基于深度学习的微生物高阶逻辑关系分析方法  

Deep Learning-Based Higher-Order Logical Relationship Analysis of Microorganism

作　　者：刘芃兰 孙硕男 Liu Penglan;Sun Shuonan(School of Electronic and Information Engineering,Liaoning Technical University,Huludao 125105)

机构地区：[1]辽宁工程技术大学电子与信息工程学院,葫芦岛125105

出　　处：《现代计算机》2022年第13期1-9,共9页Modern Computer

摘　　要：微生物群体是所有生物体和生态系统保持健康稳定的核心组分,分析生态系统中微生物的互作用,挖掘微生物群落共现模块,可以加深对微生物群落的认知,提高利用和改造微生物群落的能力,为生态修复、疾病治疗和药物研发提供新的手段。传统的微生物交互关系研究是以“成对关系”为基础的,然而,越来越多的研究发现高阶交互关系对于解释群落的多样性、复杂性具有重要作用。本文提出了一种基于深度学习的超图聚类模型(DeepHC),用来挖掘微生物之间的高阶逻辑关系。DeepHC利用信息熵挖掘微生物间普遍存在的高阶逻辑关系,通过深度神经网络获取样本的低维非线性表示,通过基于最大模块度的K均值聚类来自适应地执行聚类分析。实验结果表明,相对于其他模型,DeepHC具有更好的聚类效果,可以作为高阶逻辑关系挖掘的有效工具。Microbial community is the core component for the health and stability of all organisms and ecosystems.Analyzing the interaction of microorganisms in the ecosystem and mining the co-occurrence module of microbial communities can deepen the understanding of microbial communities,improve the ability to utilize and transform microbial communities,and provide new means for ecological restoration,disease treatment and drug development.Traditional research on microbial interaction is based on the“pairwise relationship”.However,more and more studies found that the higher-order interaction plays an important role in explaining the diversity and complexity of the community.In this paper,we propose a hypergraph clustering model based on deep learning(DeepHC)to analyze the high-order logical relationship of microorganisms.Specifically,DeepHC uses information entropy to mine higher-order logical relationships prevalent among microorganisms,obtains low-dimensional nonlinear representations of samples through deep neural networks,and adaptively performs clustering analysis through K-means clustering based on maximum modularity.Experimental results show that DeepHC has better clustering effect than other models and could be used as an effective tool for higher-order logical relationship analysis.

关 键 词：高阶逻辑关系 超图聚类 深度神经网络 最大模块度 

分 类 号：Q938[生物学—微生物学] TP18[自动化与计算机技术—控制理论与控制工程]