Directional information from PE reads?

[tboothby]

Hello,
Recently the lab I am in has done some RNAseq. We prepped with the Illumina TruSeq kit and did paired end reads.

I am curious if there is a way with the sequenced fragments that we have already generated to glean any sort of directional information on our sequences?

I know there are several pre-existing methods using other kits, but if we could get information from the data we have available that would obviously be preferable (even if it is only preliminary).

Thanks,
T

[csquared]

Short answer is no or at least not without other data to go along with the RNAseq data.

The production of the double stranded cDNA from RNA followed by the random orientation of the adaptor ligation means all directionality is lost in the final library.

[tboothby]

Thanks for your response. Do you (or anyone else with experience/knowledge) think or best option at this point would be to use Illumina's direction RNAseq protocol?

On their website they say that the primers used in this protocol have not been tested with the RNA truseq kit, so we would need to get the small RNA seq kit?

Cheers,
T

[pmiguel]

Unless the TruSeq primers/adapters have Uridine in them, I don't see what the problem would be.

I was looking at the TruSeq RNA prep kit protocol and it looked do-able adding strand-specificity to it. Maybe a little laborious though.

--
Phillip

[protist]

Not that laborious to modify the TruSeq protocol for 2nd strand dUTP and digestion of 2nd strand - I would argue it is easier doing those adjustments than going the RNA adapter route. Companies are offering it now - see https://www.fasteris.com/pdf/2011-11...eris_Basel.pdf

[pmiguel]

Not that laborious to modify the TruSeq protocol for 2nd strand dUTP and digestion of 2nd strand - I would argue it is easier doing those adjustments than going the RNA adapter route. Companies are offering it now - see https://www.fasteris.com/pdf/2011-11...eris_Basel.pdf

Yeah, I meant more laborious that doing non-strand specific libraries. But that was partially based on the idea that an AMpure clean-up after 1st strand synthesis would not be enough to remove all dTTPs prior to 2nd strand synthesis. But, now that I think about it, I guess it is not necessary to completely remove dTTP. Dropping the concentration down a couple of orders of magnitude should be sufficient to ensure that all 2nd strand get at least 1 dUTP incorporated in them.

--
Phillip

[frozenlyse]

Hello,
I am curious if there is a way with the sequenced fragments that we have already generated to glean any sort of directional information on our sequences?

Actually yes there is - reads which span a canonical splicing site can be assigned a strand due to the splcing site signals - the TopHat/Cufflinks pipeline uses this information I believe.

[pmiguel]

Actually yes there is - reads which span a canonical splicing site can be assigned a strand due to the splcing site signals - the TopHat/Cufflinks pipeline uses this information I believe.

Good point. But it does presuppose that a genome sequence exists.

--
Phillip

[csquared]

Actually yes there is - reads which span a canonical splicing site can be assigned a strand due to the splcing site signals - the TopHat/Cufflinks pipeline uses this information I believe.

Not sure I agree. The presence of the splice site sequence still doesn't tell you which strand the original transcript came from. Only that the read was generated from that strand. The production of ds cDNA followed by adaptor ligation and subsequent amplification of both strands means that reads can be generated from either direction.

I may be missing something and would be interested in a description of how it is possible to infer strandedness from a standard library prep for RNASeq.

[pmiguel]

Not sure I agree. The presence of the splice site sequence still doesn't tell you which strand the original transcript came from. Only that the read was generated from that strand. The production of ds cDNA followed by adaptor ligation and subsequent amplification of both strands means that reads can be generated from either direction.

I may be missing something and would be interested in a description of how it is possible to infer strandedness from a standard library prep for RNASeq.

Just trivially, if you align your RNA sequence to a genome and see that the genomic sequence contains intron sequence, not present in the the RNA, you can make an inference. The strand where the putative intron begins with a GT and ends with an AG, is likely the plus strand.

This is more an example of using an RNA sequence to annotate a genome, without having to know the strandedness of the read.

Of course, that read may have derived from an anti-sense RNA, and you would have no idea that was the case unless you have a strand specific library. So read counts of sense vs antisense would not be available to you.

--
Phillip

[frozenlyse]

Good point. But it does presuppose that a genome sequence exists.

True - I get a bit of tunnel vision since I only work in human/mouse.

I may be missing something and would be interested in a description of how it is possible to infer strandedness from a standard library prep for RNASeq.

>99% of splice sites (ie canonical splicing) have the intron sequence start with GU and end with AG. As these sequences are not reversible/complementary you can assign a strand to reads crossing such an intron (as long as you have the reference sequence). See http://en.wikipedia.org/wiki/RNA_splicing

[csquared]

>99% of splice sites (ie canonical splicing) have the intron sequence start with GU and end with AG. As these sequences are not reversible/complementary you can assign a strand to reads crossing such an intron (as long as you have the reference sequence). See http://en.wikipedia.org/wiki/RNA_splicing

Completely agree but as far as inferring strandedness from an area of the genome with unknown/no annotation, the canonical splicing sites don't help much. For example, non-coding RNA. A standard library prep will not tell you which strand it came off of. Even in protein coding regions, you would need a good annotation of the ORF and predicted splicing sites based on the canonical sites to be able to identify those splice sites in the RNA. But you would already know the + strand from the annotation and not need to infer the strand.

Of course the RNA information in combination with genomic sequence to generate annotation of a genomic region is valuable and I agree that strandedness can be inferred for many regions by combining the genomic and RNA data together.