AI-Powered AiCE Method Revolutionizes Protein Engineering

AI-powered strategy streamlines protein engineering by integrating structural and evolutionary constraints

A team of researchers has developed a method that could transform the field of protein engineering. The new approach, called AI-informed Constraints for protein Engineering (AiCE), enables rapid and efficient protein evolution by integrating structural and evolutionary constraints into a generic inverse folding model -- without the need to train specialized artificial intelligence (AI) models. The study, published in Cell on July 7, addresses many of the challenges associated with traditional protein engineering techniques. The team was led by Prof. Gao Caixia from the Institute of Genetics and Developmental Biology (IGDB) of the Chinese Academy of Sciences. The ideal protein engineering strategy would achieve optimal performance with minimal effort. However, existing approaches are often limited in terms of cost, efficiency, and scalability. Current AI-based protein engineering methods are often computationally intensive, underscoring the need for more accessible and user-friendly alternatives that preserve predictive accuracy and enable broader adoption across the research community. In this study, the researchers first developed AiCE, a module designed to predict high-fitness (HF) single amino acid substitutions. It enhances prediction accuracy by extensively sampling inverse folding models -- AI models that generate compatible amino acid sequences based on protein 3D structures -- while incorporating structural constraints. Benchmarking against 60 deep mutational scanning (DMS) datasets demonstrated that AiCE outperforms other AI-based methods by 36-90%. Its effectiveness for complex proteins and protein-nucleic acid complexes was also validated. Notably, incorporating structural constraints alone yielded a 37% improvement in accuracy. To address the challenge of negative epistatic interactions in combinatorial mutations, the researchers developed the AiCE module, which integrates evolutionary coupling constraints. This allows for accurate prediction of multiple high-fitness mutations at minimal computational cost, expanding the tool's versatility and practical utility. Using the AiCE framework, the researchers successfully evolved eight proteins with diverse structures and functions, including deaminases, nuclear localization sequences, nucleases, and reverse transcriptases. These engineered proteins have enabled the creation of several next-generation base editors for applications in precision medicine and molecular breeding. These include: enABE8e, a cytosine base editor with a ~50% narrower editing window; enSdd6-CBE, an adenine base editor with 1.3× higher fidelity; and enDdd1-DdCBE, a mitochondrial base editor showing a 13× increase in activity. AiCE represents a simple, efficient and broadly applicable strategy for protein engineering. By unlocking the potential of existing AI models, it offers a promising new direction for the field and enhances the interpretability of AI-driven protein redesign.

New AI-informed method accelerates protein engineering

Chinese Academy of SciencesJul 7 2025 A team of Chinese researchers led by Prof. GAO Caixia from the Institute of Genetics and Developmental Biology (IGDB) of the Chinese Academy of Sciences has developed a groundbreaking method that could transform the field of protein engineering. The new approach, called AI-informed Constraints for protein Engineering (AiCE), enables rapid and efficient protein evolution by integrating structural and evolutionary constraints into a generic inverse folding model-without the need to train specialized artificial intelligence (AI) models. The study, published in Cell on July 7, addresses many of the challenges associated with traditional protein engineering techniques. The ideal protein engineering strategy would achieve optimal performance with minimal effort. However, existing approaches are often limited in terms of cost, efficiency, and scalability. Current AI-based protein engineering methods are often computationally intensive, underscoring the need for more accessible and user-friendly alternatives that preserve predictive accuracy and enable broader adoption across the research community. In this study, the researchers first developed AiCEsingle, a module designed to predict high-fitness (HF) single amino acid substitutions. It enhances prediction accuracy by extensively sampling inverse folding models-AI models that generate compatible amino acid sequences based on protein 3D structures-while incorporating structural constraints. Benchmarking against 60 deep mutational scanning (DMS) datasets demonstrated that AiCEsingle outperforms other AI-based methods by 36-90%. Its effectiveness for complex proteins and protein-nucleic acid complexes was also validated. Notably, incorporating structural constraints alone yielded a 37% improvement in accuracy. To address the challenge of negative epistatic interactions in combinatorial mutations, the researchers developed the AiCEmulti module, which integrates evolutionary coupling constraints. This allows for accurate prediction of multiple high-fitness mutations at minimal computational cost, expanding the tool's versatility and practical utility. Using the AiCE framework, the researchers successfully evolved eight proteins with diverse structures and functions, including deaminases, nuclear localization sequences, nucleases, and reverse transcriptases. These engineered proteins have enabled the creation of several next-generation base editors for applications in precision medicine and molecular breeding. These include: enABE8e, a cytosine base editor with a ~50% narrower editing window; enSdd6-CBE, an adenine base editor with 1.3× higher fidelity; and enDdd1-DdCBE, a mitochondrial base editor showing a 13× increase in activity. AiCE represents a simple, efficient and broadly applicable strategy for protein engineering. By unlocking the potential of existing AI models, it offers a promising new direction for the field and enhances the interpretability of AI-driven protein redesign. Chinese Academy of Sciences Journal reference: Fei, H., et al. (2025). Advancing protein evolution with inverse folding models integrating structural and evolutionary constraints. Cell. doi.org/10.1016/j.cell.2025.06.014.