Differential Gene Expression Analysis using Resampling

Abstract

The primary objective in high-throughput sequencing is to identify differentially expressed genes, which provides substantial information of gene expression and regulation. A persistent challenge in this field is the bias caused by limited replicates. To address this, we have developed a novel Bootstrapping method. This approach enhances the power of DEG detection by augmenting the sample data points. New data points are generated through weighted geometric combinations of bootstrap samples and a pooled distribution. The pooled distribution consists of sample data from genes with similar expression levels.

Through simulation tests and evaluations on real-world data, our proposed Bootstrapping method exhibited competitive performance compared to common DEA tools (edgeR, DESeq2, Limma-Voom).

A key advantage of the Bootstrapping method is its independence from any assumptions about the sample distribution. This independence avoids the bias raised from inaccurate assumptions, offering the potential for broad application across various areas of genomic research.