miMatch: a microbial metabolic background matching tool for mitigating host confounding in metagenomics research.

Metagenomic research faces a persistent challenge due to the low concordance across studies. While matching host confounders can mitigate the impact of individual differences, the influence of factors such as genetics, environment, and lifestyle habits on microbial profiles makes it exceptionally challenging to create fully matched cohorts. The microbial metabolic background, which modulates microbial composition, reflects a cumulative impact of host confounders, serving as an ideal baseline for microbial sample matching. In this study, we introduced miMatch, an innovative metagenomic sample-matching tool that uses microbial metabolic background as a comprehensive reference for host-related variables and employs propensity score matching to build case-control pairs, even in the absence of host confounders. In the simulated datasets, miMatch effectively eliminated individual metabolic background differences, thereby enhancing the accuracy of identifying differential microbial patterns and reducing false positives. Moreover, in real metagenomic data, miMatch improved result consistency and model generalizability across cohorts of the same disease. A user-friendly web server (https://www.biosino.org/iMAC/mimatch) has been established to promote the integration of multiple metagenomic cohorts, strengthening causal relationships in metagenomic research.