Plotting multiple signal files against a single genomic region

Question

[keller.mckowen]

I would like to create a sorted heatmap around a single gene region, across over 1600 experimental samples. In each experiment, a TF has been knocked down in a cell line. I have processed the data and normalized the samples against a negative control for each replicate plate. I now have normalized bigwig files.

I would like to make a heatmap in which each signal line comes from a different bigwig file in my sample library. The region would be the same for every line, my gene of interest. Using this, I could sort the experiments by which had a strong effect on my gene.

This seems like a useful function to add to an analysis suite such as Deeptools. I will try to implement a solution myself. But I am certain someone here can come up with a more elegant and integrated solution that I could.

Thank You for your consideration. -Keller

Answer 1

[Alex Reynolds]

Make a custom track in the UCSC Genome Browser, setting bigWig-sourced signal tracks in "dense" mode, and arrange or sort tracks in your desired ordering. Signal for each row is taken from your bigWig files.

The "dense" mode will give you a figure very heatmap-like in appearance, e.g.:

You can modify the custom track to add color to rows on a per-signal basis. You might assign colors to groups or categories of sample types/conditions/treatments/etc.

If there are annotation tracks you want to add, you can put them into the browser.

Once you have your browser and custom track file arranged the way you want, you can use a tool like soda.py to generate a gallery of browser snapshots, taken at each position in an input BED file.

This can be a single region, per your question, or multiple regions. For example, to make the figure above, I simply echo'ed an ad-hoc BED interval to a text file called adHoc.bed, and used that as my input:

$ echo -e "chr9\t133252000\t133280861" > adHoc.bed

If you have multiple regions of interest, this input BED file does not need to be in sort-bed order. Often it is useful to sort regions by other criteria, such as ID value in the fourth column, or score or signal value in the fifth column. This determines the order of snapshots in the gallery.