Hallucinating symmetric protein assemblies

A self-assembled protein β-helix as a self-contained biofunctional motif

Nature constructs matter by employing protein folding motifs, many of which have been synthetically reconstituted to exploit function. A less understood motif whose structure-function relationships remain unexploited is formed by parallel β-strands arranged in a helical repetitive pattern, termed a β-helix. Herein we reconstitute a protein β-helix by design and endow it with biological function. Unlike β-helical proteins, which are contiguous covalent structures, this β-helix self-assembles from an elementary sequence of 18 amino acids. Using a combination of experimental and computational methods, we demonstrate that the resulting assemblies are discrete cylindrical structures exhibiting conserved dimensions at the nanoscale. We provide evidence for the structures to form a carpet-like three-dimensional scaffold promoting and inhibiting the growth of human and bacterial cells, respectively, while being able to mediate intracellular gene delivery. The study introduces a self-assembled β-helix as a self-contained bio- and multi-functional motif for exploring and exploiting mechanistic biology.

Computational design of serine hydrolases

The design of enzymes with complex active sites that mediate multistep reactions remains an outstanding challenge. With serine hydrolases as a model system, we combined the generative capabilities of RFdiffusion with an ensemble generation method for assessing active site preorganization at each step in the reaction to design enzymes starting from minimal active site descriptions. Experimental characterization revealed catalytic efficiencies (kcat/Km) up to 2.2 × 105 M-1 s-1 and crystal structures that closely match the design models (Cα root mean square deviations <1 angstrom). Selection for structural compatibility across the reaction coordinate enabled identification of new catalysts remove with five different folds distinct from those of natural serine hydrolases. Our de novo approach provides insight into the geometric basis of catalysis and a roadmap for designing enzymes that catalyze multistep transformations.

De novo design of transmembrane fluorescence-activating proteins

The recognition of ligands by transmembrane proteins is essential for the exchange of materials, energy and information across biological membranes. Progress has been made in the de novo design of transmembrane proteins1-6, as well as in designing water-soluble proteins to bind small molecules7-12, but de novo design of transmembrane proteins that tightly and specifically bind to small molecules remains an outstanding challenge13. Here we present the accurate design of ligand-binding transmembrane proteins by integrating deep learning and energy-based methods. We designed pre-organized ligand-binding pockets in high-quality four-helix backbones for a fluorogenic ligand, and generated a transmembrane span using gradient-guided hallucination. The designer transmembrane proteins specifically activated fluorescence of the target fluorophore with mid-nanomolar affinity, exhibiting higher brightness and quantum yield compared to those of enhanced green fluorescent protein. These proteins were highly active in the membrane fraction of live bacterial and eukaryotic cells following expression. The crystal and cryogenic electron microscopy structures of the designer protein-ligand complexes were very close to the structures of the design models. We showed that the interactions between ligands and transmembrane proteins within the membrane can be accurately designed. Our work paves the way for the creation of new functional transmembrane proteins, with a wide range of applications including imaging, ligand sensing and membrane transport.

Atomic context-conditioned protein sequence design using LigandMPNN

Protein sequence design in the context of small molecules, nucleotides and metals is critical to enzyme and small-molecule binder and sensor design, but current state-of-the-art deep-learning-based sequence design methods are unable to model nonprotein atoms and molecules. Here we describe a deep-learning-based protein sequence design method called LigandMPNN that explicitly models all nonprotein components of biomolecular systems. LigandMPNN significantly outperforms Rosetta and ProteinMPNN on native backbone sequence recovery for residues interacting with small molecules (63.3% versus 50.4% and 50.5%), nucleotides (50.5% versus 35.2% and 34.0%) and metals (77.5% versus 36.0% and 40.6%). LigandMPNN generates not only sequences but also sidechain conformations to allow detailed evaluation of binding interactions. LigandMPNN has been used to design over 100 experimentally validated small-molecule and DNA-binding proteins with high affinity and high structural accuracy (as indicated by four X-ray crystal structures), and redesign of Rosetta small-molecule binder designs has increased binding affinity by as much as 100-fold. We anticipate that LigandMPNN will be widely useful for designing new binding proteins, sensors and enzymes.

Advances in designed bionanomolecular assemblies for biotechnological and biomedical applications

Recent advances in protein engineering have revolutionized the design of bionanomolecular assemblies for functional therapeutic and biotechnological applications. This review highlights the progress in creating complex protein architectures, encompassing both finite and extended assemblies. AI tools, including AlphaFold, RFDiffusion, and ProteinMPNN, have significantly enhanced the scalability and success of de novo designs. Finite assemblies, like nanocages and coiled-coil-based structures, enable precise molecular encapsulation or functional protein domain presentation. Extended assemblies, including filaments and 2D/3D lattices, offer unparalleled structural versatility for applications such as vaccine development, responsive biomaterials, and engineered cellular scaffolds. The convergence of artificial intelligence-driven design and experimental validation promises strong acceleration of the development of tailored protein assemblies, offering new opportunities in synthetic biology, materials science, biotechnology, and biomedicine.

