The Baker Lab's latest research, published in Science, demonstrates a significant advancement in AI-driven enzyme design, focusing on creating serine hydrolases with complex active sites1. This breakthrough allows for the generation of efficient protein catalysts tailored to specific chemical reactions, particularly the breaking of ester bonds.

Key aspects of this research include:

• Over 300 computer-generated proteins were tested, with a subset showing reactivity with chemical probes.

• Iterative rounds of design and screening led to highly efficient catalysts, some surpassing prior computationally designed esterases.

• Crystal structures of the designed enzymes closely matched their computational models, deviating by less than 1 Å.

• The research integrates deep learning-based protein design with a novel assessment tool for evaluating catalytic preorganization across multiple reaction states.

This work builds on other recent AI breakthroughs in the lab, including the development of tools for modeling protein-small molecule conformational ensembles and creating retroaldolase enzymes with higher catalytic efficiencies1. The potential applications of this technology are vast, ranging from pharmaceutical compound synthesis to microplastic degradation, paving the way for a greener economy and more sustainable industrial processes.

The catalytic cycle of serine hydrolases can be divided into four distinct steps:

1. Substrate binding and initial attack: The substrate binds to the apoenzyme, and the catalytic serine, deprotonated by the catalytic histidine, attacks the carbonyl carbon of the ester. This forms the first tetrahedral intermediate (TI1).1

2. Acyl-enzyme intermediate formation: The catalytic histidine protonates the leaving group oxygen, promoting its departure. This leaves the active site serine covalently linked to the acyl group of the substrate, forming the acyl-enzyme intermediate (AEI).1

3. Water activation and second tetrahedral intermediate: The histidine deprotonates a water molecule, which then attacks the AEI to generate a second tetrahedral intermediate (TI2).1

4. Enzyme regeneration: The final step involves histidine-mediated protonation of serine and release of the acyl group. This reconstitutes the free enzyme, completing the catalytic cycle.1

Throughout this process, negatively charged transition states and intermediates are stabilized by a pair of hydrogen bond donors that constitute the oxyanion hole, ensuring efficient catalysis.1

The research methodology for designing AI-driven serine hydrolases involved a multi-step process combining advanced computational tools with experimental validation:

1. Initial design: Researchers used RFdiffusion, a generative AI framework, to design proteins with complex catalytic sites1.

2. Active site evaluation: A novel machine learning model called PLACER (Protein-Ligand Atomistic Conformational Ensemble Reproduction) was developed to predict and evaluate active site conformations of designed enzymes1.

3. Iterative refinement: The team initially created simplified versions with three of the five catalytic groups found in natural enzymes, then progressed to more complex designs incorporating all five groups1.

4. Experimental characterization: Designed enzymes were tested in vitro, with selected candidates further analyzed through crystal structure analysis to confirm their structural accuracy1.

5. Screening and optimization: Enzymes were screened for active site preorganization to identify designs with a higher likelihood of success in real-world applications1.

This methodology demonstrates the integration of AI tools with traditional experimental techniques, establishing a flexible approach for designing enzymes with tailored functions for synthetic biology applications1.

The groundbreaking research led by Nobel laureate David Baker at the University of Washington has significant implications for plastic breakdown, particularly for polyethylene terephthalate (PET). By using AI to design enzymes from scratch, the team has opened up new possibilities for creating tailored enzymes that can efficiently break down specific types of plastics1.

The AI-designed serine hydrolases, which are capable of breaking ester bonds, show promise for PET degradation2. PET, a common plastic used in packaging and textiles, contains ester bonds that could potentially be targeted by these custom-designed enzymes. Kiera Sumida, a co-lead author of the study, is already applying these methods to tackle plastic degradation, demonstrating the broad potential of this approach2.

• The ability to design enzymes with complex active sites could lead to more efficient PET-degrading enzymes.

• The high catalytic efficiency achieved in the study suggests potential for faster plastic breakdown rates.

• The flexibility of the AI-driven design process allows for rapid iteration and optimization of enzymes for specific plastic types.

This research represents a significant step towards developing more effective solutions for plastic waste management, potentially revolutionizing recycling processes and contributing to a more sustainable, circular economy3.