A0166
Title: Nonparametric composition-on-composition regression analysis for high dimensional microbiome data
Authors: Xiang Zhan - Peking University (China) [presenting]
Abstract: High-dimensional compositional data are frequently encountered nowadays in scientific research of many disciplines, such as microbiome research. Regression analysis with compositional data being either responses or predictors has been well studied. However, when both responses and predictors are compositional, the inventory of analysis tools is surprisingly limited. Among the few existing methods, most of them rely on a log-ratio transformation to move compositional data analysis from simplex to real. Yet, a serious weakness of these methods is the failure to handle the substantial fraction of zeroes observed in microbiome data. To investigate associations between multiple high-dimensional microbial compositions, a nonparametric composition-on-composition (NCOC) regression analysis method is proposed, which does not require log-ratio transformations and hence can handle zeroes in the data. To account for high dimensionality, regression coefficients are estimated using a penalized estimation equation approach to improve its accuracy. Finally, statistical inference procedures are proposed to quantify uncertainty in the model predictions. The superior performance of NCOC and the validity and potential usefulness of the inference procedures are demonstrated through comprehensive numerical simulation studies, real data applications, and case studies.