AlphaFold as a Prior: Experimental Structure Determination Conditioned on a Pretrained Neural Network

Advances in machine learning have transformed structural biology, enabling swift and accurate prediction of protein structure from sequence. However, challenges persist in capturing sidechain packing, condition-dependent conformational dynamics, and biomolecular interactions, primarily due to scarcity of high-quality training data. Emerging techniques, including cryo-electron tomography (cryo-ET) and high-throughput crystallography, promise vast new sources of structural data, but translating experimental observations into mechanistically interpretable atomic models remains a key bottleneck. Here, we address these challenges by improving the efficiency of structural analysis through combining experimental measurements with a landmark protein structure prediction method - AlphaFold2. We present an augmentation of AlphaFold2, ROCKET, that refines its predictions using cryo-EM, cryo-ET, and X-ray crystallography data, and demonstrate that this approach captures biologically important structural variation that AlphaFold2 does not. By performing structure optimization in the space of coevolutionary embeddings, rather than Cartesian coordinates, ROCKET automates difficult modeling tasks, such as flips of functional loops and domain rearrangements, that are beyond the scope of current state-of-the-art methods and, in some instances, even manual human modeling. The ability to efficiently sample these barrier-crossing rearrangements unlocks a new horizon for scalable and automated model building. Crucially, ROCKET does not require retraining of AlphaFold2 and is readily adaptable to multimers, ligand-cofolding, and other data modalities. Conversely, our differentiable crystal-lographic and cryo-EM target functions are capable of augmenting other structure prediction methods. ROCKET thus provides an extensible framework for the integration of experimental observables with biomolecular machine learning.

Computational scoring and experimental evaluation of enzymes generated by neural networks

In recent years, generative protein sequence models have been developed to sample novel sequences. However, predicting whether generated proteins will fold and function remains challenging. We evaluate a set of 20 diverse computational metrics to assess the quality of enzyme sequences produced by three contrasting generative models: ancestral sequence reconstruction, a generative adversarial network and a protein language model. Focusing on two enzyme families, we expressed and purified over 500 natural and generated sequences with 70-90% identity to the most similar natural sequences to benchmark computational metrics for predicting in vitro enzyme activity. Over three rounds of experiments, we developed a computational filter that improved the rate of experimental success by 50-150%. The proposed metrics and models will drive protein engineering research by serving as a benchmark for generative protein sequence models and helping to select active variants for experimental testing.

Accurately Models the Relationship Between Physical Response and Structure Using Kolmogorov-Arnold Network

Artificial intelligence (AI) in science is a key area of modern research. However, many current machine learning methods lack interpretability, making it difficult to grasp the physical mechanisms behind various phenomena, which hampers progress in related fields. This study focuses on the Poisson's ratio of a hexagonal lattice elastic network as it varies with structural deformation. By employing the Kolmogorov-Arnold Network (KAN), the transition of the network's Poisson's ratio from positive to negative as the hexagonal structural element shifts from a convex polygon to a concave polygon was accurately predicted. The KAN provides a clear mathematical framework that describes this transition, revealing the connection between the Poisson's ratio and the geometric properties of the hexagonal element, and accurately identifying the geometric parameters at which the Poisson's ratio equals zero. This work demonstrates the significant potential of the KAN network to clarify the mathematical relationships that underpin physical responses and structural behaviors.

Massively parallel assessment of designed protein solution properties using mass spectrometry and peptide barcoding

Library screening and selection methods can determine the binding activities of individual members of large protein libraries given a physical link between protein and nucleotide sequence, which enables identification of functional molecules by DNA sequencing. However, the solution properties of individual protein molecules cannot be probed using such approaches because they are completely altered by DNA attachment. Mass spectrometry enables parallel evaluation of protein properties amenable to physical fractionation such as solubility and oligomeric state, but current approaches are limited to libraries of 1,000 or fewer proteins. Here, we improved mass spectrometry barcoding by co-synthesizing proteins with barcodes optimized to be highly multiplexable and minimally perturbative, scaling to libraries of >5,000 proteins. We use these barcodes together with mass spectrometry to assay the solution behavior of libraries of de novo-designed monomeric scaffolds, oligomers, binding proteins and nanocages, rapidly identifying design failure modes and successes.

Rapid and accurate prediction of protein homo-oligomer symmetry using Seq2Symm

The majority of proteins must form higher-order assemblies to perform their biological functions, yet few machine learning models can accurately and rapidly predict the symmetry of assemblies involving multiple copies of the same protein chain. Here, we address this gap by finetuning several classes of protein foundation models, to predict homo-oligomer symmetry. Our best model named Seq2Symm, which utilizes ESM2, outperforms existing template-based and deep learning methods achieving an average AUC-PR of 0.47, 0.44 and 0.49 across homo-oligomer symmetries on three held-out test sets compared to 0.24, 0.24 and 0.25 with template-based search. Seq2Symm uses a single sequence as input and can predict at the rate of ~80,000 proteins/hour. We apply this method to 5 proteomes and ~3.5 million unlabeled protein sequences, showing its promise to be used in conjunction with downstream computationally intensive all-atom structure generation methods such as RoseTTAFold2 and AlphaFold2-multimer. Code, datasets, model are available at: https://github.com/microsoft/seq2symm .

AI-assisted protein design to rapidly convert antibody sequences to intrabodies targeting diverse peptides and histone modifications

Intrabodies are engineered antibodies that function inside living cells, enabling therapeutic, diagnostic, and imaging applications. While powerful, their development has been hindered by challenges associated with their folding, solubility, and stability in the reduced intracellular environment. Here, we present an AI-driven pipeline integrating AlphaFold2, ProteinMPNN, and live-cell screening to optimize antibody framework regions while preserving epitope-binding complementarity-determining regions. Using this approach, we successfully converted 19 out of 26 antibody sequences into functional single-chain variable fragment (scFv) intrabodies, including a panel targeting diverse histone modifications for real-time imaging of chromatin dynamics and gene regulation. Notably, 18 of these 19 sequences had failed to convert using the standard approach, demonstrating the unique effectiveness of our method. As antibody sequence databases expand, our method will accelerate intrabody design, making their development easier, more cost-effective, and broadly accessible for biological research.

