Hi! I am using deepTools to generate heatmaps for different serine phosphorilations (ser2, ser5, ser7).
I have observed a behavior that I don't understand or I can't explain, so maybe you could help me with it.
I am splitting the heatmap into 4 different clusters, with --kmeans 4 argument. The problem is that, with the same bigWig file as input, I get two different clustering solutions, when I cluster a phosphorilation as a standalone or together with the other two phosphorilations. I am aware of the fact that, by default, deepTools takes all samples into consideration when clustering several samples. That's why when clustering all three phosphorilations together, I set the option --clusterUsingSamples accordingly. Still, I am not able to reproduce the clustering (nor the signal pattern) that I observe when looking at the phosphorilations individually. Of course, I use the same settings for both analysis, only differing in the two extra bigWigs provided to -S and that I provide the extra clusterUsingSamples option (in the example provided below, focusing on Serine 5, I used clusterUsingSamples 2).
So basically, my question would be, why do I see a different behavior not only in the clusters but also in the signal when using just one or all the different samples I have.
See attached as an example the plots of the serine 5 phosphorilation.
Thank you in advance,
Jordi
