刘磊教授学术报告：A Marginalized Two-Part Beta Regression Model for Microbiome Compositional Data
报告题目：A MarginalizedTwo-Part Beta Regression Model for Microbiome Compositional Data
报告摘要：Inmicrobiome studies, one important goal is to detect differential abundance ofmicrobes across clinical conditions and treatment options. However, themicrobiome compositional data (denoted by relative abundance) are highlyskewed, bounded in [0, 1), and often with many zeros. A two-part model iscommonly used to separate zeros and positive values explicitly by twosubmodels: a logistic model for the probability of a specie being present inPart I, and a Beta regression model for the relative abundance conditional onthe presence of the specie in Part II. However, the regression coefficients inPart II cannot provide a marginal (unconditional) interpretation of covariate effectson the microbial abundance, which is of great interest in many applications. Inthis paper, we propose a marginalized two-part Beta regression model whichcaptures the zero-inflation and skewness of microbiome data and also allowsinvestigators to examine covariate effects on the marginal (unconditional)mean. We demonstrate its practical performance using simulation studies andapply the model to a real metagenomic dataset on mouse skin microbiota. We findthat under the proposed marginalized model, without loss in power, thelikelihood ratio test performs better in controlling the type I error thanthose under conventional methods.