InteractiveComplexHeatmap on DESeq2 object with more than 2 groups

Hello all, I'm writing with the hope someone can clarify a doubt I have about the differential heatmap generated by the package InteractiveComplexHeatmap via the simple command interactivate(dds).

I read the package documentation at https://bioconductor.org/packages/release/bioc/html/InteractiveComplexHeatmap.html, but I couldn't find the answer I am looking for, because in the examples depicted the dds only contain two groups. What is the information shown by the differential heatmap in case of more than two groups?

I have a dataset with 3 groups, as you can see here: enter image description here

By selecting a FDR cutoff equal to that chosen when extracting results from pairwise comparisons (i.e. LFS1 vs LFC and LFS2 vs LFC) I see 271 significant genes. Anyway, I don't understand what they are. At first, I guessed this is the unique + common genes resulting from the two pairwise comparisons, but that doesn't seem to be the case.

Also, what are all the legends referring to? What is the data represented in the MAplot and volcano plot by default in case of multiple groups?

Looking forward to your reply. Cheers,

deseq2 interactivecomplexheatmap heatmap

• 1,555 views • link

updated 4.2 years ago by jared.andrews07

• written 4.2 years ago by miccoli.andrea89 •

1 answer

The comparison shown is just for whatever comes out of results(dds). It's not going to combine different comparisons for you, but it will show the results for all samples because it can't really read your mind about what you want or how the comparison is set up.

The volcano and MA plot for a given results would show the same comparison. You may be well served by reading the DESeq2 vignette.

The legends are just referring to all the annotations shown. The sample groups, the log2FoldChange is showing, well, the log2FoldChange as calculated by DESeq2 for whatever comparison is being used. The log10(baseMean+1) is basically the average expression across the groups in the comparison, again calculated by DESeq2. That's the far right row annotation. The sizeFactor is the scaling factor used for normalization by DESeq2. And the z-score are row-wise values compared to the mean of the row, a common way to show group-wise differences in a heatmap like this.

• 0 views • link

written 4.2 years ago by jared.andrews07

Thanks Jared.

Based on your reply and the dds vignette, I suppose that the log fold change(base 10 and 2) as well as the plots) refer to the last of the factor levels compared to the reference group.

However, I still have doubts as to what are the 271 genes. When extracting the results of the contrast above (i.e LFS2 vs. LFC) from DESeq2 and choosing a FDR cutoff of 0.05, I get a different number of modulated genes (156 up + 162 down), which doesn't add up to 271. Any clue?

• 0 views • link

written 4.2 years ago by miccoli.andrea89 •

I'd check resultsNames(dds), it'll be the first one in that list I think.

• 0 views • link

written 4.2 years ago by jared.andrews07