Fig. 1

The workflow of the microbial risk score (MRS) framework. Data input: a phyloseq-class object is needed, which consists of a feature table (observed count table), a sample metadata, a taxonomy table (optional), and a phylogenetic tree (optional). MRS Algorithm has two steps: Step 1 is to identify a sub-community consisting of the signature microbial taxa with the P+T method and AUC evaluation in the discovery cohort. The black ROC curve which has the largest AUC determines the optimal p value cutoff. Step 2 is to integrate the identified microbial taxa into a continuous score, i.e., calculate the MRS value for each sample by calculating the diversity of the identified sub-community with the Shannon index. In addition, the constructed MRS is independently validated in the validation cohort. Application: In this manuscript, we perform multi-omics data integration for disease prediction by jointly modeling the proposed MRS and other risk scores constructed from other omics data in two real data cohorts