Artificial intelligence in peptide-based drug design

Protein-protein interactions (PPIs) are fundamental to a variety of biological processes, but targeting them with small molecules is challenging because of their large and complex interaction interfaces. However, peptides have emerged as highly promising modulators of PPIs, because they can bind to protein surfaces with high affinity and specificity. Nonetheless, computational peptide design remains difficult, hindered by the intrinsic flexibility of peptides and the substantial computational resources required. Recent advances in artificial intelligence (AI) are paving new paths for peptide-based drug design. In this review, we explore the advanced deep generative models for designing target-specific peptide binders, highlight key challenges, and offer insights into the future direction of this rapidly evolving field.

Parametrically guided design of beta barrels and transmembrane nanopores using deep learning

Francis Crick's global parameterization of coiled coil geometry has been widely useful for guiding design of new protein structures and functions. However, design guided by similar global parameterization of beta barrel structures has been less successful, likely due to the deviations from ideal barrel geometry required to maintain inter-strand hydrogen bonding without introducing backbone strain. Instead, beta barrels have been designed using 2D structural blueprints; while this approach has successfully generated new fluorescent proteins, transmembrane nanopores, and other structures, it requires expert knowledge and provides only indirect control over the global shape. Here we show that the simplicity and control over shape and structure provided by parametric representations can be generalized beyond coiled coils by taking advantage of the rich sequence-structure relationships implicit in RoseTTAFold based design methods. Starting from parametrically generated barrel backbones, both RFjoint inpainting and RFdiffusion readily incorporate backbone irregularities necessary for proper folding with minimal deviation from the idealized barrel geometries. We show that for beta barrels across a broad range of beta sheet parameterizations, these methods achieve high in silico and experimental success rates, with atomic accuracy confirmed by an X-ray crystal structure of a novel barrel topology, and de novo designed 12, 14, and 16 stranded transmembrane nanopores with conductances ranging from 200 to 500 pS. By combining the simplicity and control of parametric generation with the high success rates of deep learning based protein design methods, our approach makes the design of proteins where global shape confers function, such as beta barrel nanopores, more precisely specifiable and accessible.

Target-conditioned diffusion generates potent TNFR superfamily antagonists and agonists

Despite progress in designing protein-binding proteins, the shape matching of designs to targets is lower than in many native protein complexes, and design efforts have failed for the tumor necrosis factor receptor 1 (TNFR1) and other protein targets with relatively flat and polar surfaces. We hypothesized that free diffusion from random noise could generate shape-matched binders for challenging targets and tested this approach on TNFR1. We obtain designs with low picomolar affinity whose specificity can be completely switched to other family members using partial diffusion. Designs function as antagonists or as superagonists when presented at higher valency for OX40 and 4-1BB. The ability to design high-affinity and high-specificity antagonists and agonists for pharmacologically important targets in silico presages a coming era in protein design in which binders are made by computation rather than immunization or random screening approaches.

Optimisation strategies for directed evolution without sequencing

Directed evolution can enable engineering of biological systems with minimal knowledge of their underlying sequence-to-function relationships. A typical directed evolution process consists of iterative rounds of mutagenesis and selection that are designed to steer changes in a biological system (e.g. a protein) towards some functional goal. Much work has been done, particularly leveraging advancements in machine learning, to optimise the process of directed evolution. Many of these methods, however, require DNA sequencing and synthesis, making them resource-intensive and incompatible with developments in targeted in vivo mutagenesis. Operating within the experimental constraints of established sorting-based directed evolution techniques (e.g. Fluorescence-Activated Cell Sorting, FACS), we explore approaches for optimisation of directed evolution that could in future be implemented without sequencing information. We then expand our methods to the context of emerging experimental techniques in directed evolution, which allow for single-cell selection based on fitness objectives defined from any combination of measurable traits. Finally, we explore these alternative strategies on the GB1 and TrpB empirical landscapes, demonstrating that they could lead to up to 19-fold and 7-fold increases respectively in the probability of attaining the global fitness peak.

Design of high specificity binders for peptide-MHC-I complexes

Class I MHC molecules present peptides derived from intracellular antigens on the cell surface for immune surveillance, and specific targeting of these peptide-MHC (pMHC) complexes could have considerable utility for treating diseases. Such targeting is challenging as it requires readout of the few outward facing peptide antigen residues and the avoidance of extensive contacts with the MHC carrier which is present on almost all cells. Here we describe the use of deep learning-based protein design tools to denovo design small proteins that arc above the peptide binding groove of pMHC complexes and make extensive contacts with the peptide. We identify specific binders for ten target pMHCs which when displayed on yeast bind the on-target pMHC tetramer but not closely related peptides. For five targets, incorporation of designs into chimeric antigen receptors leads to T-cell activation by the cognate pMHC complexes well above the background from complexes with peptides derived from proteome. Our approach can generate high specificity binders starting from either experimental or predicted structures of the target pMHC complexes, and should be widely useful for both protein and cell based pMHC targeting.

Design of facilitated dissociation enables control over cytokine signaling duration

Protein design has focused primarily on the design of ground states, ensuring they are sufficiently low energy to be highly populated1. Designing the kinetics and dynamics of a system requires, in addition, the design of excited states that are traversed in transitions from one low-lying state to another2,3. This is a challenging task as such states must be sufficiently strained to be poorly populated, but not so strained that they are not populated at all, and because protein design methods have generally focused on creating near-ideal structures4-7. Here we describe a general approach for designing systems which use an induced-fit power stroke8 to generate a structurally frustrated9 and strained excited state, allosterically driving protein complex dissociation. X-ray crystallography, double electron-electron resonance spectroscopy, and kinetic binding measurements demonstrate that incorporating excited states enables design of effector-induced increases in dissociation rates as high as 6000-fold. We highlight the power of this approach by designing cytokine mimics which can be dissociated within seconds from their receptors.

