Strand-specific vs. non-strand-specific RNAseq

Question

[Y Tb]

I want more details about Strand-specific vs. Non strand-specific protocols in RNAseq. I know that the benefit of strand-specific is to know whether the read originated from the +ve or - ve strand, also it helps identify antisense RNA, and I know it costs more than non-strand-specific. So could you please give me more information about this protocol. I am going to do RNA-seq analysis for human and mouse to detect antisense.

[Sam]

The strand specific library should only be slightly more expensive than the non-strand specific library from what I remember. So I always prefer the stranded library. What kind of information do you want? The procedure of the protocol? Or the type of strand specific library?

You might want to read this discussion

[Y Tb]

Dear Sam,

Could you please send me the information about the procedure of the protocol because I am not biologist major and I want to know more about it. Also, about the type of strand specific library.

BTW, the discussion link doesn't work

[Obi Griffith]

The link works for me.

Answer 1

[Obi Griffith]

You might start by just checking out the literature for these two popular protocols from Illumina.

• The TruSeq Stranded Total RNA Sample Prep Kit
• TruSeq Stranded mRNA Sample Prep Kit

They share the same Product Literature and it should give you a good introduction to "information about the procedure of the protocol".

As mentioned by Sam you can use STAR for alignment and HTSeq for counting with strand-specific modes specified for both. You could then use edgeR or a number of alternate BioConductor packages for differential expression analysis. Alternatively, you can use TopHat for alignment (in stranded mode) and Cufflinks for estimation of isoform expression and more.

You could take or review materials from our course at the Canadian Bioinformatics Workshops on RNAseq analysis.

Answer 2

[Istvan Albert]

If you need to detect antisense transcripts then it makes no sense (pun intended) to use non-sense reads .

[Y Tb]

Dear Albert,

Could you please let me know what is the best way to detect antisense and count the reads (which sense and which antisense). My RNA-seq data are for human/mouse.

Answer 3

[Sam]

If you want an illustration of the protocol, you can read it here

As for the analysis, you can use STAR for alignment and HTSeq for counting. Both of them supports strand specific library preparation.

[Lhl]

STAR manual says 'If you have stranded RNA-seq data, you do not need to use any specific STAR options.' I do not understand how it distinguishes between SS Lib and Non-SS lib.

[Obi Griffith]

I don't think it does distinguish between them. It simply assumes strand-specific by default. If you have unstranded (non-strand-specific) then you must supply extra options. It also doesn't distinguish between different types of strand-specific libraries. It is up to you to know this information about your RNAseq library and use the appropriate options for STAR and Cufflinks.

From the STAR manual:

For unstranded RNA-seq data, Cufflinks/Cuffdiff require spliced alignments with XS strand attribute, which STAR will generate with --outSAMstrandField intronMotif option. As required, the XS strand attribute will be generated for all alignments that contain splice junctions. The spliced alignments that have undefined strand (i.e. containing only non-canonical unannotated junctions) will be suppressed.

If you have stranded RNA-seq data, you do not need to use any specific STAR options. Instead, you need to run Cufflinks with the library option --library-type options. For example, cufflinks... --library-type fr-firststrand should be used for the standard dUTP protocol, including Illumina's stranded Tru-Seq. This option has to be used only for Cufflinks runs and not for STAR runs.