ribosomal RNA contamination

Question

[lkmklsmn]

Hi,
I am working on a RNAseq data set of about 20 human tissue samples. However, some of the samples have a very low alignment rate (~30-40%) while the rest has a a decent alignment rate around 80%. I aligned the reads using STAR with default settings. The log file tells me that for those samples with low alignment rate many reads did not map due to multimapping. My first guess is that there could have been an issue with the rRNA depletion for those samples. How can I show that the rRNA depletion failed?
Take unmapped reads and align to rRNA sequences? Has anybody had similar issues before?
Any feedback is greatly appreciated!

Answer 1

[Devon Ryan]

Yes, you can align against the 45S sequence, noting that this will (in my experience) somewhat underestimate the actual amount of rRNA in the sample.

As an aside, I should note that rRNA depletion only seems to work well if you do it on fresh samples. If you freeze them beforehand then the % of rRNA remaining is more likely to be high (this is purely observational).

[lkmklsmn]

Thanks! Its true that some of the samples had low RINs. Is there any way to correct for this? And still use the samples. As of right now, PCA clustering does not reveal my treatment structure and I am worried these samples are unusable.

[Devon Ryan]

Low RINs tend to correlate with degradation, so you might give salmon a try with whatever the position bias correction option is. Perhaps that'd help.