A cross‐omics data analysis strategy for metabolite‐microbe pair identification

Abstract

Given the pivotal roles of metabolomics and microbiomics, numerous data mining approaches aim to uncover their intricate connections. However, the complex many‐to‐many associations between metabolome‐microbiome profiles yield numerous statistically significant but biologically unvalidated candidates. To address these challenges, we introduce BiOFI, a strategic framework for identifying metabolome‐microbiome correlation pairs (Bi‐Omics). BiOFI employs a comprehensive scoring system, incorporating intergroup differences, effects on feature correlation networks, and organism abundance. Meanwhile, it establishes a built‐in database of metabolite‐microbe‐KEGG functional pathway linking relationships. Furthermore, BiOFI can rank related feature pairs by combining importance scores and correlation strength. Validation on a dataset of cesarean‐section infants confirms the strategy's validity and interpretability. The BiOFI R package is freely accessible at https://github.com/chentianlu/BiOFI.