Using log2 read counts for RNA-seq heatmap

[[email protected]]

I want to plot a heatmap for my deferentially expressed genes. I used partek genomics to get read counts for each gene and used EdgeR to identify significant genes. After log2-transformation of the read counts with or without quantile normalization, I plot a heatmap in R with heatmap2 function. However, the red and green color are not that distinctively different between treatment groups (ideally, I want to have red for one panel and green for the other). I think the reason is that some genes are highly expressed whereas some have lower expression (like 8000 reads vs. 10 reads/a million counts for different genes). How can I adjust the data for a nicer heatmap (attached)? Also, I found that some papers used raw z-score for plotting. May I ask what raw z-score is and how to calculate that? Thanks a bunch!!

[TiborNagy]

I have usually set the parameter scale='none'.
The standard normal distribution (mean = 0, standard deviation = 1) known as z-distribution. So the scale of this distribution is called z-score.
Coverting raw scores to z-score is very easy:
z-score = (x - mean) / sd, where x is the raw score, mean is the mean of x and sd is the standard deviation of x.

[jcrow06]

Hey thanks for this post.
Did this work for you?

I have a similar situation and like Tibor I have used the scale="none".
I am trying to examine my RNAseq samples with RNAseq data from TCGA.
I have their count table and mine and have performed the DE analysis using DESeq. I have generated nice results of significantly differentially expressed genes and would now like to compare the two on a heat map for genes that are considered differentially expressed in both experiments.

In doing so obviously the scales are different when I use the "getVarianceStabilizedData(myData)" option in DESeq. I am thinking that z-score normalizing would be the best approach and would love to know if this can be confirmed.

Cheers