How does heatmap in DESeq2 manual is helping the quality assessment?

Question

[Parham]

Hi,

I am practicing DESeq2 manual. Under section 2.2.1 it plots some heatmaps which the purpose of doing it is some quality assessment according to the paragraph before. However it doesn't provide any conclusion for it. Whether the samples are good to do further analysis on or not. Or how to interpret the heatmaps. Since the three replicates of same condition are not showing similar patterns of expression and it is the same when I test it on my own data, I concluded that it doesn't have to show a similar pattern between replicates, but then I am wondering what is the aim of plotting such a heatmap? Any explanation on this is very appreciated.

Answer 1

[Devon Ryan]

The general thing we hope to see with heatmaps or PCA plots is whether the samples are clustering in a strange way or if any of them cluster very far away from the other. In the first case, this can give evidence for the presence of a batch effect (e.g., library extraction or creation date). in the second case, samples clustering well away from the others are probably outliers and are subject for exclusion. In a perfect world, you'd see samples clustering by treatment group, but that's often not the case and not something to worry about when it doesn't happen.

[Parham]

It's somehow confusing to me the way to deal with it. Since in my final analysis (enriched GO terms) I see terms that makes sense to my experiment but the heatmaps are not clustering same conditions together, at least one replicate of each. It would be very helpful if the manual would discuss it more or refer to some article that could help. Thanks for your answer Devon.