[Gonza]

i guess, on the positive side, is better to have an 'exotic' contamination rather than the usual e.coli

Would you suggest dumping these 2 libraries? As i mentioned, I still map 8MR to the TAIR10 genome. Is that 8MR usable ? What would you do?

thanks.

[GenoMax]

8M may be enough for Arabidopsis but is the difference very large for other samples (8M vs how many mapped)? edgeR/DESeq2 should be able to account for these differences during analysis (https://support.bioconductor.org/p/41719/).

If the samples are critical then asking your sequence provider about what happened and perhaps re-making the libraries for those two samples would be an option.

[Brian Bushnell]

Removing reads with quality below 30 could seriously bias any RNA-seq quantification you are trying to do, due to the sequence dependence of Illumina quality values.

[Gonza]

thanks, appreciate the feedback. Two more things ....

* this problematic libraries has 4 million reads less than the others. Will cufflinks account for these differences too?

* should i keep reads with values >Q20 instead of Q>30?

G

[Brian Bushnell]

I don't really see an upside for throwing away low-quality reads before mapping, though quality-trimming is often useful, to perhaps Q10 or so. To minimize bias, though, I think it's generally better to use a more sensitive aligner if the quality is very low.