[snetmcom]

I was able to get an ambion rep to explain that adapter B is not as well suited as adapter mix A. It has nothing to do with RNAse III. Avoid adapter mix B unless you need to check the 5' tags.

[pmiguel]

I was able to get an ambion rep to explain that adapter B is not as well suited as adapter mix A. It has nothing to do with RNAse III. Avoid adapter mix B unless you need to check the 5' tags.

Hmmm, "not as well suited" sounds more like a sound byte than an explanation. Here let me translate for you:

AMBION: "Do not use adapter mix B for an undisclosed reason."

There, see how my translation doesn't leave the subtle implication that an explanation has been given? It might even pique your interest.

But even better:

SEQANSWERS: "Do not use adapter mix B because it will result in a much higher mismatch rate in the first 6 bases of your reads."

And, if you want to dig a little deeper into the thread you might summarize as follows:

"The random pentamer ligation splints anneal somewhat promiscuously to the ends of your RNAs. One splint is incorporated into your amplicon during cDNA synthesis, so that stretch of sequence will less faithfully represent the original RNA than the rest of the cDNA. When using adapter mix B the splint-derived sequence is adjacent to the P1 adaptor and will be sequenced. When using adapter mix A the splint-derived sequence is forced to the far end of your insert. There it may not be sequenced at all if you are doing SWTAK and your inserts are longer than 55 bases. With SREK, presumably you would be stuck sequencing splint whether you use adapter mix A or B."

Nothing to do with RNAse III? Yep, I agree. Mid-thread Pascal switched topics to focus on the lack of polyA in his SWTAK data set. RNAse III bias was a possible explanation.

--
Phillip

[scooter]

You are exactly correct. The errors you are observing are due to mismatch priming by reverse transcriptase. Typically the random primers are hexamers, so you would expect the error rate to be higher in the final 6 bases of your reads and the rate can be quite high.

A Nature methods pub from the Grimmond lab (cloonan et al) has a supplementary figure that describes this phenomenon.

[pmiguel]

You are exactly correct. The errors you are observing are due to mismatch priming by reverse transcriptase. Typically the random primers are hexamers, so you would expect the error rate to be higher in the final 6 bases of your reads and the rate can be quite high.

A Nature methods pub from the Grimmond lab (cloonan et al) has a supplementary figure that describes this phenomenon.

The SOLiD Whole Transcriptome Analysis Kit (SWTAK) and the closely related Small RNA Expression Kit (SREK), effectively prime reverse transcription off a pentamer. That said, I think this pentamer would have a six base error signature in color space because of dual base encoding.

I write "effectively" because both kits actually use a ligase (I presume T4 RNA ligase 2) to ligate RNA to 3' and 5' adaptors. Said adaptors having random 5 base 5' and 3' (respectively) single stranded overhangs.

My guess is that this ligation-mediated methodology somewhat exacerbates the higher error rate you might expect at the very terminus of a random primed reverse transcription first or second strand.

--
Phillip

[Nitrogen-DNE-sulfer]

I was told it was a hexamer but one of the colors accounts for the vector to insert color. Not certain that clarifies anything but it may support the depletion story a bit more (ie 4096 as opposed to 1024).

One can call it a ligation mismatch issue but that generalizes the problem such that one questions why it doesnt happen on the other end of the molecule with the P1 adaptation (in Mix A).
It 's really the use of the lower strand hexamer in the P2 adaptor to act as the 3' primer for cDNA synthesis. If there are any bubbles on this guide adaptor they then become truth of the template after cDNA extension. The P1 adaptor (In MixA) bottom strand is temporary and blow away once the cDNA extension is done. One of the users at the European user meeting had nice definitive data on this (need to confirm who).

As for polyAs missing, depletion is one possibility. I like Phillips theory on the degradation as well. The other thing to consider is the length of the poly A's. If RNAse III doesnt cut in this sequence then they polyAs will be largely uncut and thus longer and there is clearly a selection against longer RNA throughout various steps in the protocol.

I hear the issue is going be addressed in a new kit in spring 2010 but havent seen the exact details.

[scooter]

The issue is indeed caused by the lower strand of the 3' adaptor incorporating some mismatch bases due to the promiscuity of reverse transcriptase. In the SREK procedure the amount of error is significantly LESS than that seen with true random priming and this is because with SREK a ligation step is also involved. If there are any mismatches in the 6 N's for SREK than ligation is less efficient and the probability of inclusion into the final library is reduced.

