ComBat-seq: batch effect adjustment for RNA-seq count data.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Genomic data integration offers increased statistical power but is often compromised by batch effects.
• Existing batch correction methods frequently assume Gaussian distributions, which are unsuitable for skewed RNA-seq count data.
• Inappropriate assumptions can lead to erroneous results in genomic data analysis.

Purpose of the Study:

• To develop a novel batch correction method for RNA-seq data that addresses the limitations of existing approaches.
• To ensure adjusted data remains compatible with downstream differential expression analysis tools requiring integer counts.
• To improve the accuracy and reliability of genomic data analysis by effectively removing batch effects.

Main Methods:

• Developed ComBat-seq, a batch correction method utilizing a negative binomial regression model.
• The method preserves the integer nature of RNA-seq count data.
• Evaluated ComBat-seq performance through realistic simulations and a real-world data example.

Main Results:

• ComBat-seq demonstrated superior performance in simulations, enhancing statistical power and controlling false positives in differential expression.
• The method successfully removed batch effects in a real RNA-seq dataset.
• Biological signals within the data were effectively recovered after ComBat-seq adjustment.

Conclusions:

• ComBat-seq offers a robust solution for batch effect correction in RNA-seq studies.
• The negative binomial regression approach accurately models count data, leading to more reliable genomic analyses.
• ComBat-seq improves the utility of integrated genomic datasets by mitigating technical variation.