Synthetic promoter design in Escherichia coli based on multinomial diffusion model

Generative design of promoters has enhanced the efficiency of de novo creation of functional sequences. Though several deep generative models have been employed in biological sequence generation, including variational autoencoder (VAE) or Wasserstein generative adversarial network (WGAN), these models might struggle with mode collapse and low sample diversity. In this study, we introduce the multinomial diffusion model (MDM) for promoter sequence design and propose a structured set of criteria for effectively comparing the performance of generative models. In silico experiments demonstrate that MDM outperforms existing generative AI approaches. MDM demonstrates superior performance in various computational evaluations, remains robust during the training process, and exhibits a strong ability in capturing weak signals. In addition, we experimentally validated that the majority of our model designed promoters have expression activities in vivo, indicating the practicality and potential of MDM for bioengineering.

Natural Language Prompts Guide the Design of Novel Functional Protein Sequences

The advent of natural language interaction with machines has ushered in new innovations in text-guided generation of images, audio, video, and more. In this arena, we introduce Bio logical M ulti- M odal M odel ( BioM3 ), as a novel framework for designing functional proteins via natural language prompts. This framework integrates natural language with protein design through a three-stage process: aligning protein and text representations in a joint embedding space learned using contrastive learning, refinement of the text embeddings, and conditional generation of protein sequences via a discrete autoregressive diffusion model. BioM3 synthe-sizes protein sequences with detailed descriptions of the protein structure, lineage, and function from text annotations to enable the conditional generation of novel sequences with desired attributes through natural language prompts. We present in silico validation of the model predictions for subcellular localization prediction, reaction classification, remote homology detection, scaffold in-painting, and structural plausibility, and in vivo and in vitro experimental tests of natural language prompt-designed synthetic analogs of Src-homology 3 (SH3) domain proteins that mediate signaling in the Sho1 osmotic stress response pathway in baker's yeast. BioM3 possesses state-of-the-art performance in zero-shot prediction and homology detection tasks, and generates proteins with native-like tertiary folds and wild-type levels of experimentally assayed function.

Designed miniproteins potently inhibit and protect against MERS-CoV

Middle-East respiratory syndrome coronavirus (MERS-CoV) is a zoonotic pathogen with 36% case-fatality rate in humans. No vaccines or specific therapeutics are currently approved to use in humans or the camel host reservoir. Here, we computationally designed monomeric and homo-oligomeric miniproteins binding with high affinity to the MERS-CoV spike (S) glycoprotein, the main target of neutralizing antibodies and vaccine development. We show that these miniproteins broadly neutralize a panel of MERS-CoV S variants, spanning the known antigenic diversity of this pathogen, by targeting a conserved site in the receptor-binding domain (RBD). The miniproteins directly compete with binding of the DPP4 receptor to MERS-CoV S, thereby blocking viral attachment to the host entry receptor and subsequent membrane fusion. Intranasal administration of a lead miniprotein provides prophylactic protection against stringent MERS-CoV challenge in mice motivating future clinical development as a next-generation countermeasure against this virus with pandemic potential.

The Nobel Prize in Chemistry: past, present, and future of AI in biology

A Comment on the transformative progress of artificial intelligence for structural and protein biology, referencing the 2024 Nobel Prize in Chemistry.

Scalable protein design using optimization in a relaxed sequence space

Machine learning (ML)-based design approaches have advanced the field of de novo protein design, with diffusion-based generative methods increasingly dominating protein design pipelines. Here, we report a "hallucination"-based protein design approach that functions in relaxed sequence space, enabling the efficient design of high-quality protein backbones over multiple scales and with broad scope of application without the need for any form of retraining. We experimentally produced and characterized more than 100 proteins. Three high-resolution crystal structures and two cryo-electron microscopy density maps of designed single-chain proteins comprising up to 1000 amino acids validate the accuracy of the method. Our pipeline can also be used to design synthetic protein-protein interactions, as validated experimentally by a set of protein heterodimers. Relaxed sequence optimization offers attractive performance with respect to designability, scope of applicability for different design problems, and scalability across protein sizes.

Bond-centric modular design of protein assemblies

We describe a modular bond-centric approach to protein nanomaterial design inspired by the rich diversity of chemical structures that can be generated from the small number of atomic valencies and bonding interactions. We design protein building blocks with regular coordination geometries and bonding interactions that enable the assembly of a wide variety of closed and opened nanomaterials using simple geometrical principles. Experimental characterization confirms successful formation of more than twenty multi-component polyhedral protein cages, 2D arrays, and 3D protein lattices, with a high (10-50 %) success rate and electron microscopy data closely matching the corresponding design models. Because of the modularity, individual building blocks can assemble with different partners to generate distinct regular assemblies, resulting in an economy of parts and enabling the construction of reconfigurable systems.

Connecting the dots: Computational network analysis for disease insight and drug repurposing

Network biology is a powerful framework for studying the structure, function, and dynamics of biological systems, offering insights into the balance between health and disease states. The field is seeing rapid progress in all of its aspects: data availability, network synthesis, network analytics, and impactful applications in medicine and drug development. We review the most recent and significant results in network biomedicine, with a focus on the latest data, analytics, software resources, and applications in medicine. We also discuss what in our view are the likely directions of impactful development over the next few years.

