Casestudy of GSE159751
TPM, FPKM, and RPKM are essential normalization metrics in RNA-Seq analysis that adjust for library size and gene length. Yet, in complex datasets, these linear methods may not fully correct systematic biases, especially nonlinear biases arising from library preparation, RNA degradation, or sample heterogeneity.
For example, in GSE159751, FPKMs are provided via the GEO database. Even after TPM/FPKM normalization, gene expression distributions vary substantially across samples. The distribution histograms show both single-peaked and bimodal shapes. Such patterns may reflect technical factors such as RNA degradation, batch differences, or library quality rather than biological variation. This example illustrates the value of visual aids in omics data analysis.
We also examined whether quantile normalization—a method that forces expression distributions across samples to be similar—can remove non-linear biases. Though it delivers distribution shape resemblance, it didn’t help remove the non-linear bias.
Thanks to bioinformaticians, we have many algorithms to tackle highly complex tasks. However, experimental biologists need to test whether these algorithms work in practice. At present, having a strong experimental design, a clear delivery plan for generating high-quality raw data, and effective monitoring tools is preferable to blindly relying on “sophisticated” algorithms.
If significant systematic errors are present in the data, it becomes difficult or impossible to perform an analysis that accounts for them. To avoid such situations, Subio strongly encourages you to conduct a preliminary assessment of your measurement techniques and design an experimental plan to minimize risks. For personalized support and to ensure reliable results, contact Subio before starting your experiments.