Predicting bacterial community assemblages using an artificial neural network approach
Predicting bacterial community assemblages using an artificial neural network approach
复制标题
DOI：
10.1038/nmeth.1975
复制
发表时间：
2012-06-01
期刊：
影响因子：
48
通讯作者：
Gilbert, Jack A.
中科院分区：
文献类型：
--
作者：
Larsen, Peter E.;Field, Dawn;Gilbert, Jack A.
了解地球微生物群与物理、化学和生物环境之间的相互作用是微生物生态学的根本目标。我们描述了一种生物气候建模方法，该方法利用人工神经网络来预测微生物群落结构作为环境参数和微生物相互作用的函数。这种方法在预测观察到的群落结构方面比没有纳入生物相互作用的几个单物种模型中的任何一个都要好。用该模型对群落结构随时间的变化进行了内插和外推，平均Bray-Curtis相似性为89.7。此外，还从地理上推断了群落结构，以创建第一张由单点观测得出的微生物图谱。这种方法可以推广到目前正在生成详细分类数据的许多微生物生态系统，为生态学研究中预测微生物分类结构提供了一种基于观测的建模技术。
Understanding the interactions between the Earth's microbiome and the physical, chemical and biological environment is a fundamental goal of microbial ecology. We describe a bioclimatic modeling approach that leverages artificial neural networks to predict microbial community structure as a function of environmental parameters and microbial interactions. This method was better at predicting observed community structure than were any of several single-species models that do not incorporate biotic interactions. The model was used to interpolate and extrapolate community structure over time with an average Bray-Curtis similarity of 89.7. Additionally, community structure was extrapolated geographically to create the first microbial map derived from single-point observations. This method can be generalized to the many microbial ecosystems for which detailed taxonomic data are currently being generated, providing an observation-based modeling technique for predicting microbial taxonomic structure in ecological studies.