicting homo-oligomeric and hetero-oligomeric contacts and used them for oligomer modeling. To build structure models, we developed an oligomer structure generation method that utilizes predicted interchain contacts to guide iterative restrained minimization from random backbone structures. We supplemented this gradient-based fold-and-dock method with template-based and ab initio docking approaches using deep learning-based subunit predictions on 29 assembly targets. These methods produced oligomer models with summed Z-scores 5.5 units higher than the next best group, with the fold-and-dock method having the best relative performance. Over the eight targets for which this method was used, the best of the five submitted models had average oligomer TM-score of 0.71 (average oligomer TM-score of the next best group: 0.64), and explicit modeling of inter-subunit interactions improved modeling of six out of 40 individual domains (ΔGDT-TS > 2.0)., (© 2021 The Authors. Proteins: Structure, Function, and Bioinformatics published by Wiley Periodicals LLC.)

There has been considerable recent progress in protein structure prediction using deep neural networks to predict inter-residue distances from amino acid sequences 1-3 . Here we investigate whether the information captured by such networks is sufficiently rich to generate new folded proteins with sequences unrelated to those of the naturally occurring proteins used in training the models. We generate random amino acid sequences, and input them into the trRosetta structure prediction network to predict starting residue-residue distance maps, which, as expected, are quite featureless. We then carry out Monte Carlo sampling in amino acid sequence space, optimizing the contrast (Kullback-Leibler divergence) between the inter-residue distance distributions predicted by the network and background distributions averaged over all proteins. Optimization from different random starting points resulted in novel proteins spanning a wide range of sequences and predicted structures. We obtained synthetic genes encoding 129 of the network-'hallucinated' sequences, and expressed and purified the proteins in Escherichia coli; 27 of the proteins yielded monodisperse species with circular dichroism spectra consistent with the hallucinated structures. We determined the three-dimensional structures of three of the hallucinated proteins, two by X-ray crystallography and one by NMR, and these closely matched the hallucinated models. Thus, deep networks trained to predict native protein structures from their sequences can be inverted to design new proteins, and such networks and methods should contribute alongside traditional physics-based models to the de novo design of proteins with new functions., (© 2021. The Author(s), under exclusive licence to Springer Nature Limited.)

Each year vast international resources are wasted on irreproducible research. The scientific community has been slow to adopt standard software engineering practices, despite the increases in high-dimensional data, complexities of workflows, and computational environments. Here we show how scientific software applications can be created in a reproducible manner when simple design goals for reproducibility are met. We describe the implementation of a test server framework and 40 scientific benchmarks, covering numerous applications in Rosetta bio-macromolecular modeling. High performance computing cluster integration allows these benchmarks to run continuously and automatically. Detailed protocol captures are useful for developers and users of Rosetta and other macromolecular modeling tools. The framework and design concepts presented here are valuable for developers and users of any type of scientific software and for the scientific community to create reproducible methods. Specific examples highlight the utility of this framework, and the comprehensive documentation illustrates the ease of adding new tests in a matter of hours., (© 2021. The Author(s).)

The ability to design stable proteins with custom-made functions is a major goal in biochemistry with practical relevance for our environment and society. Understanding and manipulating protein stability provide crucial information on the molecular determinants that modulate structure and stability, and expand the applications of de novo proteins. Since the (β/⍺) 8 -barrel or TIM-barrel fold is one of the most common functional scaffolds, in this work we designed a collection of stable de novo TIM barrels (DeNovoTIMs), using a computational fixed-backbone and modular approach based on improved hydrophobic packing of sTIM11, the first validated de novo TIM barrel, and subjected them to a thorough folding analysis. DeNovoTIMs navigate a region of the stability landscape previously uncharted by natural TIM barrels, with variations spanning 60 degrees in melting temperature and 22 kcal per mol in conformational stability throughout the designs. Significant non-additive or epistatic effects were observed when stabilizing mutations from different regions of the barrel were combined. The molecular basis of epistasis in DeNovoTIMs appears to be related to the extension of the hydrophobic cores. This study is an important step towards the fine-tuned modulation of protein stability by design., (Copyright © 2021 The Authors. Published by Elsevier Ltd.. All rights reserved.)

