Plotting chip signal at TSSs with deeptools

Question

[srhic]

Hello,

I have a very basic question about using deeptools computeMatrix function to plot ChIP/ATAC-seq signal across TSS sites. I am just curious what is the ‘standard’ way of making the bed file of the TSSs I want to plot. I can make the bed file in two ways:

1. I get the gene start coordinate of my genes of interest from an online database like ensemble and make the bed file. This gives me one start site per gene.

2. I get all ensemble annotated TSSs associated with a gene which gives me multiple regions associated with each gene. This seems like the better option as I am not ignoring any TSS but I am concerned that since most TSS are pretty close to each other I will be double counting most of them.

Since such plots are very common in literature, I was just curious what is the best practice for making them?

Answer 1

[Devon Ryan]

Within the deepTools team we use the second option you listed. Yes, this results in some duplication, but since the duplicate regions are ambiguous in terms of which transcript they're likely to affect this ends up maintaining more biological meaning.

[srhic]

Thanks. I found an old thread mentioning the same but also recommending that overlapping regions in the bed file be merged before plotting. Do you think that is a good idea?

[Devon Ryan]

There's no need to merge the regions and that's rarely a good idea.

[QX]

hi Devon Ryan,

Thank for the answer, I would like to ask how can I make the 'ensemble annotated TSSs associated with a gene' bed file? is that the region from the peak callings itself (for e.g the peak from macs2)?