【科技日报】新型蛋白质改造方法成功开发

Core Insights - The research team from the Chinese Academy of Sciences has developed a new AI-based protein modification method that utilizes existing universal protein folding AI models, allowing for efficient protein evolution simulation and functional design without the need for specialized AI model training [1][2]. Group 1: Methodology and Innovation - The new protein modification method, named AiCEsingle, leverages a universal protein folding AI model to predict possible amino acid sequences based on a given three-dimensional structure [1]. - Traditional protein modification methods are time-consuming, costly, and heavily reliant on expert experience, while existing AI prediction technologies require separate training for each protein, leading to poor generalizability and high computational resource consumption [1]. Group 2: Performance and Impact - The new method achieved a prediction accuracy of 16%, with performance improvements ranging from 36% to 90% compared to other common AI models [2]. - The method successfully modified eight proteins with various functions, including key gene-editing tool deaminases, demonstrating its broader applicability and efficiency compared to traditional methods [2]. - This innovation significantly lowers the barrier to using AI technology, allowing ordinary laboratories to benefit from intelligent predictions without the need for expensive computational resources [2].