RSeqQC infer_experiment strandedness

[lwhitmore]

Hi,

I am using RSeqQC infer_experiment.py to look at whether or not a library is stranded and I got the following results:

This is SingleEnd Data
Fraction of reads failed to determine: 0.0242
Fraction of reads explained by "++,--": 0.0398
Fraction of reads explained by "+-,-+": 0.9360


What does this mean? it appears that the reads are mapping to the reverse complement strand relative to the gene, what kind of sequencing protocol would do this? Does it have a name?

Thanks for any help,
Leanne

[Michael.Ante]

Hi Leanne,

The result shows you, that you have a stranded library. It also means, as you already said, that the reads are the reverse complement of the transcripts.
A couple of library preps produce reads in that flavour:
TruSeq stranded, ScriptSeq, Sense, ....
Just use the strandedness as detected in your analyses (e.g. htseq-count ... -s reverse ..).
Cheers,
Michael

[dpryan]

The "+-,-+" syntax is described here and has got to be about the most cryptic way of conveying things that's possible. What this means is that you have strand-specific data made with the dUTP method. This is the most common type these days and means that if you need to do counting, that you should use the "-s reverse" option in htseq-count or "-s 2" in featureCounts. If you happen to be using tophat2 for alignment, you want the "fr-firststrand" library type.

Edit: I should have refreshed, Michael beat me to it!

[lwhitmore]

Thanks guys! this was very helpful! The paper says an Illumina TruSeq Stranded mRNA LT kit so based on what Michael said it is to be expected that the read are the reverse complement of the gene. Another sequencing set from the same paper ( which was also indicated to use the Illumina TruSeq Stranded mRNA kit) gave me the following RSeqQC results:


Fraction of reads explained by "++,--": 0.4775
Fraction of reads explained by "+-,-+": 0.4956

Does this mean that maybe the strandedness part of the kit didn't work as well? Or maybe a non stranded kit was used?

[dpryan]

Either the paper was wrong or their kit failed completely. My guess is that the paper was wrong and that they used an unstranded kit.

[lwhitmore]

Thanks dpryan!

I do have another question if multiple alignments is specified in the bowtie alignment (-a) will infer_experiment be able to work correctly in predicting strandedness?

[dpryan]

Most likely it'll still work, though the percentages will be different.