Reads within introns from polyA+ samples

[shambam]

Hi all,

Long time reader, first time poster.

I've been processing some RNASeq data where the RNA was isolated using a Clontech's SMART kit which selects polyA+ mRNA. This was done single end, and the reads were mapped using Tophat.

I find that many genes have significant numbers of reads in the introns which I thought would be suppressed due to polyA+ selection. I've attached an example which is a screen grab from Seqmonk. An example of the "problematic" area is marked in green.

It would be good to know what others think, and whether these reads should be used as a measure of expression, or whether it would be wiser to use only exonic reads.

I should also point out these were run on a HiSeq and multiplexed on the same lane.

Many thanks,

Sham

[Richard Finney]

This question is oft asked:
(
http://biostars.org/p/42890/

http://seqanswers.com/forums/showthread.php?t=11417

)

Short answer: it happens. No big deal. Just deal with it.
Some non-poly A reads get through "the filter".

Either method works for counting: Pick 1) count reads only hitting exons OR 2) reads (greater than or overlapping transcription start) AND (less than or overlapping transcription end).

[CPCantalapiedra]

Hi shambam,

your image seems good to me in terms of 5' or 3' bias. Have you checked this thoroughly? Do you have your reads homogeneously distributed along the genes? Just asking because we have huge 3' using Evrogen Mint-2, that could be considered analogous.

thanks
Carlos

[shambam]

Hi shambam,

your image seems good to me in terms of 5' or 3' bias. Have you checked this thoroughly? Do you have your reads homogeneously distributed along the genes? Just asking because we have huge 3' using Evrogen Mint-2, that could be considered analogous.

thanks
Carlos

Thanks Carlos. From what I see, I don't believe there is a bias (see attachment, good coverage with exons from end to end), but I have had datasets where a bias has happened. These samples were from low cell numbers, so the amplification prior to sequencing was probably the root cause of this.

Regarding my original post, I was just talking to the guy who did the extraction. He said it was a total RNA trizol prep followed by column clean-up. This was then amplified using the clontech SMART kit which primes off the ployA. I understand that nascent RNA will be in there, but I'm failing to understand why i see bursts of reads within introns, rather than low level intron wide coverage.

My concern is really whether these should be counted as the measure of expression, or as Richard said, just count within exons.

[CPCantalapiedra]

I would say the intron you marked and the two at its left are at the same distance than the expected exons but shifted to the left? Am I right?

[shambam]

I would say the intron you marked and the two at its left are at the same distance than the expected exons but shifted to the left? Am I right?

I see what you mean, but when I zoom in the distances between exons do not exactly match where I see reads. Just to be on the same side I just double-checked the genome/annotation versions and they all match. It looks the same on UCSC too.

Thanks for pointing this out.