# **Enzymes Designed by AI: Opening New Avenues in Biocatalysis**
Enzymes are the catalysts provided by nature, facilitating vital biochemical processes with extraordinary efficiency and precision. These proteins, made up of common elements, assist in tasks from energy transformation to molecular formation. However, natural enzymes, despite their adaptability, have drawbacks—they are not available for every reaction that scientists aim to catalyze, such as decomposing plastics or effectively capturing carbon dioxide.
Recent breakthroughs in artificial intelligence (AI)-enabled protein design are transforming enzyme engineering, allowing for the development of novel enzymes with capabilities that exceed those present in nature. A recent research study showcases a significant advancement in creating an AI-generated enzyme that can break down plastic, highlighting both the possibilities and challenges of this innovative technique.
—
## **The Difficulty of Enzyme Design**
The mechanisms of enzymes are intrinsically intricate. Even reactions that appear straightforward often encompass multiple stages, demanding exact atomic configurations and intermediate forms. Conventional methods of enzyme engineering, like directed evolution, have shown some success in adjusting existing enzymes for different functions. Nevertheless, crafting entirely new enzymes from the ground up has proven to be a daunting challenge.
One of the most desired enzymatic capabilities is the ability to degrade ester bonds, which are prevalent in both natural molecules and synthetic materials like plastics. Polyester, for instance, gets its name from the multitude of ester bonds present in its structure. While some ester-degrading enzymes can be found in nature, they often lack efficiency for industrial use, necessitating the creation of more potent alternatives.
—
## **Enzyme Engineering Powered by AI**
To address this challenge, scientists utilized AI-centric protein design tools to generate a new enzyme proficient at breaking ester bonds. The approach involved several AI models, each playing a role in different facets of enzyme architecture and functionality:
1. **RFDiffusion** – This AI application produced varied protein structures based on the average positions of amino acids found in known ester-cleaving enzymes.
2. **PLACER** – A generative AI model trained on established protein structures, PLACER refined enzyme designs to guarantee they could achieve the required conformations for catalytic activity.
Initially, from 129 engineered proteins, only two showed catalytic capability. Acknowledging the necessity for additional refinement, scientists integrated PLACER to boost enzyme adaptability, resulting in a threefold enhancement in functional enzymes.
—
## **Surmounting Functional Challenges**
Despite these advancements, the AI-crafted enzymes initially faltered after a single reaction. Instead of functioning as genuine catalysts, they became chemically bonded to reaction intermediates, leaving them ineffective. To resolve this, researchers employed PLACER to evaluate structures that could assume critical intermediate states, eventually identifying two enzymes—referred to as “super” and “win”—that could perform multiple reaction cycles.
By persistently refining enzyme structures through AI tools, the team successfully engineered an enzyme with activity on par with naturally occurring enzymes. Moreover, they showcased the capability to develop an esterase that could degrade PET, a commonly utilized plastic.
—
## **Consequences and Future Pathways**
This breakthrough highlights the potential of AI in enzyme design, presenting a promising method for developing biocatalysts for uses such as plastic degradation, carbon sequestration, and sustainable chemistry. Nonetheless, the procedure remains complicated, necessitating extensive computational and experimental verification.
Looking forward, scientists might investigate the incorporation of AI-designed enzymes into living cells, enabling natural evolution to further improve their functionality. By merging AI-driven design with biological evolution, researchers could unlock even more effective and versatile enzymes, setting the stage for sustainable approaches to urgent environmental and industrial issues.
As AI technologies progress, the aspiration of designing custom enzymes for nearly any reaction may soon turn into reality, revolutionizing sectors from biotechnology to materials science.
—
**Reference:**
*Science*, 2024. DOI: [10.1126/science.adu2454](http://dx.doi.org/10.1126/science.adu2454)
