Harmonizing 16S rRNA Sequencing Data: Benchmarking Batch Effect Removal algorithms

Abstract

[Abstract]: This study evaluates batch effect removal (BER) algorithms’ efficacy in harmonizing 16S rRNA sequencing data, crucial for accurate microbiome analysis. Despite remarkable progress in generating microbiome datasets, challenges such as batch effects can confound analyses, emphasizing the need for robust BER algorithms. We benchmarked five BER algorithms using publicly available datasets: ComBat, limma, FAbatch, pnorm, and Mmuphin. Utilizing a scoring system based on machine learning classification accuracy, we assessed their ability to mitigate inter-cohort variability. Our results demonstrate that FAbatch effectively reduces batch effects, surpassing ComBat. However, further investigation into biological signal variation is warranted. This study underscores the importance of rigorous BER algorithms in ensuring data integrity and reliability in microbiome research.