Comparative Evaluation of Filtering Strategies in Differential Gene Expression Analysis of RNA Sequencing Data

Abstract

Differential gene expression (DGE) analysis identifies genes expressed at varying levels between conditions, offering valuable insights into affected biological processes. RNA Sequencing (RNA-Seq) DGE analysis usually includes a filtering step to remove genes having low expression from the count data matrix. This study assesses the impact of different filtering strategies on DGE analysis. RNA-Seq read counts of the GSE150706 (n = 72) and TARGET (Therapeutically Applicable Research to Generate Effective Treatments) neuroblastoma (n = 84) datasets were used for analysis. DGE analysis was performed between the Pulled and Close-out groups in GSE150706 and between the MYCN-amplified and non-amplified groups in the TARGET neuroblastoma datasets. The effect of filtering strategies (filterByExpr, count, minimal, and no filtering) was assessed on the count data matrix, the number of low-count genes, the number of differentially expressed genes (DEGs) identified, and enrichment analysis. An adjusted p-value < 0.05 was set as the significance threshold for DGE analysis and enrichment analysis. For the GSE150706 dataset, 222, 288, 289, and 208 DEGs were identified from the filterByExpr, none, minimal, and count filtered matrices, respectively, while for the neuroblastoma dataset, 1662, 2059, 2075, and 1579 DEGs were identified from the filterByExpr, none, minimal, and count filtered matrices, respectively. FilterByExpr and count filtering returned no outliers and low counts at the end of DGE analysis. The filtering strategy also influenced enrichment analysis results. Filtering is an important step in DGE analysis with a significant impact on DGE output and downstream analysis. It is recommended to use filterByExpr or count filtering in DGE analysis of RNA-Seq data.