I understand that Sean Grimmond's group looked into this issue and determined that SREK had ~1/3 of the error rate under the hexamer compared to a typical random priming protocol.

Ambion is soon coming out with a new version of SREK that will correct this issue. I heard it was scheduled for March launch, but who knows for sure......

[pmiguel]

If there are any mismatches in the 6 N's for SREK than ligation is less efficient and the probability of inclusion into the final library is reduced.

5 N's. ABI doesn't specify the lengths of the random primers, but every diagram I have ever seen of the process depicts 5 N's. (See attachment.)

--
Phillip

[hingamp]

5 N's. ABI doesn't specify the lengths of the random primers, but every diagram I have ever seen of the process depicts 5 N's. (See attachment.)

--
Phillip

True that diagrams suggest pentamers, but the error rate analysis we did suggests hexamers (fig from thread begining brought forward)? We're seing ABI people thursday about this adaptor issue and the mysterious polyA tail disappearance act. I'll keep you posted.

[pmiguel]

True that diagrams suggest pentamers, but the error rate analysis we did suggests hexamers (fig from thread begining brought forward)? We're seing ABI people thursday about this adaptor issue and the mysterious polyA tail disappearance act. I'll keep you posted.

The pentamers are in sequence space, but your diagram, I presume, is in (dual encoded) base space. Those 5 sequence space bases "project" onto 6 color space bases. That is, the 5th base of the pentamer would be interrogated by ligations that produce color bases 5 and 6.

So your diagram is consistent with random pentamers, no?

--
Phillip

[hingamp]

Short answers from meeting with ABI people:
-high error rate under "mix B" first 6 nucleotides is expected and known (just not documented). New kit to released soon has no "mix B" approach, which is one way to solve issue...
-missing polyA tails: unexpected, ABI will look into it.

We also mentioned an observation that could hold the key if confirmed: we see very significant read coverage correlation with underlying DNA GC content, both in SWTAK and standard genomic fragment analysis. This has been observed by ABI (Genome Res. 2009 Sep;19(9):1527-41. Epub 2009 Jun 22.) but might become an issue with AT rich regions or genomes, or polyA tails...

[pmiguel]

Short answers from meeting with ABI people:
-high error rate under "mix B" first 6 nucleotides is expected and known (just not documented). New kit to released soon has no "mix B" approach, which is one way to solve issue...
-missing polyA tails: unexpected, ABI will look into it.

We also mentioned an observation that could hold the key if confirmed: we see huge re

???
Looks like your post was truncated?

--
Phillip

[hingamp]

???
Looks like your post was truncated?

--
Phillip

Sorry if it looks like our sentence suffered the same fate as our polyA tails - my finger just slipped on the enter key... Here's the full post:

Short answers from meeting with ABI people:
-high error rate under "mix B" first 6 nucleotides is expected and known (just not documented). New kit to released soon has no "mix B" approach, which is one way to solve issue...
-missing polyA tails: unexpected, ABI will look into it.

We also mentioned an observation that could hold the key if confirmed: we see very significant read coverage correlation with underlying DNA GC content, both in SWTAK and standard genomic fragment analysis (orders of magnitude). A milder version of this has been observed by ABI (Genome Res. 2009 Sep;19(9):1527-41. Epub 2009 Jun 22.) but might become an issue with AT rich regions or genomes, or polyA tails...

[pmiguel]

The issue is indeed caused by the lower strand of the 3' adaptor incorporating some mismatch bases due to the promiscuity of reverse transcriptase. In the SREK procedure the amount of error is significantly LESS than that seen with true random priming and this is because with SREK a ligation step is also involved. If there are any mismatches in the 6 N's for SREK than ligation is less efficient and the probability of inclusion into the final library is reduced.

I understand that Sean Grimmond's group looked into this issue and determined that SREK had ~1/3 of the error rate under the hexamer compared to a typical random priming protocol.

Ambion is soon coming out with a new version of SREK that will correct this issue. I heard it was scheduled for March launch, but who knows for sure......

I just looked at the manual for the new kit. Elegant.

After ligation of the adaptors the "NNNNN" terminated 3' oligo is melted away from the adaptor oligo terminated RNA fragment and the RT rxn mix (prior to RT enzyme addition) allowed to cool in the presence of a (an excess, one presumes) of "SOLiD™ RT Primer". Sequence not specified, but no reason it would terminate with the 5 problematic N's at this juncture...

This is especially relevant with the new paired end chemistry coming on line with v4.

--
Phillip