Microbial risk score for capturing microbial characteristics, integrating multi-omics data, and predicting disease risk

• DOI: 10.1186/s40168-022-01310-2
• Publication Date: 2022-08-05T06:08:51Z
• Authors: Chan Wang; Leopoldo N. Segal; Jiyuan Hu; Boyan Zhou; Richard B. Hayes; Jiyoung Ahn; Huilin Li
• Keywords: Omics

Abstract Background With the rapid accumulation of microbiome-wide association studies, a great amount microbiome data are available to study microbiome’s role in human disease and advance potential use for prediction. However, unique features hinder its utility Methods Motivated from polygenic risk score framework, we propose microbial (MRS) framework aggregate complicated profile into summarized that can be used measure predict susceptibility. Specifically, MRS algorithm involves two steps: (1) identifying sub-community consisting signature taxa associated with (2) integrating identified continuous score. The first step is carried out using existing sophisticated tests pruning thresholding method discovery samples. second constructs community-based by calculating alpha diversity on validation Moreover, multi-omics integration jointly modeling proposed other scores constructed omics Results Through three comprehensive real-data analyses NYU Langone Health COVID-19 cohort, gut health index (GMHI) multi-study large type 1 diabetes cohort separately, exhibit evaluate prediction integration. In addition, disease-specific MRSs colorectal adenoma, cancer, Crohn’s disease, rheumatoid arthritis based relative abundances 5, 6, 12, 6 taxa, respectively, created validated GMHI cohort. Especially, achieves AUCs 0.88 (0.85–0.91) 0.86 (0.78–0.95) cohorts, respectively. Conclusions sheds light provides understanding diagnosis prognosis.