Multistate and functional protein design using RoseTTAFold sequence space diffusion

Protein denoising diffusion probabilistic models are used for the de novo generation of protein backbones but are limited in their ability to guide generation of proteins with sequence-specific attributes and functional properties. To overcome this limitation, we developed ProteinGenerator (PG), a sequence space diffusion model based on RoseTTAFold that simultaneously generates protein sequences and structures. Beginning from a noised sequence representation, PG generates sequence and structure pairs by iterative denoising, guided by desired sequence and structural protein attributes. We designed thermostable proteins with varying amino acid compositions and internal sequence repeats and cage bioactive peptides, such as melittin. By averaging sequence logits between diffusion trajectories with distinct structural constraints, we designed multistate parent-child protein triples in which the same sequence folds to different supersecondary structures when intact in the parent versus split into two child domains. PG design trajectories can be guided by experimental sequence-activity data, providing a general approach for integrated computational and experimental optimization of protein function.

AI in cellular engineering and reprogramming

During the last decade, artificial intelligence (AI) has increasingly been applied in biophysics and related fields, including cellular engineering and reprogramming, offering novel approaches to understand, manipulate, and control cellular function. The potential of AI lies in its ability to analyze complex datasets and generate predictive models. AI algorithms can process large amounts of data from single-cell genomics and multiomic technologies, allowing researchers to gain mechanistic insights into the control of cell identity and function. By integrating and interpreting these complex datasets, AI can help identify key molecular events and regulatory pathways involved in cellular reprogramming. This knowledge can inform the design of precision engineering strategies, such as the development of new transcription factor and signaling molecule cocktails, to manipulate cell identity and drive authentic cell fate across lineage boundaries. Furthermore, when used in combination with computational methods, AI can accelerate and improve the analysis and understanding of the intricate relationships between genes, proteins, and cellular processes. In this review article, we explore the current state of AI applications in biophysics with a specific focus on cellular engineering and reprogramming. Then, we showcase a couple of recent applications where we combined machine learning with experimental and computational techniques. Finally, we briefly discuss the challenges and prospects of AI in cellular engineering and reprogramming, emphasizing the potential of these technologies to revolutionize our ability to engineer cells for a variety of applications, from disease modeling and drug discovery to regenerative medicine and biomanufacturing.

The JAK-STAT pathway: from structural biology to cytokine engineering

The Janus kinase-signal transducer and activator of transcription (JAK-STAT) pathway serves as a paradigm for signal transduction from the extracellular environment to the nucleus. It plays a pivotal role in physiological functions, such as hematopoiesis, immune balance, tissue homeostasis, and surveillance against tumors. Dysregulation of this pathway may lead to various disease conditions such as immune deficiencies, autoimmune diseases, hematologic disorders, and cancer. Due to its critical role in maintaining human health and involvement in disease, extensive studies have been conducted on this pathway, ranging from basic research to medical applications. Advances in the structural biology of this pathway have enabled us to gain insights into how the signaling cascade operates at the molecular level, laying the groundwork for therapeutic development targeting this pathway. Various strategies have been developed to restore its normal function, with promising therapeutic potential. Enhanced comprehension of these molecular mechanisms, combined with advances in protein engineering methodologies, has allowed us to engineer cytokines with tailored properties for targeted therapeutic applications, thereby enhancing their efficiency and safety. In this review, we outline the structural basis that governs key nodes in this pathway, offering a comprehensive overview of the signal transduction process. Furthermore, we explore recent advances in cytokine engineering for therapeutic development in this pathway.

AlphaFold two years on: Validation and impact

Two years on from the initial release of AlphaFold, we have seen its widespread adoption as a structure prediction tool. Here, we discuss some of the latest work based on AlphaFold, with a particular focus on its use within the structural biology community. This encompasses use cases like speeding up structure determination itself, enabling new computational studies, and building new tools and workflows. We also look at the ongoing validation of AlphaFold, as its predictions continue to be compared against large numbers of experimental structures to further delineate the model's capabilities and limitations.

Opportunities and challenges in design and optimization of protein function

The field of protein design has made remarkable progress over the past decade. Historically, the low reliability of purely structure-based design methods limited their application, but recent strategies that combine structure-based and sequence-based calculations, as well as machine learning tools, have dramatically improved protein engineering and design. In this Review, we discuss how these methods have enabled the design of increasingly complex structures and therapeutically relevant activities. Additionally, protein optimization methods have improved the stability and activity of complex eukaryotic proteins. Thanks to their increased reliability, computational design methods have been applied to improve therapeutics and enzymes for green chemistry and have generated vaccine antigens, antivirals and drug-delivery nano-vehicles. Moreover, the high success of design methods reflects an increased understanding of basic rules that govern the relationships among protein sequence, structure and function. However, de novo design is still limited mostly to α-helix bundles, restricting its potential to generate sophisticated enzymes and diverse protein and small-molecule binders. Designing complex protein structures is a challenging but necessary next step if we are to realize our objective of generating new-to-nature activities.

Design of amyloidogenic peptide traps

Segments of proteins with high β-strand propensity can self-associate to form amyloid fibrils implicated in many diseases. We describe a general approach to bind such segments in β-strand and β-hairpin conformations using de novo designed scaffolds that contain deep peptide-binding clefts. The designs bind their cognate peptides in vitro with nanomolar affinities. The crystal structure of a designed protein-peptide complex is close to the design model, and NMR characterization reveals how the peptide-binding cleft is protected in the apo state. We use the approach to design binders to the amyloid-forming proteins transthyretin, tau, serum amyloid A1 and amyloid β1-42 (Aβ42). The Aβ binders block the assembly of Aβ fibrils as effectively as the most potent of the clinically tested antibodies to date and protect cells from toxic Aβ42 species.