DeepMind presented notably accurate predictions at the recent 14th Critical Assessment of Structure Prediction (CASP14) conference. We explored network architectures that incorporate related ideas and obtained the best performance with a three-track network in which information at the one-dimensional (1D) sequence level, the 2D distance map level, and the 3D coordinate level is successively transformed and integrated. The three-track network produces structure predictions with accuracies approaching those of DeepMind in CASP14, enables the rapid solution of challenging x-ray crystallography and cryo-electron microscopy structure modeling problems, and provides insights into the functions of proteins of currently unknown structure. The network also enables rapid generation of accurate protein-protein complex models from sequence information alone, short-circuiting traditional approaches that require modeling of individual subunits followed by docking. We make the method available to the scientific community to speed biological research., (Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.)

The phyllosilicate mineral muscovite mica is widely used as a surface template for the patterning of macromolecules, yet a molecular understanding of its surface chemistry under varying solution conditions, required to predict and control the self-assembly of adsorbed species, is lacking. We utilize all-atom molecular dynamics simulations in conjunction with an electrostatic analysis based in local molecular field theory that affords a clean separation of long-range and short-range electrostatics. Using water polarization response as a measure of the electric fields that arise from patterned, surface-bound ions that direct the adsorption of charged macromolecules, we apply a Landau theory of forces induced by asymmetrically polarized surfaces to compute protein-surface interactions for two muscovite-binding proteins (DHR10-mica6 and C98 RhuA). Comparison of the pressure between surface and protein in high-concentration KCl and NaCl aqueous solutions reveals ion-specific differences in far-field protein-surface interactions, neatly capturing the ability of ions to modulate the surface charge of muscovite that in turn selectively attracts one binding face of each protein over all others., Competing Interests: The authors declare no competing interest.

We previously elucidated principles for designing ideal proteins with completely consistent local and non-local interactions which have enabled the design of a wide range of new αβ-proteins with four or fewer β-strands. The principles relate local backbone structures to supersecondary-structure packing arrangements of α-helices and β-strands. Here, we test the generality of the principles by employing them to design larger proteins with five- and six- stranded β-sheets flanked by α-helices. The initial designs were monomeric in solution with high thermal stability, and the nuclear magnetic resonance (NMR) structure of one was close to the design model, but for two others the order of strands in the β-sheet was swapped. Investigation into the origins of this strand swapping suggested that the global structures of the design models were more strained than the NMR structures. We incorporated explicit consideration of global backbone strain into the design methodology, and succeeded in designing proteins with the intended unswapped strand arrangements. These results illustrate the value of experimental structure determination in guiding improvement of de novo design, and the importance of consistency between local, supersecondary, and global tertiary interactions in determining protein topology. The augmented set of principles should inform the design of larger functional proteins.

Protein nanomaterial design is an emerging discipline with applications in medicine and beyond. A long-standing design approach uses genetic fusion to join protein homo-oligomer subunits via α-helical linkers to form more complex symmetric assemblies, but this method is hampered by linker flexibility and a dearth of geometric solutions. Here, we describe a general computational method for rigidly fusing homo-oligomer and spacer building blocks to generate user-defined architectures that generates far more geometric solutions than previous approaches. The fusion junctions are then optimized using Rosetta to minimize flexibility. We apply this method to design and test 92 dihedral symmetric protein assemblies using a set of designed homodimers and repeat protein building blocks. Experimental validation by native mass spectrometry, small-angle X-ray scattering, and negative-stain single-particle electron microscopy confirms the assembly states for 11 designs. Most of these assemblies are constructed from designed ankyrin repeat proteins (DARPins), held in place on one end by α-helical fusion and on the other by a designed homodimer interface, and we explored their use for cryogenic electron microscopy (cryo-EM) structure determination by incorporating DARPin variants selected to bind targets of interest. Although the target resolution was limited by preferred orientation effects and small scaffold size, we found that the dual anchoring strategy reduced the flexibility of the target-DARPIN complex with respect to the overall assembly, suggesting that multipoint anchoring of binding domains could contribute to cryo-EM structure determination of small proteins., Competing Interests: The authors declare no competing interest.

The FastDesign protocol in the molecular modeling program Rosetta iterates between sequence optimization and structure refinement to stabilize de novo designed protein structures and complexes. FastDesign has been used previously to design novel protein folds and assemblies with important applications in research and medicine. To promote sampling of alternative conformations and sequences, FastDesign includes stages where the energy landscape is smoothened by reducing repulsive forces. Here, we discover that this process disfavors larger amino acids in the protein core because the protein compresses in the early stages of refinement. By testing alternative ramping strategies for the repulsive weight, we arrive at a scheme that produces lower energy designs with more native-like sequence composition in the protein core. We further validate the protocol by designing and experimentally characterizing over 4000 proteins and show that the new protocol produces higher stability proteins., (© 2020 Wiley Periodicals LLC.)

