Batch effects removal for microbiome data via conditional quantile regression

Area of Science:

• Microbiome research
• Bioinformatics
• Statistical modeling

Background:

• Microbiome data often contain batch effects due to specimen processing.
• Existing methods for genomic data are inadequate for zero-inflated, over-dispersed microbiome data.
• Current microbiome-specific methods have limited applications in association testing or specialized designs.

Purpose of the Study:

• To develop a novel method for mitigating batch effects in microbiome data.
• To create a flexible approach applicable to general study designs and analytical goals.
• To generate batch-corrected microbiome data suitable for downstream analyses.

Main Methods:

• Conditional Quantile Regression (ConQuR) approach.
• Utilizes a two-part quantile regression model.
• Employs non-parametric modeling to handle complex microbial read count distributions.

Main Results:

• ConQuR effectively removes batch effects from microbiome datasets.
• The method preserves biologically relevant signals of interest.
• Generated batch-removed data are suitable for subsequent analyses.

Conclusions:

• ConQuR offers a comprehensive solution for microbiome batch effect correction.
• The approach accommodates the unique characteristics of microbiome data.
• Enables broader applications of microbiome data analysis beyond association testing.