21/11/2025
𝐄𝐙𝐒𝐩𝐞𝐜𝐢𝐟𝐢𝐜𝐢𝐭𝐲: 𝐇𝐨𝐰 𝐀𝐈 𝐢𝐬 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐧𝐳𝐲𝐦𝐞 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡.
Enzymes are the workhorses of biochemistry, catalyzing reactions essential for life and industrial processes alike. However, identifying the perfect enzyme-substrate combination is a challenging task. Enzyme active sites are not static; they change shape upon substrate binding—a phenomenon called induced fit. Moreover, some enzymes are promiscuous, capable of catalyzing multiple reactions. These complexities make predicting enzyme-substrate compatibility difficult, and traditional experimental approaches are often slow and limited.
To address this challenge, researchers at the University of Illinois Urbana-Champaign, led by Professor Huimin Zhao, have developed EZSpecificity, an artificial intelligence (AI)-powered tool that predicts which substrates are most likely to fit a given enzyme. The tool combines a machine learning algorithm with a large dataset derived from both experimental and computational studies, making it far more accurate than previous models. The team collaborated with Professor Diwakar Shukla’s group, performing millions of docking simulations to examine how enzymes of different classes adapt their shapes around various substrates. By integrating enzyme sequence, structural data, and flexibility information, the AI model can now predict enzyme-substrate specificity with unprecedented accuracy.
EZSpecificity was tested against ESP, the leading enzyme specificity model, in four scenarios designed to simulate real-world applications. The results demonstrated a clear advantage for EZSpecificity, which achieved 91.7% accuracy for top substrate predictions, compared to 58.3% accuracy for ESP. Experimental validation with eight halogenase enzymes and 78 substrates further confirmed the model’s reliability. This validation highlights the tool’s potential not only for research but also for practical applications in medicine, biotechnology, and industrial catalysis.
The availability of EZSpecificity has important implications for the scientific community. Researchers can now input a substrate and an enzyme sequence into the user-friendly interface to predict the likelihood of a successful pairing. This capability accelerates the process of enzyme selection, facilitating the design of efficient catalysts for drug development, synthesis of bioactive molecules, and various manufacturing processes. Looking ahead, the researchers plan to expand EZSpecificity to analyze enzyme selectivity, predicting which specific sites on a substrate an enzyme prefers, and to refine the model further with more experimental data.
EZSpecificity represents a powerful integration of biochemistry and AI, demonstrating how computational tools can complement experimental research to solve complex biological problems. By combining large datasets, machine learning, and detailed simulations, the tool offers a faster and more reliable way to explore enzyme-substrate interactions. Its development not only advances our understanding of enzyme specificity but also opens new avenues for innovation across medicine, biotechnology, and industrial applications.
References :
Zhao, H.; Shukla, D. EZSpecificity: AI-powered prediction of enzyme-substrate specificity. Nature, 2025.
University of Illinois Urbana-Champaign. AI Tool Predicts Enzyme-Substrate Compatibility. 2025.
𝐵𝑆/2023/077
1𝑠𝑡 𝑦𝑒𝑎𝑟
𝐾𝑢𝑙𝑎𝑡ℎ𝑢𝑛𝑔𝑒 𝐾.𝐺.𝐻.𝐾.