Anna Lauko’s research while affiliated with University of Washington and other places

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 to design enzymes starting from minimal active site descriptions. Experimental characterization revealed catalytic efficiencies ( k cat / K m ) up to 2.2x10 ⁵ M ⁻¹ s ⁻¹ and crystal structures that closely match the design models (Cα RMSDs < 1 Å). Selection for structural compatibility across the reaction coordinate enabled identification of new catalysts in low-throughput screens 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.

New enzymes can be designed by starting from a description of an ideal active site composed of catalytic residues surrounding the reaction transition state(s) and identifying or generating a protein scaffold that supports the site, but there are a few current limitations. First, the catalytic efficiencies achieved by such efforts have generally been quite low, and considerable optimization by directed evolution has been required to reach activities typical of native enzymes. Second, generative AI methods such as RFdiffusion now enable the direct generation of proteins around active sites, but to date, such scaffolding has required specification of both the position in the sequence and the backbone coordinates of the catalytic residue, which complicates sampling. Here we introduce a generative AI method called RFflow that overcomes these limitations and use it to design zinc metallohydrolases starting from a density functional theory description of active site geometry. Of 96 designs tested experimentally, the most active has a kcat/KM of 23,000 M ⁻¹ s ⁻¹ , orders of magnitude higher than previously designed metallohydrolases. This 148 amino acid protein has a novel fold with an enclosed chamber which positions the reaction substrate nearly perfectly for attack by a catalytic water molecule activated by the bound metal and is predicted by ChemNet to have a highly preorganized active site. The ability to generate high activity catalysts starting from quantum chemistry calculated active site geometries without experimental optimization should open the door to a new generation of potent designer enzymes.

Modeling the conformational heterogeneity of protein-small molecule systems is an outstanding challenge. We reasoned that while residue level descriptions of biomolecules are efficient for de novo structure prediction, for probing heterogeneity of interactions with small molecules in the folded state an entirely atomic level description could have advantages in speed and generality. We developed a graph neural network called ChemNet trained to recapitulate correct atomic positions from partially corrupted input structures from the Cambridge Structural Database and the Protein Data Bank; the nodes of the graph are the atoms in the system. ChemNet accurately generates structures of diverse organic small molecules given knowledge of their atom composition and bonding, and given a description of the larger protein context, and builds up structures of small molecules and protein side chains for protein-small molecule docking. Because ChemNet is rapid and stochastic, ensembles of predictions can be readily generated to map conformational heterogeneity. In enzyme design efforts described here and elsewhere, we find that using ChemNet to assess the accuracy and pre-organization of the designed active sites results in higher success rates and higher activities; we obtain a preorganized retroaldolase with a kcat/KM of 11000 M-1min-1, considerably higher than any pre-deep learning design for this reaction. We anticipate that ChemNet will be widely useful for rapidly generating conformational ensembles of small molecule and small molecule-protein systems, and for designing higher activity preorganized enzymes.

Enzymes that proceed through multistep reaction mechanisms often utilize complex, polar active sites positioned with sub-angstrom precision to mediate distinct chemical steps, which makes their de novo construction extremely challenging. We sought to overcome this challenge using the classic catalytic triad and oxyanion hole of serine hydrolases as a model system. We used RFdiffusion ¹ to generate proteins housing catalytic sites of increasing complexity and varying geometry, and a newly developed ensemble generation method called ChemNet to assess active site geometry and preorganization at each step of the reaction. Experimental characterization revealed novel serine hydrolases that catalyze ester hydrolysis with catalytic efficiencies ( k cat / K m ) up to 3.8 × 10 ³ M ⁻¹ s ⁻¹ , closely match the design models (Cα RMSDs < 1 Å), and have folds distinct from natural serine hydrolases. In silico selection of designs based on active site preorganization across the reaction coordinate considerably increased success rates, enabling identification of new catalysts in screens of as few as 20 designs. Our de novo buildup approach provides insight into the geometric determinants of catalysis that complements what can be obtained from structural and mutational studies of native enzymes (in which catalytic group geometry and active site makeup cannot be so systematically varied), and provides a roadmap for the design of industrially relevant serine hydrolases and, more generally, for designing complex enzymes that catalyze multi-step transformations.

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

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

We use computational design coupled with experimental characterization to systematically investigate the design principles for macrocycle membrane permeability and oral bioavailability. We designed 184 6–12 residue macrocycles with a wide range of predicted structures containing noncanonical backbone modifications and experimentally determined structures of 35; 29 are very close to the computational models. With such control, we show that membrane permeability can be systematically achieved by ensuring all amide (NH) groups are engaged in internal hydrogen bonding interactions. 84 designs over the 6–12 residue size range cross membranes with an apparent permeability greater than 1 × 10⁻⁶ cm/s. Designs with exposed NH groups can be made membrane permeable through the design of an alternative isoenergetic fully hydrogen-bonded state favored in the lipid membrane. The ability to robustly design membrane-permeable and orally bioavailable peptides with high structural accuracy should contribute to the next generation of designed macrocycle therapeutics.

Cryo-electron microscopy of protein complexes often leads to moderate resolution maps (4–8 Å), with visible secondary-structure elements but poorly resolved loops, making model building challenging. In the absence of high-resolution structures of homologues, only coarse-grained structural features are typically inferred from these maps, and it is often impossible to assign specific regions of density to individual protein subunits. This paper describes a new method for overcoming these difficulties that integrates predicted residue distance distributions from a deep-learned convolutional neural network, computational protein folding using Rosetta, and automated EM-map-guided complex assembly. We apply this method to a 4.6 Å resolution cryoEM map of Fanconi Anemia core complex (FAcc), an E3 ubiquitin ligase required for DNA interstrand crosslink repair, which was previously challenging to interpret as it comprises 6557 residues, only 1897 of which are covered by homology models. In the published model built from this map, only 387 residues could be assigned to the specific subunits with confidence. By building and placing into density 42 deep-learning-guided models containing 4795 residues not included in the previously published structure, we are able to determine an almost-complete atomic model of FAcc, in which 5182 of the 6557 residues were placed. The resulting model is consistent with previously published biochemical data, and facilitates interpretation of disease-related mutational data. We anticipate that our approach will be broadly useful for cryoEM structure determination of large complexes containing many subunits for which there are no homologues of known structure.

Cryo-electron microscopy of protein complexes often leads to moderate resolution maps (4-8 Å), with visible secondary structure elements but poorly resolved loops, making model-building challenging. In the absence of high-resolution structures of homologues, only coarse-grained structural features are typically inferred from these maps, and it is often impossible to assign specific regions of density to individual protein subunits. This paper describes a new method for overcoming these difficulties that integrates predicted residue distance distributions from a deep-learned convolutional neural network, computational protein folding using Rosetta, and automated EM-map-guided complex assembly. We apply this method to a 4.6 Å resolution cryoEM map of Fanconi Anemia core complex (FAcc), an E3 ubiquitin ligase required for DNA interstrand crosslink repair, which was previously challenging to interpret as it is comprised of 6557 residues, only 1897 of which are covered by homology models. In the published structure built from this map, only 387 residues could be assigned to specific subunits. By building and placing into density 42 deep-learning guided models containing 4795 residues not included in the previously published structure, we are able to determine an almost-complete atomic model of FAcc, in which 5182 of the 6557 residues were placed. The resulting model is consistent with previously published biochemical data, and facilitates interpretation of disease related mutational data. We anticipate that our approach will be broadly useful for cryoEM structure determination of large complexes containing many subunits for which there are no homologues of known structure.