Tutorial:A simple tutorial for a complex ComplexHeatmap

Question

[Kevin Blighe]

This is being added in response to my message on Twitter and LinkedIn about ComplexHeatmap. The new message is here.

Many folks seemed interested in understanding how such figures could be made; so, I have decided to make the rheumatoid arthritis gene expression data available and provide a 'fool-proof' walk-through of how the heatmap can be generated.

The tutorial is on GitHub at: https://github.com/kevinblighe/E-MTAB-6141

Kevin

[ATpoint]

By the way, cudos for the blue-black-yellow color scheme, very appealing to the eye to spot differences.

[Kevin Blighe]

I think that I may have taken it from Michael Love, but cannot recall.

[Rob]

Hi Kevin This tutorial is very helpful. I wonder how can I do hierarchical clustering here?

[Rob]

Hi Kevin Blighe

In your complexheatmap workflow, you scale data to z-score. do we still need vst() or rlog() transformation before scaling to z-score or just it is just enough? Can we do vst() and then scale to z-score?

Thanks

[ATpoint]

You always normalize data before z-scoring it. The Z-score just calculate like the "deviation" of every sample from the mean of all samples for the particular gene. If the samples are not normalized then neither the mean nor the expression values for each sample are accurate and therefore the Z-score makes no sense, so yes, use either of the two transformations prior to running the Z-scoring. In the end you Z-score as you want for each gene a fixed and standardized metric irrespective of the expression level of the gene. Otherwise highly expressed genes would dominate the color scale of the heatmap and the clustering would simply reflect the expression level of the gene rather than the samplewise differences.

[Rob]

Great! Thank you ATpoint