机器学习在肠道菌群宿主表型预测中的应用

Application of Machine Learning in Phenotypic Prediction of Gut Microbiota

随着第二代DNA测序技术的发展,研究人员积累了大量的肠道菌群数据,研究表明肠道菌群与宿主健康状况存在密切联系,因此如何对复杂、高维的肠道菌群数据进行建模分析,是当前生物信息学研究中的重要挑战。人工智能的兴起为处理肠道菌群数据,揭示肠道菌群与宿主表型之间的复杂关系提供了可能。综述了现阶段肠道菌群与宿主表型之间的相关研究,重点介绍了常用的5种机器学习算法(线性回归、支持向量机、K-近邻、随机森林、人工神经网络)的理论原理及在相关研究中的应用,对预测宿主表型的机器学习算法选择提出了建议,并对该领域的未来发展进行了展望,以期为利用机器学习对肠道菌群宿主表型预测提供参考依据。

With the development of second-generation DNA sequencing technology,a large amount of gut microbiota data has been accumulated.The studies showed that gut microbiota were closely related to the health status of the host.Therefore how to model and analyze the complex and high-dimensional gut microbiota data has been an important challenge faced by bioinformatics at present.The rise of artificial intelligence had made it possible to process gut microbiota data and revealed the complex relationship between gut microbiota and host phenotypes.The paper summarized the present stage of gut microbiota and phenotypic correlation study among five kinds of machine learning algorithm(commonly used linear regression,support vector machine(SVM),K-nearest neighbor,random forests and artificial neural network),introduced five kinds of machine learning algorithms of theory and application in the related research,and to choose what kind of machine learning algorithms to predict recommendations to host phenotype.Finally,the future development of this field was prospected to provide a reference for predicting host phenotypes using machine learning based on gut microbiota data.