tool for checking the stranded vs unstranded library

[MichalO]

Assuming that we have by chance a FASTQ from a library that we do not know if was done with stranded or unstranded protocol - how to check this fact?

The library was most likely done as paired end with Tru Seq - whole transcript RNAseq.

A tool for counting antisense matches probably would do - what would you suggest? Meci!

[Simon Anders]

What about simply _looking_ at your data instead of pulling through some script? You shouldn't go ahead with an analysis without having made at least two or three spot check with a genome browser (e.g., IGV), and that tells you not only about strandedness.

[mknut]

Here is the answer (with pretty pictures). I agree with simon - checking manually a few reads will not only tell you about the strandedness, but will also give you an idea about the insert length.

[MichalO]

Thanks, MK!

Yup, Simon - looking in-depth is the first choice for me too
But we see some strange, antisense-looking thingies in IGV in some of the places- and some not very relaxed people call for a "metrics"

[Simon Anders]

And these "not very relaxed" persons think that a good way of dealing with such strange looking worrying features is to condense them into a single number in the hope that that makes them go ahead? That actually rather sounds too relaxed to me. ;-)

If you must, you could count everything with htseq-count, once with '-s yes' and once with '-s reverse' and make a scatter plot of counts per gene on the sense versus on the antisense strand.

BTW, don't forget that there are lots of antisense transcripts in many species.

[kcchan]

If you want a very quick and dirty check for strandedness, you can look for A/T and G/C skew in your reads. Stranded libraries will have a visible skew while unstranded libraries should have an even A/T and G/C skew. Of course, this isn't fool proof, so definitely go more in depth as others have suggested if you want to make sure.