Targeted quantitative RNA-sequencing with interlibrary concordance and reduced cost

[thomasblomquist]

http://www.plosone.org/article/info%...l.pone.0079120

Hello SEQanswers Forum. Long time follower, occasional thread responder. I have been in the qPCR world for some years now. One need in the RNA-seq world I've noticed is the need for the assays to be, 1) targeted, 2) quantitative, 3) reproducible in quantitation, and 4) at a drastically reduced cost.

The work in the study presented above represents our approach to make targeted RNA-sequencing highly quantitative, reproducible, at a drastically reduced need for sequencing depth for large dynamic ranges of templates.

Hope you all enjoy. Any questions as to how this study can be applied to your research needs, I would be more than happy to address.

-Tom Blomquist
University of Toledo

[KristenC]

Interesting!

Hi Tom,

I really enjoyed reading your paper! You touch on a couple of very important issues present in targeted RNA-sequencing.
I am currently working on a targeted RNA-sequencing assay, and am exploring possibilities for using reference genes to normalize the counts. Since your assay does not even need reference genes, I was wondering if you at some point looked into it and maybe dismissed it for certain reasons? Could you comment on why you moved away from possible adaptations of the more traditional RNA-seq approaches?

Kind regards,

Kristen

[thomasblomquist]

Hi Tom,

I really enjoyed reading your paper! You touch on a couple of very important issues present in targeted RNA-sequencing.
I am currently working on a targeted RNA-sequencing assay, and am exploring possibilities for using reference genes to normalize the counts. Since your assay does not even need reference genes, I was wondering if you at some point looked into it and maybe dismissed it for certain reasons? Could you comment on why you moved away from possible adaptations of the more traditional RNA-seq approaches?

Kind regards,

Kristen

Hi Kristen,

You are correct in your statement that we can provide back either median normalized abundance, reference gene normalized abundance, or absolute copies based on the internal standard input amount (i.e. cDNA or gDNA copies per uL input). The choice for median normalized abundance for comparison of samples C and D was purely so that the intercept was near 0 (i.e. aesthetic reasons). For the interplatform concordance assessment, it made sense to perform a normalization of our data to the median measurement because TaqMan as well as Illumina RNA seq data was without denominator units as well. It was purely for aesthetics in some instances.

Another item to consider was the fact that different reference samples A versus B versus C, D, etc., as seen in the clinic, will have different RNA content amounts, different cell counts, etc. And as biologists and clinicians, we are making a best attempt to compare between samples in our studies and in the clinical world. The choice of a house-keeping gene as a reference normalizer, a set of reference genes for geometric normalization, or whether you decide to do a median abundance normalization, or if you decide to normalize on a negative binomial distribution fit (which may ultimately be the best for RNA-seq studies), can seem to be arbitrary. I took a look at some of your previous posts, and I think your other thoughts on trying to find a somewhat invariant set of expressed targets is probably the best. Try, if you can, to identify the targets that are invariant relative to cell-count input into the assay. That way your normalized measurements down the line are relative ~~~to cell input. Normalizing to RNA input amounts, although commonly employed, has a lot of poor assumptions, and largely is dependant on rRNA level which can vary WIDELY!!! Bustin on the A-Z of qPCR has a few good sections on this topic of normalization between samples for transcript measurement.

But you are correct, our data follows a very closely a Poisson sampling distribution in measurement variance. And our experience in the qPCR world has demonstrated that with extreme limiting dilution to stochastic sampling range it follows a Digital PCR phenomenon, which you can also base your calculations on. So, theoretically, if you control for RT efficiency, you will have an absolute measure of RNA transcripts. This is important for clinical assays for viral load, etc.

Think pan-viral assay measurement with reporting back to a clinician absolute copy numbers, and because our method pretty much eliminates the need for deep sequencing, you could multiplex the heck out of an ion torrent chip and get results back ASAP!!! It is very good stuff.

Thank you again for your interest. Best of luck on your research.

-Tom Blomquist

[thomasblomquist]

Hi Tom,

I really enjoyed reading your paper! You touch on a couple of very important issues present in targeted RNA-sequencing.
I am currently working on a targeted RNA-sequencing assay, and am exploring possibilities for using reference genes to normalize the counts. Since your assay does not even need reference genes, I was wondering if you at some point looked into it and maybe dismissed it for certain reasons? Could you comment on why you moved away from possible adaptations of the more traditional RNA-seq approaches?

Kind regards,

Kristen

As to comment on why we moved away from traditional RNA-seq. Some preliminary data from colleagues using unsupervised biomarker identification in RNA seq libraries from different clinical samples was differentiating based on the DAY or the BATCH the RNA-seq library was prepped with!!! EEK!!!

This is not surprising as ligation efficiency is highly sequence dependent, and combined with fragmentation efficiency that can be highly dependent on protocols, we opted to develop a more focused and hopefully standardized approach that can answer focused hypotheses in our clinical samples. We work with small bronchial brushings with only 10-100 ng of RNA input, and we needed to be able to have reproducible and robust measurement on a couple hundred assay targets.

There are other reasons as well... But those were the big ones.

-Tom