CoMeDA: an integrated platform for compositional analysis of metabarcoding data with batch correction

Abstract Background Metabarcoding data generated from 16S rRNA and internal transcribed spacer (ITS) amplicon sequencing are inherently compositional, with observed abundances representing relative rather than absolute information. Ignoring this compositional structure can lead to spurious results in downstream analyses. In addition, multi-cohort studies frequently suffer from batch effects arising from differences in sample processing or sequencing platforms, which may confound true biological signals if not properly addressed. Despite the availability of multiple microbiome analysis platforms, these analytical considerations are not always incorporated together as central components of user-accessible workflows. Results We developed CoMeDA, an integrated web-based platform that treats the compositional structure of metabarcoding data and batch effects as coordinated components of a single analytical workflow. The platform supports both raw FASTQ files and pre-classified taxa tables, making it compatible with both short-read and long-read sequencing technologies. CoMeDA performs zero handling through probabilistic estimation and applies centered log-ratio (CLR) transformation as the default normalization strategy, followed by automated batch effect correction that removes technical variation while preserving biological signals. Downstream analyses, comprising diversity assessment using compositionally appropriate metrics, differential abundance analysis with effect size estimation, co-occurrence network inference, and functional prediction for both bacterial and fungal communities, are performed within the same compositional framework. CoMeDA also implements split-CLR transformation for cross-dataset analyses, enabling cross-kingdom and paired-condition correlation analyses while reducing spurious correlations introduced by directly combining datasets with different compositional structures. Using simulated datasets and a mock community with known ground truth, we evaluated the contribution of individual analytical components, the behavior of the automated batch-correction strategy, and taxonomic classification accuracy. We further reanalyzed a published multi-cohort gout dataset, a pediatric Crohn’s disease bacterial-fungal dataset, and a PacBio long-read periodontitis dataset. Using these datasets, CoMeDA recovered biologically coherent patterns while also highlighting additional literature-supported findings across different study designs. Conclusions CoMeDA provides a compositionally coherent workflow in which normalization, batch correction, and downstream analysis follow a consistent analytical logic. The platform features an interactive web interface and requires no programming expertise. CoMeDA is freely available online without registration, and a standalone Docker image is provided for local deployment.