Heatmap with just two columns

Question

[Mozart]

Hello everyone, I have been asked to summarise information from my count matrix - obtained from 4 replicates - in just two columns (one column per group). What's the best approach to do this? I usually use the following steps to get a classical heatmap containing as many tiles as per replicate

top <- head(order(-rowVars(assay(rld))),100)

and then I take out the row mean

mat <- assay(rld)[ top, ]
mat <- mat - rowMeans(mat)

Once I got what I want I use heatmap.2 to plot my results.

Since I have already applied the mean centring here, if I wanted to present my data as just two separate columns (thus merging values from each replicate in just one tile/condition), I should take the average from each group (i.e. condition 1 and condition 2); in this way I always get almost 'specular' results and I don't think it is correct to do this way. Any ideas?

Answer 1

[ATpoint]

Z-scaling with only two columns is not very informative imho since you will see only a bimodal high and low pattern. Why don't you do it like everyone does, so perform standard DE analysis, take the differential genes and then show a heatmap for those genes based on the Z-score for all samples.

[Mozart]

Thanks so much for your reply. Sometimes it is hard to change someone else's mind, if you know what I mean! It's all about being persuasive and kind at the same time but it's just hard sometimes. But thanks a lot, I do appreciate the fact that the common way of doing this is actually the usual approach I use..

[ATpoint]

I have the experience that PIs can be easily persuaded with nice figures rather than words. Show them a proper alternative and things will be ok ;-)

[Mozart]

Thanks a lot!! I will do!