This website requires JavaScript.

A Graphical Model for Fusing Diverse Microbiome Data

Mehmet AktukmakHaonan ZhuMarc G. ChevretteJulia NepperJo HandelsmanAlfred Hero
Aug 2022
摘要
This paper develops a Bayesian graphical model for fusing disparate types ofcount data. The motivating application is the study of bacterial communitiesfrom diverse high dimensional features, in this case transcripts, collectedfrom different treatments. In such datasets, there are no explicitcorrespondences between the communities and each correspond to differentfactors, making data fusion challenging. We introduce a flexiblemultinomial-Gaussian generative model for jointly modeling such count data.This latent variable model jointly characterizes the observed data through acommon multivariate Gaussian latent space that parameterizes the set ofmultinomial probabilities of the transcriptome counts. The covariance matrix ofthe latent variables induces a covariance matrix of co-dependencies between allthe transcripts, effectively fusing multiple data sources. We present acomputationally scalable variational Expectation-Maximization (EM) algorithmfor inferring the latent variables and the parameters of the model. Theinferred latent variables provide a common dimensionality reduction forvisualizing the data and the inferred parameters provide a predictive posteriordistribution. In addition to simulation studies that demonstrate thevariational EM procedure, we apply our model to a bacterial microbiome dataset.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答