A new paradigm in protein engineering – developed by researchers at UC Berkeley and dubbed AiCE – is on the verge of transforming the discipline. It combines structural and evolutionary constraints into a generalized inverse folding model. AiCE was just recently developed by a team led by Professor Gao Caixia at the Institute of Genetics and Developmental Biology (IGDB) from the Chinese Academy of Sciences. This new tool simplifies protein evolution while avoiding the requirements of sophisticated artificial intelligence models.
AiCE is a major step forward, enabling researchers to quickly and efficiently evolve proteins with a wide range of structures and functions. The platform has now walked successfully through eight different proteins so far from deaminases, to nuclear localization sequences, nucleases and reverse transcriptases. This powerful approach may be the most intuitive of all CRISPR’s contributions, but it is the most universal—affecting nearly every facet of protein engineering.
The Development of AiCE
The AiCE platform, spearheaded by research team leader Professor Gao’s lab, aims to simplify the protein engineering process. Yet conventional approaches necessitate intensive training on proprietary AI models, which can be a tedious and resource-heavy process. By contrast, AiCE lets you skip all of that through the magic of structural and evolutionary inference.
By considering these historical limitations, AiCE significantly improves the realism of protein evolution. In first benchmarking against 60 deep mutational scanning datasets, AiCE was found to significantly outpace other AI-based methods by a staggering 36–90%. In fact, even just adding the structural constraints led to a 37% increase in accuracy, illustrating the power of this framework.
“The ideal protein engineering strategy would achieve optimal performance with minimal effort.” – Phys.org
Versatility and Applications
AiCE’s power to evolve proteins with practically any desired function makes it a broadly applicable, universal strategy for protein engineering. Its inherent versatility offers researchers the opportunity to utilize the framework in numerous domains, sparking discoveries and innovations that could benefit biotechnology and medicine. The successful evolution of many proteins including deaminases, nucleases shows its wide applicability.
Additionally, the clarity of AiCE allows researchers with varying degrees of comprehension to utilize it effectively. This democratization of technology ignites further waves of innovation within the industry. With more scientists adopting this convenient method of protein engineering, we’ll continue to see encouraging innovations.
Publication and Future Prospects
The paper describing AiCE was just published in the journal Cell, giving an in-depth look at AiCE’s development and how it can be used. Our publication DOI 10.1016/j.cell.2025.06.014 has been made available here for anyone who wants to read more and dive deeper into this exciting new framework.
The possible effects of AiCE on the protein engineering profession are enormous. Expediting the melding of structural and evolutionary constraints increases productivity. It begins to open doors to some truly promising new directions in research and development. As scientists continue to explore its capabilities, AiCE may pave the way for groundbreaking discoveries that could reshape our understanding of protein functions.