【新华社】科研团队成功利用人工智能蛋白语言模型揭示生命演化奥秘

Core Insights - The research reveals the significant role of protein high-order features in functional adaptive convergent evolution, providing a new perspective on the mystery of life evolution [1][6] - The study utilizes an artificial intelligence-based protein language model to uncover the molecular mechanisms behind convergent evolution, moving beyond traditional methods that focus solely on single amino acid changes [2][4] Group 1: Research Methodology - The research team developed a computational analysis framework named "ACEP," which leverages pre-trained protein language models to capture complex contextual information and high-order features from protein sequences [2][4] - ACEP's analysis process involves calculating the real distance of homologous protein embedding vectors, simulating neutral evolution to construct background distance distributions, and statistically testing for significant high-order feature convergence signals [4][5] Group 2: Findings and Implications - The ACEP framework successfully identified significant high-order feature convergence signals in known cases, such as the Prestin protein in echolocating mammals and PEPC/PPCK proteins in Crassulacean acid metabolism plants [5] - The framework also identified hundreds of candidate genes with convergent signals in bats and toothed whales, with some genes significantly associated with sensory perception functions related to echolocation [5][6] - This research marks a paradigm shift in evolutionary biology by systematically revealing the importance of protein high-order feature convergence in adaptive evolution, showcasing the potential of AI technology in addressing complex biological questions [6]