microarray heatmap and clustering

Question

[liyi.1989]

In the microarray analysis, what is the use of heatmap(is the color grid in the color map really useful?)?

I think the dendrogram on the top(column) is enough.

1. Is the colorful heatmap(gene expression value) really useful for providing some imformation? In some paper, it takes more than half page to present the beautiful coloful heatmap. But the analysis seems only based on the clustering of the column.
2. Is the clustering(dendrogram) for the row(gene) useful? I saw some paper also provide dendrogram on both dimensions, but what is the usage of the row clustering?

Thank you very much!

[linus]

Did you try a heatmap of microarray data without row clustering? Most of the times it looks quite messy. Even if they do not show the dendrogram of the rows, they are often clustered just for the sake of clarity. The dendrogram is often removed since a dendrogram of probably 10000 genes is quite confusing.

1. You could also just show the dendrogram itself, but a colorful map is always much nicer to look at. And since a paper is also about presenting the results, i would prefer a heatmap.

[liyi.1989]

Thank you for your advice! I am new to this aera : )

Answer 1

[Devon Ryan]

The actual heatmap is mostly useful to ensure that one or two probes aren't driving all of the clustering. The row dendrograms are mostly useful to see vague structure in the correlations. So you can see that a module is shared in groups A and B and then different between them and groups C and D. If the dendrograms are tightly clustered, then you might have nicely co-regulated differences. Having said that, it's best not to read too much into dendrograms like that.

Often I'd agree that the column dendrogram is sufficient to convey the information ...and the rest is just "oooo, pretty picture!"

[liyi.1989]

Thank you so much for your answer!