Cross Sequencing Integration of Compositional Microbiome Data in Cancer

Abstract

[Abstract]: High-throughput sequencing has revolutionized our understanding of the human microbiome, providing detailed insights into microbial communities under various health and disease conditions. Among the most common strategies for studying the microbiome are 16S rRNA amplicon sequencing and whole genome shotgun sequencing (WGS), each with its own advantages and limitations. However, integrating and comparing results from data obtained through these two sequencing techniques presents a challenge due to the inherent differences in methods and discrepancies among datasets and their sources. This work evaluates batch effect removal (BER) methods for integrating microbiome composition data from different sequencing platforms. Using data from ten different cohorts, we applied BER methods such as Combat, Limma, FAbatch, MMUPHin, and Percentile-normalization. Our results demonstrate the effectiveness of these methods in reducing batch effects. However, it remains unclear whether the remaining biological signal is reliable, which is critical. Additionally, we compared GG2 with standard databases (SILVA for 16S and WoL for WGS), showing that GG2 enables more unified analysis (increasing the number of taxa shared among cohorts from 94 genera and 58 species to 215 and 210, respectively). In conclusion, our findings suggest that appropriate BER methods can harmonize microbiome data from diverse sequencing platforms, but further experiments are needed to reliably understand how the biological signal is modulated in the process.