Relative abundance heatmap interpretation

Question

[dpc]

HI community!!! I have a very basic question regarding heatmap. I have never done heatmap analysis before. I have generated a relative abundance heatmap from test and control samples. Can you please tell how much colour difference should I consider to conclude that there is an increase or decrease in relative abundance? I am seeing some taxa are having very little change in there colour difference which is observable only when I zoom out or look very closely. Should I consider those changes also? (I am using a log scale)

Thanks and Regards, DC7

Answer 1

[antonioggsousa]

Hi,

Can you provide an example? Can you post your figure or part of it, online? (even with fake data, but resembling your data) Otherwise it's a bit hard to help you properly.

Did you try to transform your data first, using Z-Score transformation/standardization (mean 0 and standard deviation 1). Depending on what do you have in your columns and rows heatmap, you can scale (Z-Score transform) the columns or rows highlighting row- or column-wise differences, respectively, in terms of increased or decreased abundances.

If you're working with gene expression data (RNA-seq) or OTU tables (from 16S rRNA amplicon data) a few numbers of features will have extreme high values. When you log-transform you reduce these extreme high values as well as the right long tail of the distributed values. However, still most of your values will have a small log-transformed number and, therefore, most of your data will have a similar color on a heatmap, highlighting basically the most extreme high values. However, when you scale the log-transformed data you'll be able to highlight differences among rows (scale columns) or columns (scale rows) in terms of the most abundant and less abundant features among rows or columns, respectively.

I hope this answers your question,

António