Rna-Seq Expression For 3' Transcripts

Question

[Sander Timmer]

I've RNA-Seq measurements for 1 sample using 2 different libraries. As one of the libraries is only build to capture the 3' of the transcripts I want to make a 3' expression comparison between the 2 different methods (the other library is a more default Illumina one, so should have reads for the whole transcript).

So-far I've just been using Bowtie and then Cufflinks to compare FPKM values for both methods, though this makes no sense as this isn't a flexible method that can deals with the more specific 3' library we have been using.

Is there any tool or setting that I can use to do this quickly just to have a preliminary comparison of how our new library performs?

[Damian Kao]

What exactly do you want to compare? Do you basically just want to look at the tag counts at the 3' end of transcripts between the two libraries? Do you already have a set of transcript annotations? You can modify your annotation file to leave out 1/2 or 2/3 of the 5' end and remap to just the 3' end. Then get tag counts and compare.

[Sander Timmer]

Yes, basically I just want to look/compare the tag counts at the 3' end of the transcripts. So how would I modify my annotation file? I'm using the Drosophila melanogaster Ensembl annotation file (from here http://cufflinks.cbcb.umd.edu/igenomes.html), is there an easy way to automatically leave out information from there or do I need to hack this myself?

Answer 1

[Damian Kao]

A more systematic way to do this might be to convert your BAM files into a wig/bed coverage file that shows coverage per base of your transcripts. Then you can weigh each coverage score by distance from the 3' end. Then get an average of the weighted coverage and compare between the two libraries. So for example:

-you have a transcript of sequence length 100

-weigh the coverage score by distance from the 3' end by: coverage of base * distance to base / total length

-for example, the 4th base from the 3' end would be: coverage score * 97/100. The first base from the 5' end would be: coverage score * 1/100.

-so the coverage of bases farther from the 3' end would be weighed less than coverage near the 3' end.

-take an average of the weighted coverage. The 3' library should on average have higher coverage.

And remember to normalize the coverage scores by library size.

[Rm]

@DK: nice explaination...

[Sander Timmer]

Thanks, really useful!