How much rRNA sequence is reasonable in an RNA-seq dataset?

[Nigel Saunders]

Dear all,
I have just received two RNA-seq datasets from a bacterial comparision. The reads are at least 50% (have not been able to look deeper yet) of 16 and 23S rRNA (i.e. all the top over-represented sequences by fastqc are rRNA). The centre that processed this insisted on doing the RNA depletion as part of their pipeline, and I believe used the Ribo-Zero (Epicentre) methodology which is supposed to be a good one.
What percentage post depletion should I have typically expected? Is this normal, or have they failed to do the rRNA depletion properly?
Nigel

[Melissa]

I think this topic has been discussed in the old thread. Back in the days when people are still using Invitrogen Ribominus kit, 30-40% rRNA in a RNA-Seq data is quite good. The new kits should perform better. In my own experience, percentage of rRNA in the data depends very much on the mapping parameters you use. I can map 30% if I allow 2 mismatch in 50bp reads but up to 50% if I use a min of 30bp extract matching. That's a huge difference.

[Lucilia]

I think this topic has been discussed in the old thread. Back in the days when people are still using Invitrogen Ribominus kit, 30-40% rRNA in a RNA-Seq data is quite good. The new kits should perform better. In my own experience, percentage of rRNA in the data depends very much on the mapping parameters you use. I can map 30% if I allow 2 mismatch in 50bp reads but up to 50% if I use a min of 30bp extract matching. That's a huge difference.

Hi Mellissa
I have a worse situation than Nigel. My sequencing results showed 86% of rRNA ( I have to check the parameters used). The service provider showed me the mRNA agilent profile and it indicated 11% of rRNA contamination (poly A purification). How did it become 86%? What percentage of ribosomal RNA contamination is acceptable in the purified mRNA for 454 sequencing?

[pmiguel]

Hi Mellissa
I have a worse situation than Nigel. My sequencing results showed 86% of rRNA ( I have to check the parameters used). The service provider showed me the mRNA agilent profile and it indicated 11% of rRNA contamination (poly A purification). How did it become 86%? What percentage of ribosomal RNA contamination is acceptable in the purified mRNA for 454 sequencing?

Probably you were shown an Agilent Bioanalyzer RNA Nano chip run in "mRNA" mode. The analysis must presume that the ribosomal RNA is still intact. If any degradation happened along the way (or was present initially) then rRNA starts to look like just plain old RNA. Hence, not an accurate measurement of the amount of rRNA in the polyA purified fraction.

I don't think there is any standard as to what is "acceptable". Personally I have found that a single cycle of rRNA depletion does a poor job of removing all the rRNA. But yields from doing 2 cycles are terrible usually.

The amount of polyA in your sample will be organism and tissue specific. And this probably plays an enormous role in amount of rRNA left after purification. Actually I mentioned in another thread (some years ago?) a paper where they posed the following thought experiment. Imagine a sample that is 95% rRNA. Say you have a protocol that will remove 90% of rRNA from a sample during a single cycle of depletion. You might naively expect only 10% rRNA to remain after one cycle. But this is mistaken.

What you expect is that 90% of the rRNA is removed. What does that mean in real terms? Think of 100 RNA molecules -- 95 are rRNA and 5 are non-rRNA. 90% of 95 is about 86 molecules of rRNA molecules being removed. That leaves 9 (95-86) molecules of rRNA and the 5 non-rRNA molecules. So your sample is still 65% rRNA (9/16). That is the theoretical best case. So in cases where your polyA RNA makes up a very small fraction of the total RNA, either you do 2 cycles of depletion and your yields suck, or your do 1 cycle and sequence mostly ribosomal.

--
Phillip

[Lucilia]

Probably you were shown an Agilent Bioanalyzer RNA Nano chip run in "mRNA" mode. The analysis must presume that the ribosomal RNA is still intact. If any degradation happened along the way (or was present initially) then rRNA starts to look like just plain old RNA. Hence, not an accurate measurement of the amount of rRNA in the polyA purified fraction.

I don't think there is any standard as to what is "acceptable". Personally I have found that a single cycle of rRNA depletion does a poor job of removing all the rRNA. But yields from doing 2 cycles are terrible usually.

The amount of polyA in your sample will be organism and tissue specific. And this probably plays an enormous role in amount of rRNA left after purification. Actually I mentioned in another thread (some years ago?) a paper where they posed the following thought experiment. Imagine a sample that is 95% rRNA. Say you have a protocol that will remove 90% of rRNA from a sample during a single cycle of depletion. You might naively expect only 10% rRNA to remain after one cycle. But this is mistaken.

What you expect is that 90% of the rRNA is removed. What does that mean in real terms? Think of 100 RNA molecules -- 95 are rRNA and 5 are non-rRNA. 90% of 95 is about 86 molecules of rRNA molecules being removed. That leaves 9 (95-86) molecules of rRNA and the 5 non-rRNA molecules. So your sample is still 65% rRNA (9/16). That is the theoretical best case. So in cases where your polyA RNA makes up a very small fraction of the total RNA, either you do 2 cycles of depletion and your yields suck, or your do 1 cycle and sequence mostly ribosomal.

--
Phillip

Hi Phillip, Thanks for the considerations.
You are right! The profile is mRNA mode. I also checked the RIN number of the samples and it is bellow 8 (a number that I´ve read here would be indicated). The service provider assure that they purified twice and don´t know why the reason of the huge rRNA contamination in the results. Attached are the obtained profiles of total and mRNA - maybe someone has an insight of what is going on..
The service provider would like to try ribominus (what would cost me more) and I´m in doubt if it would solve the problem....

[pmiguel]

Well, the chips are pretty much useless for estimating rRNA % of a sample. All they show is the % of intact rRNA in the sample. If the rRNA degrades at any point, it will not be identifiable as such by the chip software.

Obviously ribo-depletion was not successful for you samples.

--
Phillip

[Lucilia]

So... That's what I' m thinking to do: purify more total RNA, considering the degradation problem, and perform another purification step ( invisorb) before mRNA isolation (I'm guessing that some impurities in the total RNA samples are disturbing the rRNA depletion). I hope it works...
Lucilia