Differences between gene expression expressed as z-score or log2 values in Allen Brain Atlas

Hi everyone,

I know this is a very basic question, but why is that log2 gene expression values are sometimes very different even for similar z-score values? For example:

I noticed that log2 values are quite stable within a given probe, while z-score values are more stable between probes. Is this because they are normalized by different factors?

Which score would be better to use if I want to compare expression differences between genes?

Thank you.

Hi @spraguedawley. Z scores are defined as the number of standard deviations a value is away from the mean expression level, which makes it a good measure of differential expression. So if a gene has higher expression in hippocampus (for example) as compared to the rest of the brain, then the z score will be high regardless of what the average expression is in brain. I would not recommend comparing absolute expression differences between genes at all using microarray data. That said, the log2 expression values would be better to use than z scores for comparing such average expression levels between genes. Differential expression values (e.g., how much higher is gene x in hippocampus than in frontal cortex) are more reliable and can be compared between genes. This thread on data normalization and z scores goes into a bit more detail about some components of this response, but feel free to reply if you still have questions.

1 Like