Context-aware geometric deep learning for protein sequence design

Protein design and engineering are evolving at an unprecedented pace leveraging the advances in deep learning. Current models nonetheless cannot natively consider non-protein entities within the design process. Here, we introduce a deep learning approach based solely on a geometric transformer of atomic coordinates and element names that predicts protein sequences from backbone scaffolds aware of the restraints imposed by diverse molecular environments. To validate the method, we show that it can produce highly thermostable, catalytically active enzymes with high success rates. This concept is anticipated to improve the versatility of protein design pipelines for crafting desired functions.

Noncanonical Amino Acids in Biocatalysis

In recent years, powerful genetic code reprogramming methods have emerged that allow new functional components to be embedded into proteins as noncanonical amino acid (ncAA) side chains. In this review, we will illustrate how the availability of an expanded set of amino acid building blocks has opened a wealth of new opportunities in enzymology and biocatalysis research. Genetic code reprogramming has provided new insights into enzyme mechanisms by allowing introduction of new spectroscopic probes and the targeted replacement of individual atoms or functional groups. NcAAs have also been used to develop engineered biocatalysts with improved activity, selectivity, and stability, as well as enzymes with artificial regulatory elements that are responsive to external stimuli. Perhaps most ambitiously, the combination of genetic code reprogramming and laboratory evolution has given rise to new classes of enzymes that use ncAAs as key catalytic elements. With the framework for developing ncAA-containing biocatalysts now firmly established, we are optimistic that genetic code reprogramming will become a progressively more powerful tool in the armory of enzyme designers and engineers in the coming years.

Modulation of FGF pathway signaling and vascular differentiation using designed oligomeric assemblies

Many growth factors and cytokines signal by binding to the extracellular domains of their receptors and driving association and transphosphorylation of the receptor intracellular tyrosine kinase domains, initiating downstream signaling cascades. To enable systematic exploration of how receptor valency and geometry affect signaling outcomes, we designed cyclic homo-oligomers with up to 8 subunits using repeat protein building blocks that can be modularly extended. By incorporating a de novo-designed fibroblast growth factor receptor (FGFR)-binding module into these scaffolds, we generated a series of synthetic signaling ligands that exhibit potent valency- and geometry-dependent Ca2+ release and mitogen-activated protein kinase (MAPK) pathway activation. The high specificity of the designed agonists reveals distinct roles for two FGFR splice variants in driving arterial endothelium and perivascular cell fates during early vascular development. Our designed modular assemblies should be broadly useful for unraveling the complexities of signaling in key developmental transitions and for developing future therapeutic applications.

De Novo Design of Integrin α5β1 Modulating Proteins for Regenerative Medicine

Integrin α5β1 is crucial for cell attachment and migration in development and tissue regeneration, and α5β1 binding proteins could have considerable utility in regenerative medicine and next-generation therapeutics. We use computational protein design to create de novo α5β1-specific modulating miniprotein binders, called NeoNectins, that bind to and stabilize the open state of α5β1. When immobilized onto titanium surfaces and throughout 3D hydrogels, the NeoNectins outperform native fibronectin and RGD peptide in enhancing cell attachment and spreading, and NeoNectin-grafted titanium implants outperformed fibronectin and RGD-grafted implants in animal models in promoting tissue integration and bone growth. NeoNectins should be broadly applicable for tissue engineering and biomedicine.

Computational design of soluble and functional membrane protein analogues

De novo design of complex protein folds using solely computational means remains a substantial challenge1. Here we use a robust deep learning pipeline to design complex folds and soluble analogues of integral membrane proteins. Unique membrane topologies, such as those from G-protein-coupled receptors2, are not found in the soluble proteome, and we demonstrate that their structural features can be recapitulated in solution. Biophysical analyses demonstrate the high thermal stability of the designs, and experimental structures show remarkable design accuracy. The soluble analogues were functionalized with native structural motifs, as a proof of concept for bringing membrane protein functions to the soluble proteome, potentially enabling new approaches in drug discovery. In summary, we have designed complex protein topologies and enriched them with functionalities from membrane proteins, with high experimental success rates, leading to a de facto expansion of the functional soluble fold space.

Accelerated Engineering of ELP-Based Materials through Hybrid Biomimetic-De Novo Predictive Molecular Design

Efforts to engineer high-performance protein-based materials inspired by nature have mostly focused on altering naturally occurring sequences to confer the desired functionalities, whereas de novo design lags significantly behind and calls for unconventional innovative approaches. Here, using partially disordered elastin-like polypeptides (ELPs) as initial building blocks this work shows that de novo engineering of protein materials can be accelerated through hybrid biomimetic design, which this work achieves by integrating computational modeling, deep neural network, and recombinant DNA technology. This generalizable approach involves incorporating a series of de novo-designed sequences with α-helical conformation and genetically encoding them into biologically inspired intrinsically disordered repeating motifs. The new ELP variants maintain structural conformation and showed tunable supramolecular self-assembly out of thermal equilibrium with phase behavior in vitro. This work illustrates the effective translation of the predicted molecular designs in structural and functional materials. The proposed methodology can be applied to a broad range of partially disordered biomacromolecules and potentially pave the way toward the discovery of novel structural proteins.

