Fig. 2
The optimal p value thresholds by P+T method for including taxa in MRSα, MRSwS, and MRSunwS, separately, using the metagenomic data in the NYULH COVID-19 cohort. Specifically, given a cut-off, the taxa with p values less than the cut-off were selected and defined as a sub-community. The p values were obtained by ANCOM-BC method. The leave-one-out CV was used for the predictions. MRSα: the negative alpha diversity (Shannon index) was calculated for each sample on the selected sub-community; MRSwS: the weighted sum of relative abundances of the selected taxa with the weights being the coefficients estimated from the ANCOM-BC log-linear model; MRSunwS: the sum of relative abundances of the selected taxa