RNA-seq with paired-end reads

[xuying]

anyone's doing RNA-seq with paired-end reads now? how to use pairing information is a challenge. any good ideas?

[dvh]

Yes. We've done paired end RNAseq - partly because of mixed sample type (some DNAseq etc) flow cells.

The paired end info is useful for novel gene expression and splice junction discovery.

For simple RPKM etc we just treated each read as single end and aligned to genome+splice junctions seperately - this is likely to be suboptimal although easy approach. I'd be very interested in people's comments...

david

[sci_guy]

Rna-pet

Paired-end ditags (PET) is an interesting strategy. It is very economical way of finding splice junction variants. Of course it doesn't offer the same resolution as full paired end RNA-Seq. I believe the team at GIS have migrated to Illumina and SOLiD in particular for RNA-PET now.

[monad]

PE and long read would be better to detect splicing junction and SNPs.

[ieuanclay]

Hi all, We have had success with PE RNAseq (illumina).

Some of the main pitfalls that we found:
-when you make your cDNA, if you use random hexamers (best results for us), your first strand synthesis reaction will tend to produce sequences that start/end in GC rich sequences. We have found that these tend to be non-accurate annealling products, and you can improve your mapping efficiency by stripping the first/last few bases.
- TopHAT (see bowtie/ben langmeads posts) has also been very useful (and extremely fast for mapping!!
- Creating 'spliced' reference genomes is very useful for retaining the benefits of PE mapping, without having to introduce huge 'insert' sizes...
- Avoid PET tags - the extra ligation steps, if you are not careful, can introduce a high percentage of 'false' splice events... With the illumina PE strategy there is no need for this...

Has anyone had any luck with barcoding RNAseq samples?

Ieuan

[seq_GA]

Hi all, We have had success with PE RNAseq (illumina).

Some of the main pitfalls that we found:
-when you make your cDNA, if you use random hexamers (best results for us), your first strand synthesis reaction will tend to produce sequences that start/end in GC rich sequences. We have found that these tend to be non-accurate annealling products, and you can improve your mapping efficiency by stripping the first/last few bases.
- TopHAT (see bowtie/ben langmeads posts) has also been very useful (and extremely fast for mapping!!
- Creating 'spliced' reference genomes is very useful for retaining the benefits of PE mapping, without having to introduce huge 'insert' sizes...
- Avoid PET tags - the extra ligation steps, if you are not careful, can introduce a high percentage of 'false' splice events... With the illumina PE strategy there is no need for this...

Has anyone had any luck with barcoding RNAseq samples?

Ieuan

Hi Ieuan,

It is interesting to see PE RNAseq.
I still ahve some doubts regarding mapping PE reads. Because ELAND can support eland_rna analysis for SR(single read) only.
Any tool to map tags with splice juction or can you please explain about the mapping part of PE RNA seq tags?

Thanks.

[ieuanclay]

Check the forums here for TopHat if you want to know about mapping RNAseq!

The actual way that you map is very dependent on the source of your RNA, so i would need to know more about your experiments! My main experience is with nascent RNA, so we PE map againstt a genomic reference. Spliced RNA can be mapped against spliced genomes (again see TopHat). Or a combination of both can be used, i.e. mapping against the genomic reference using a sensible PE insert limit, to find nascent RNA, then mapping the reads that don't map to the reference genome against a spliced genome. This way you can flag potentially spliced and unspliced transcripts.

There really are too many ways of proceeding, you just have to pick a processing pipeline that makes sense for your material/experiment and that you can justify when it comes to publishing.

Hope this helps!?

[baohua100]

no one show how to use paired-end info to infer junction and alternative splicing?

[karthikprabu]

hai ....I would like to know is it possible to use Bowtie to map the reads which does not have its genome sequenced?

[Pepe]

You can't use Bowtie without a reference genome. But there are others that can (Velvet, etc.)... just search for them in this forum...