Validation of de novo designed water-soluble and transmembrane β-barrels by in silico folding and melting

In silico validation of de novo designed proteins with deep learning (DL)-based structure prediction algorithms has become mainstream. However, formal evidence of the relationship between a high-quality predicted model and the chance of experimental success is lacking. We used experimentally characterized de novo water-soluble and transmembrane β-barrel designs to show that AlphaFold2 and ESMFold excel at different tasks. ESMFold can efficiently identify designs generated based on high-quality (designable) backbones. However, only AlphaFold2 can predict which sequences have the best chance of experimentally folding among similar designs. We show that ESMFold can generate high-quality structures from just a few predicted contacts and introduce a new approach based on incremental perturbation of the prediction ("in silico melting"), which can reveal differences in the presence of favorable contacts between designs. This study provides a new insight on DL-based structure prediction models explainability and on how they could be leveraged for the design of increasingly complex proteins; in particular membrane proteins which have historically lacked basic in silico validation tools.

Protein Design Using Structure-Prediction Networks: AlphaFold and RoseTTAFold as Protein Structure Foundation Models

Designing proteins with tailored structures and functions is a long-standing goal in bioengineering. Recently, deep learning advances have enabled protein structure prediction at near-experimental accuracy, which has catalyzed progress in protein design as well. We review recent studies that use structure-prediction neural networks to design proteins, via approaches such as activation maximization, inpainting, or denoising diffusion. These methods have led to major improvements over previous methods in wet-lab success rates for designing protein binders, metalloproteins, enzymes, and oligomeric assemblies. These results show that structure-prediction models are a powerful foundation for developing protein-design tools and suggest that continued improvement of their accuracy and generality will be key to unlocking the full potential of protein design.

An integrative approach to protein sequence design through multiobjective optimization

With recent methodological advances in the field of computational protein design, in particular those based on deep learning, there is an increasing need for frameworks that allow for coherent, direct integration of different models and objective functions into the generative design process. Here we demonstrate how evolutionary multiobjective optimization techniques can be adapted to provide such an approach. With the established Non-dominated Sorting Genetic Algorithm II (NSGA-II) as the optimization framework, we use AlphaFold2 and ProteinMPNN confidence metrics to define the objective space, and a mutation operator composed of ESM-1v and ProteinMPNN to rank and then redesign the least favorable positions. Using the two-state design problem of the foldswitching protein RfaH as an in-depth case study, and PapD and calmodulin as examples of higher-dimensional design problems, we show that the evolutionary multiobjective optimization approach leads to significant reduction in the bias and variance in RfaH native sequence recovery, compared to a direct application of ProteinMPNN. We suggest that this improvement is due to three factors: (i) the use of an informative mutation operator that accelerates the sequence space exploration, (ii) the parallel, iterative design process inherent to the genetic algorithm that improves upon the ProteinMPNN autoregressive sequence decoding scheme, and (iii) the explicit approximation of the Pareto front that leads to optimal design candidates representing diverse tradeoff conditions. We anticipate this approach to be readily adaptable to different models and broadly relevant for protein design tasks with complex specifications.

De novo design of alpha-beta repeat proteins

Proteins composed of a single structural unit tandemly repeated multiple times carry out a wide range of functions in biology. There has hence been considerable interest in designing such repeat proteins; previous approaches have employed strict constraints on secondary structure types and relative geometries, and most characterized designs either mimic a known natural topology, adhere closely to a parametric helical bundle architecture, or exploit very short repetitive sequences. Here, we describe Rosetta-based and deep learning hallucination methods for generating novel repeat protein architectures featuring mixed alpha-helix and beta-strand topologies, and 25 new highly stable alpha-beta proteins designed using these methods. We find that incorporation of terminal caps which prevent beta strand mediated intermolecular interactions increases the solubility and monomericity of individual designs as well as overall design success rate.

A suite of designed protein cages using machine learning and protein fragment-based protocols

Designed protein cages and related materials provide unique opportunities for applications in biotechnology and medicine, but their creation remains challenging. Here, we apply computational approaches to design a suite of tetrahedrally symmetric, self-assembling protein cages. For the generation of docked conformations, we emphasize a protein fragment-based approach, while for sequence design of the de novo interface, a comparison of knowledge-based and machine learning protocols highlights the power and increased experimental success achieved using ProteinMPNN. An analysis of design outcomes provides insights for improving interface design protocols, including prioritizing fragment-based motifs, balancing interface hydrophobicity and polarity, and identifying preferred polar contact patterns. In all, we report five structures for seven protein cages, along with two structures of intermediate assemblies, with the highest resolution reaching 2.0 Å using cryo-EM. This set of designed cages adds substantially to the body of available protein nanoparticles, and to methodologies for their creation.

Diversifying de novo TIM barrels by hallucination

De novo protein design expands the protein universe by creating new sequences to accomplish tailor-made enzymes in the future. A promising topology to implement diverse enzyme functions is the ubiquitous TIM-barrel fold. Since the initial de novo design of an idealized four-fold symmetric TIM barrel, the family of de novo TIM barrels is expanding rapidly. Despite this and in contrast to natural TIM barrels, these novel proteins lack cavities and structural elements essential for the incorporation of binding sites or enzymatic functions. In this work, we diversified a de novo TIM barrel by extending multiple βα-loops using constrained hallucination. Experimentally tested designs were found to be soluble upon expression in Escherichia coli and well-behaved. Biochemical characterization and crystal structures revealed successful extensions with defined α-helical structures. These diversified de novo TIM barrels provide a framework to explore a broad spectrum of functions based on the potential of natural TIM barrels.

