##AI助力蛋白质功能预测，上海理工、牛津等团队突破性成果登Nature子刊
蛋白质是生命活动的执行者，其功能的解析对于理解健康、疾病、进化等至关重要。然而，目前超过2亿种蛋白质的功能尚未得到有效注释，这极大地限制了相关研究的进展。近日，来自牛津大学、苏黎世联邦理工学院、上海理工大学和北京师范大学的研究团队，在蛋白质功能预测领域取得重大突破，其基于统计的图网络方法PhiGnet，成功登上了《自然通讯》期刊。
PhiGnet的独特之处在于它巧妙地利用了蛋白质序列中蕴含的进化信息，通过深度学习模型，将蛋白质序列转化为图网络，并通过图卷积网络（GCN）进行分析，最终实现对蛋白质功能的精准预测。该方法不仅在性能上超越了现有方法，更重要的是，即使在缺乏蛋白质结构信息的情况下，也能有效地缩小序列与功能之间的差距。
研究团队通过对大量蛋白质数据的分析，发现蛋白质序列中存在着丰富的进化信息，这些信息可以反映蛋白质的结构和功能。PhiGnet正是基于这一发现，将蛋白质序列转化为图网络，并利用图卷积网络来学习这些信息，从而实现对蛋白质功能的预测。
PhiGnet的应用范围十分广泛，它可以用于预测蛋白质的酶活性、基因本体注释、功能位点识别等。研究人员表示，PhiGnet的出现将为生物医学研究提供强大的工具，有助于加速蛋白质功能的解析，并为新药研发、疾病诊断等领域带来新的突破。
值得一提的是，该研究团队由来自不同国家和地区的科学家组成，体现了国际合作在科学研究中的重要作用。这项成果不仅是人工智能在生物医学领域应用的又一成功案例，也为未来蛋白质功能研究指明了新的方向。
英语如下：
##AI Makes a Breakthrough in Predicting Protein Function!
**Keywords:** Protein, AIPrediction, Functional Annotation
**News Content:**
## AI-powered Protein FunctionPrediction: Breakthrough by ShanghaiTech, Oxford, and Other Teams Published in *Nature Communications*
Proteins are the executors of life activities, and understanding their functions iscrucial for comprehending health, disease, and evolution. However, the functions of over 200 million proteins remain uncharacterized, severely hindering research progress. Recently, a research team from the University of Oxford, ETH Zurich, ShanghaiTech University, and Beijing Normal University achieved a significant breakthrough in protein function prediction. Their statistical graph network method, PhiGnet, was successfully published in *NatureCommunications*.
PhiGnet’s uniqueness lies in its ingenious utilization of evolutionary information embedded in protein sequences. Through a deep learning model, it transforms protein sequences into graph networks and analyzes them using graph convolutional networks (GCN), ultimately achievingprecise prediction of protein functions. This method not only surpasses existing methods in performance but also effectively bridges the gap between sequence and function, even in the absence of protein structural information.
By analyzing vast amounts of protein data, the research team discovered abundant evolutionary information within protein sequences, reflecting their structure and function. PhiGnetleverages this discovery, converting protein sequences into graph networks and employing graph convolutional networks to learn this information, thereby enabling protein function prediction.
PhiGnet has a wide range of applications, including predicting protein enzymatic activity, gene ontology annotation, and functional site identification. Researchers believe that PhiGnet will provide a powerful toolfor biomedical research, accelerating protein function characterization and leading to new breakthroughs in drug development, disease diagnosis, and other fields.
It’s worth noting that the research team comprises scientists from different countries and regions, highlighting the crucial role of international collaboration in scientific research. This achievement not only represents another successful application of artificialintelligence in the biomedical field but also points to a new direction for future protein function research.
【来源】https://www.jiqizhixin.com/articles/2024-08-22-3
Views: 2