bigwig and replicates dilemma

[emolinari]

Hi guys!

I am visualizing bigwig files on the genome viewer (I use IGV) from RNAseq samples of human origin.
I have two clusters of 10 replicates each. I wanted to have just the experimental vs control track (otherwise it'll be too big screenshot, and also too confusing), but I am not sure what would be the good way to produce accurate data.

Should I use samtools merge to combine all the replicates from group#1 and group#2 (: all experimental combined and all control combined), and use those .bam files as input to create bigwig files? Or is there any other way?

Thanks to all!!!

Manu

[Richard Finney]

The ideal way to visualize this might be a box and whisker type plot using coverage values.

5 wiggles (wig) or big wigs overlaid as plots representing (for each genomic location) the summary values for the samples as 5 wiggle lines: the 5 lines would be
1) max, 2) min, 3) lower quartile, 4) upper quartile and 5) median for the TEST CASES.

If you had controls, another 5 lines (in another color) could represent such CONTROLS CASES.

Calculating it is pretty straightforward but I don't know if IGV or UCSC custom would handle it.

[Wallysb01]

It sounds messy to me to have 5 wiggles (or 10) overlaid. In my experience this is hard to interpret. I would personally just average the wiggle files and have test and control tracks separated. Its just a visualization track. The more meaningful statistics of the expression will be in DE test tables and what not, not in a browser view.

You should be able to use wig.math from here (https://github.com/timpalpant/java-genomics-io) to do that averaging.

[emolinari]

It sounds messy to me to have 5 wiggles (or 10) overlaid. In my experience this is hard to interpret. I would personally just average the wiggle files and have test and control tracks separated. Its just a visualization track. The more meaningful statistics of the expression will be in DE test tables and what not, not in a browser view.

You should be able to use wig.math from here (https://github.com/timpalpant/java-genomics-io) to do that averaging.

Hi Wallysb01 and Richard,

thanks for your reply!
I agree that it is just a visualization tool...but some of the recent papers I have read on similar topics include a table with the genome browser screenshot of some DE genes. Although it's not a validation of DE - you also need to do qPCRs of course- it might be a nice thing to add to your paper.
What do you think about that?

thanks indeed for the advices, I will try wig.math!

Manu

[chickenmcfu]

Hi, good question!

I have only 2x2 samples, but ran into the same "problem". 4 tracks seperated or overlayed do not give a nice picture. And thats what you want to have. As already mentioned, its only for visualization of the statistical correct, abstract table data.

I use IGB and used the build in graph options, which can create an averaged multigraph from selected graphs. I used my normalized wig files for this and overlayed the 2 resulting tracks.