The protein design problem is to identify an amino acid sequence that folds to a desired structure. Given Anfinsen's thermodynamic hypothesis of folding, this can be recast as finding an amino acid sequence for which the desired structure is the lowest energy state. As this calculation involves not only all possible amino acid sequences but also, all possible structures, most current approaches focus instead on the more tractable problem of finding the lowest-energy amino acid sequence for the desired structure, often checking by protein structure prediction in a second step that the desired structure is indeed the lowest-energy conformation for the designed sequence, and typically discarding a large fraction of designed sequences for which this is not the case. Here, we show that by backpropagating gradients through the transform-restrained Rosetta (trRosetta) structure prediction network from the desired structure to the input amino acid sequence, we can directly optimize over all possible amino acid sequences and all possible structures in a single calculation. We find that trRosetta calculations, which consider the full conformational landscape, can be more effective than Rosetta single-point energy estimations in predicting folding and stability of de novo designed proteins. We compare sequence design by conformational landscape optimization with the standard energy-based sequence design methodology in Rosetta and show that the former can result in energy landscapes with fewer alternative energy minima. We show further that more funneled energy landscapes can be designed by combining the strengths of the two approaches: the low-resolution trRosetta model serves to disfavor alternative states, and the high-resolution Rosetta model serves to create a deep energy minimum at the design target structure., Competing Interests: The authors declare no competing interest., (Copyright © 2021 the Author(s). Published by PNAS.)

Accurate and rapid calculation of protein-small molecule interaction free energies is critical for computational drug discovery. Because of the large chemical space spanned by drug-like molecules, classical force fields contain thousands of parameters describing atom-pair distance and torsional preferences; each parameter is typically optimized independently on simple representative molecules. Here, we describe a new approach in which small molecule force field parameters are jointly optimized guided by the rich source of information contained within thousands of available small molecule crystal structures. We optimize parameters by requiring that the experimentally determined molecular lattice arrangements have lower energy than all alternative lattice arrangements. Thousands of independent crystal lattice-prediction simulations were run on each of 1386 small molecule crystal structures, and energy function parameters of an implicit solvent energy model were optimized, so native crystal lattice arrangements had the lowest energy. The resulting energy model was implemented in Rosetta, together with a rapid genetic algorithm docking method employing grid-based scoring and receptor flexibility. The success rate of bound structure recapitulation in cross-docking on 1112 complexes was improved by more than 10% over previously published methods, with solutions within <1 Å in over half of the cases. Our results demonstrate that small molecule crystal structures are a rich source of information for guiding molecular force field development, and the improved Rosetta energy function should increase accuracy in a wide range of small molecule structure prediction and design studies.

In aqueous solution, polar groups make hydrogen bonds with water, and hence burial of such groups in the interior of a protein is unfavorable unless the loss of hydrogen bonds with water is compensated by formation of new ones with other protein groups. For this reason, buried "unsatisfied" polar groups making no hydrogen bonds are very rare in proteins. Efficiently representing the energetic cost of unsatisfied hydrogen bonds with a pairwise-decomposable energy term during protein design is challenging since whether or not a group is satisfied depends on all of its neighbors. Here we describe a method for assigning a pairwise-decomposable energy to sidechain rotamers such that following combinatorial sidechain packing, buried unsaturated polar atoms are penalized. The penalty can be any quadratic function of the number of unsatisfied polar groups, and can be computed very rapidly. We show that inclusion of this term in Rosetta sidechain packing calculations substantially reduces the number of buried unsatisfied polar groups., Competing Interests: The authors have declared that no competing interests exist.

We develop a deep learning framework (DeepAccNet) that estimates per-residue accuracy and residue-residue distance signed error in protein models and uses these predictions to guide Rosetta protein structure refinement. The network uses 3D convolutions to evaluate local atomic environments followed by 2D convolutions to provide their global contexts and outperforms other methods that similarly predict the accuracy of protein structure models. Overall accuracy predictions for X-ray and cryoEM structures in the PDB correlate with their resolution, and the network should be broadly useful for assessing the accuracy of both predicted structure models and experimentally determined structures and identifying specific regions likely to be in error. Incorporation of the accuracy predictions at multiple stages in the Rosetta refinement protocol considerably increased the accuracy of the resulting protein structure models, illustrating how deep learning can improve search for global energy minima of biomolecules.