Data Normalization for RNA-seq and Heatmaps

Data Normalization for RNA-seq and Heatmaps

Online Inquiry


The exponential growth and widespread adoption of high-throughput sequencing technologies such as RNA sequencing (RNA-seq) have advanced our understanding of genomics and transcriptomics. These technologies have provided insights into the complex mechanisms of biological systems, illuminating their complexity on an unprecedented scale. However, gathering meaningful interpretations from the vast amounts of data generated by these technologies requires effective data normalization and processing algorithms. One of the most visually informative ways to represent processed data is through heatmaps, which vividly depict gene expression in different samples.

RNA-Seq Data

Normalization of RNA-seq data is a key cornerstone of gene expression analysis. It is designed to minimize potential technical bias and thus facilitate fair comparisons between samples. The primary goal of standardization is to correct for factors that may distort the interpretation of results, such as differences in the length of sequenced genes or differences in the depth of sequencing between samples.

A software package commonly used for the normalization of RNA-seq data is DESeq2. This employs an intrinsic normalization method to calculate the geometric mean of each gene across all samples. Next, the gene counts in each sample are divided by this average to generate a ratio. The median of these sample-specific ratios is designated as the "size factor" for the corresponding sample. This modus operandi employed by DESeq2 provides a powerful and effective strategy for eliminating technical noise, thus ensuring a more accurate interpretation of gene expression data.

Pipeline for benchmarking the optimal workflow for constructing coexpression networks from RNA-seq data.Fig. 1. Pipeline for benchmarking the optimal workflow for constructing coexpression networks from RNA-seq data. (Mount, D. W., 2001)

Heatmap Data Normalization

Heatmaps provide a compelling tool for visualizing RNA sequencing results, enabling rapid identification of genes that are up or down-regulated across samples or conditions. In the context of heatmaps, Z-score normalization is typically performed on the normalized read counts associated with each gene sample. This involves calculating the Z-score on a per-gene basis, subtracting the mean, and dividing by the standard deviation.

The resulting Z-score scaling indicates that genes exhibiting high expression relative to the mean have positive Z-scores (usually shown in dark red), while low-expression genes exhibit negative Z-scores (usually shown in blue). Thus, the array of colors in the heatmap reflects changes in the expression of individual genes across the sample set, providing a concise and comprehensive visualization of gene expression data.

However, it is important to recognize that "log CPM" (counts per million) values are also commonly used for normalization in heatmaps, especially when processing large-scale RNA-seq data. This process involves calculating the CPM value for each gene using the effective library size determined by normalizing the TMM (trimmed mean of M values). Subsequently, a Z-score normalization is performed in which the counts for each gene are centered on the mean and scaled to unit variance.


Essentially, the normalization process associated with RNA-Seq data and the generation of heat maps form a critical component of gene expression analysis. By applying rigorous and appropriate normalization techniques, we can extract more accurate insights from the labyrinthine maze of high-throughput sequencing data, helping to elucidate biological systems and mechanisms of disease progression. The choice of appropriate standardization techniques depends heavily on the specific experimental setup and the nature of the samples used. These considerations, therefore, form the cornerstone of RNA-seq experimental design and analysis.


  1. Johnson K A, Krishnan A. Robust normalization and transformation techniques for constructing gene coexpression networks from RNA-seq data[J]. Genome biology, 2022, 23: 1-26.
* For Research Use Only. Not for use in diagnostic procedures.
Online Inquiry