A Mixed Effect Similarity Matrix Regression Model (SMRmix) for Integrating Multiple Microbiome Datasets at Community Level and its Application in HIV.

Recent studies have highlighted the importance of human microbiota in our health and diseases. However, in many areas of research, individual microbiome studies often offer inconsistent results due to the limited sample sizes and the heterogeneity in study populations and experimental procedures. Integrative analysis of multiple microbiome datasets is necessary. However, statistical methods that incorporate multiple microbiome datasets and account for the study heterogeneity are not available in the literature. In this paper, we develop a mixed effect similarity matrix regression (SMRmix) approach for identifying community level microbiome shifts between outcomes. SMRmix has a close connection with the microbiome kernel association test, one of the most popular approaches for such a task but is only applicable when we have a single study. Via extensive simulations, we show that SMRmix has well-controlled type I error and higher power than some potential competitors. We also applied SMRmix to data from the HIV-reanalysis consortium, a collective effort that obtained all publicly available data on gut microbiome and HIV at December 2017, and obtained consistent associations of gut microbiome with HIV infection, and with MSM status (i.e. men who have sex with men).