RNA-seq is a powerful tool for gene expression analysis, but its results are highly dependent on sequencing depth (SD). This study investigates how SD affects the detection of differentially expressed genes (d.e.g.) and the accuracy of RNA-seq data analysis. The authors propose a novel nonparametric method, NOISeq, which models noise distribution from the data and is less sensitive to SD than existing methods. Unlike traditional methods that rely on parametric assumptions, NOISeq uses empirical noise modeling to better control false discovery rates and detect true differential expression. The study shows that most existing methods are highly dependent on SD, leading to an increased number of false positives as read counts increase. In contrast, NOISeq maintains stable performance across varying SDs and is more effective in identifying true differentially expressed genes. The study also highlights the challenges of detecting low-expression transcripts and the impact of sequencing depth on transcript detection and expression quantification. The results suggest that while deep sequencing improves the detection of low-expression transcripts, it also increases noise, making differential expression analysis more challenging. The study emphasizes the importance of balancing SD between experimental conditions and using appropriate normalization techniques to reduce bias. Overall, the findings indicate that NOISeq is a robust and reliable method for differential expression analysis in RNA-seq experiments.RNA-seq is a powerful tool for gene expression analysis, but its results are highly dependent on sequencing depth (SD). This study investigates how SD affects the detection of differentially expressed genes (d.e.g.) and the accuracy of RNA-seq data analysis. The authors propose a novel nonparametric method, NOISeq, which models noise distribution from the data and is less sensitive to SD than existing methods. Unlike traditional methods that rely on parametric assumptions, NOISeq uses empirical noise modeling to better control false discovery rates and detect true differential expression. The study shows that most existing methods are highly dependent on SD, leading to an increased number of false positives as read counts increase. In contrast, NOISeq maintains stable performance across varying SDs and is more effective in identifying true differentially expressed genes. The study also highlights the challenges of detecting low-expression transcripts and the impact of sequencing depth on transcript detection and expression quantification. The results suggest that while deep sequencing improves the detection of low-expression transcripts, it also increases noise, making differential expression analysis more challenging. The study emphasizes the importance of balancing SD between experimental conditions and using appropriate normalization techniques to reduce bias. Overall, the findings indicate that NOISeq is a robust and reliable method for differential expression analysis in RNA-seq experiments.