Concentration Recognition-Based Auto-Dynamic Regulation System (CRUISE) Enabling Efficient Production of Higher Alcohols

Microbial factories lacking the ability of dynamically regulating the pathway enzymes overexpression, according to in situ metabolite concentrations, are suboptimal, especially when the metabolic intermediates are competed by growth and chemical production. The production of higher alcohols (HAs), which hijacks the amino acids (AAs) from protein biosynthesis, minimizes the intracellular concentration of AAs and thus inhibits the host growth. To balance the resource allocation and maintain stable AA flux, this work utilizes AA-responsive transcriptional attenuator ivbL and HA-responsive transcriptional activator BmoR to establish a concentration recognition-based auto-dynamic regulation system (CRUISE). This system ultimately maintains the intracellular homeostasis of AA and maximizes the production of HA. It is demonstrated that ivbL-driven enzymes overexpression can dynamically regulate the AA-to-HA conversion while BmoR-driven enzymes overexpression can accelerate the AA biosynthesis during the HA production in a feedback activation mode. The AA flux in biosynthesis and conversion pathways is balanced via the intracellular AA concentration, which is vice versa stabilized by the competition between AA biosynthesis and conversion. The CRUISE, further aided by scaffold-based self-assembly, enables 40.4 g L-1 of isobutanol production in a bioreactor. Taken together, CRUISE realizes robust HA production and sheds new light on the dynamic flux control during the process of chemical production.

Generative artificial intelligence for de novo protein design

Engineering new molecules with desirable functions and properties has the potential to extend our ability to engineer proteins beyond what nature has so far evolved. Advances in the so-called 'de novo' design problem have recently been brought forward by developments in artificial intelligence. Generative architectures, such as language models and diffusion processes, seem adept at generating novel, yet realistic proteins that display desirable properties and perform specified functions. State-of-the-art design protocols now achieve experimental success rates nearing 20%, thus widening the access to de novo designed proteins. Despite extensive progress, there are clear field-wide challenges, for example, in determining the best in silico metrics to prioritise designs for experimental testing, and in designing proteins that can undergo large conformational changes or be regulated by post-translational modifications. With an increase in the number of models being developed, this review provides a framework to understand how these tools fit into the overall process of de novo protein design. Throughout, we highlight the power of incorporating biochemical knowledge to improve performance and interpretability.

Generalized biomolecular modeling and design with RoseTTAFold All-Atom

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.

Designing Multifunctional Biomaterials via Protein Self-Assembly

Protein self-assembly is a fundamental biological process where proteins spontaneously organize into complex and functional structures without external direction. This process is crucial for the formation of various biological functionalities. However, when protein self-assembly fails, it can trigger the development of multiple disorders, thus making understanding this phenomenon extremely important. Up until recently, protein self-assembly has been solely linked either to biological function or malfunction; however, in the past decade or two it has also been found to hold promising potential as an alternative route for fabricating materials for biomedical applications. It is therefore necessary and timely to summarize the key aspects of protein self-assembly: how the protein structure and self-assembly conditions (chemical environments, kinetics, and the physicochemical characteristics of protein complexes) can be utilized to design biomaterials. This minireview focuses on the basic concepts of forming supramolecular structures, and the existing routes for modifications. We then compare the applicability of different approaches, including compartmentalization and self-assembly monitoring. Finally, based on the cutting-edge progress made during the last years, we summarize the current knowledge about tailoring a final function by introducing changes in self-assembly and link it to biomaterials' performance.

Reengineering of a flavin-binding fluorescent protein using ProteinMPNN

Recent advances in machine learning techniques have led to development of a number of protein design and engineering approaches. One of them, ProteinMPNN, predicts an amino acid sequence that would fold and match user-defined backbone structure. Its performance was previously tested for proteins composed of standard amino acids, as well as for peptide- and protein-binding proteins. In this short report, we test whether ProteinMPNN can be used to reengineer a non-proteinaceous ligand-binding protein, flavin-based fluorescent protein CagFbFP. We fixed the native backbone conformation and the identity of 20 amino acids interacting with the chromophore (flavin mononucleotide, FMN) while letting ProteinMPNN predict the rest of the sequence. The software package suggested replacing 36-48 out of the remaining 86 amino acids so that the resulting sequences are 55%-66% identical to the original one. The three designs that we tested experimentally displayed different expression levels, yet all were able to bind FMN and displayed fluorescence, thermal stability, and other properties similar to those of CagFbFP. Our results demonstrate that ProteinMPNN can be used to generate diverging unnatural variants of fluorescent proteins, and, more generally, to reengineer proteins without losing their ligand-binding capabilities.

Rapid and automated design of two-component protein nanomaterials using ProteinMPNN

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.

An integrative approach to protein sequence design through multiobjective optimization

With recent methodological advances in the field of computational protein design, in particular those based on deep learning, there is an increasing need for frameworks that allow for coherent, direct integration of different models and objective functions into the generative design process. Here we demonstrate how evolutionary multiobjective optimization techniques can be adapted to provide such an approach. With the established Non-dominated Sorting Genetic Algorithm II (NSGA-II) as the optimization framework, we use AlphaFold2 and ProteinMPNN confidence metrics to define the objective space, and a mutation operator composed of ESM-1v and ProteinMPNN to rank and then redesign the least favorable positions. Using the multistate design problem of the foldswitching protein RfaH as an in-depth case study, we show that the evolutionary multiobjective optimization approach leads to significant reduction in the bias and variance in RfaH native sequence recovery, compared to a direct application of ProteinMPNN. We suggest that this improvement is due to three factors: (i) the use of an informative mutation operator that accelerates the sequence space exploration, (ii) the parallel, iterative design process inherent to the genetic algorithm that improves upon the ProteinMPNN autoregressive sequence decoding scheme, and (iii) the explicit approximation of the Pareto front that leads to optimal design candidates representing diverse tradeoff conditions. We anticipate this approach to be readily adaptable to different models and broadly relevant for protein design tasks with complex specifications.

Blueprinting extendable nanomaterials with standardized protein blocks

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 structures1. 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.