spike-in controls

[phillyjoeremarkable]

I have just ordered the following spike-in controls for future profiling expts.

AUGUACACGGCGAAGUCCCGCAACGCGA
AUGUACACGGAUGGCACCCGCAACGCGA
AUGUACACGGGCAAUGCCCGCAACGCGA

They are based on the Ingolia 28mer marker (AUGUACACGGAGUCGACCCGCAACGCGA) and do not align to any known sequence so could be used as controls for any profiling expt regardless of organism. I have kept the 5' and 3' nucleotides and the %GC content the same within this set to reduce linker ligation or PCR bias. My intention is to use these at different concentrations within each library.

[chewgl]

Single spike-in controls have been tried, with limited success. The problem is that you only have spike-ins of a single length, but you cannot control the length distribution of nucleotide fragments that you cut out of the gel. Therefore, your spike-ins are unlikely to be a quantitative measure of how large your ribosome profiling library is.

What has been tried (with some success) is spiking in entire libraries of another organism, e.g. when doing ribosome profiling in yeast, spike-in a known quantity of a HeLa cell library. One would have to cut a large enough gel band to encompass the known size distribution of ribosome protected fragments from both organisms (since mammalian ribosome protected fragments seem to be a tad larger than yeast), but it should be possible.

[First citizen of GWIPS]

The cells of another organism is an interesting idea indeed. However, I am not sure it would work accurately if rRNA depletion is attempted. Suppose I have cells from an organism A under control and stress condition and I spyke-in it with cells of an organism B. In control I have footprints, rRNA spike-in footprints in a ratio of x:y:z (x+y+z=const due to normalization). Under stress conditions the ratio gets to 0.1x:y:z ((x+y+z=const) due to global 10-fold suppression of organism A translation. Here I can figure out the level of suppression under assumption that y (level of rRNA is not changing y=const). However, once I carry out rRNA depletion, y become unknown because the amount of depleted rRNA may not be the same in stress and control.

I wonder if it would make sense to use 5S rRNA as a spyke-in. Say in addition to cutting out a gel band corresponding to ~30 nt cut out ~120 and mix it into sequencing at a cretain ratio. Assuming that 5S rRNA concentration is unaffected, the ration of footprints to 5S may permit to estimate the level of suppression. Of course, this would work only under short term stress, since under long-term stress rRNA levels may change, but this would effect the above spyke-in as well anyway. Any thoughts?

[plutopluto]

Wow, just found out about this site from a recent Nature Communication paper. What a great job for the "ribosome profiling" community.

I'd like to chime in on the spike-in control discussion, as in a lot of cases we want to look at datasets where the global translational rate is changed, therefore, I am afraid just by normalizing by total library sizes may change "mildly down regulated genes" into "upregulated genes".

For the 5S rRNA method, would the difference of the library size (supposedly ~180bp for RPFs, and ~270bp of 5S) affect the sequencing if you mix them? Please forgive me if this is a naive question, as I am more a wet lab researcher than sequencing/bioinformatics guru.

And for the differences in rRNA depletion, it may not matter as long as the x:z ratio keeps the same, since no matter how much y is left, you are going to discard them in the downstream analysis? People use spike-ins for regular RNA-seq, and whether you use Ribozero or oligodT beads to enrich mRNAs doesn't seem to affect the intent, but maybe it's because the left-over rRNA reads are very limited there.

My anxiety is at which step do we spike in the controls. You want to do this at a step where you are quite certain of the equal loading of the mRNAs (by cell number?), which is not easy to achieve with certain samples, and also after the RNase digestion, ultracentrifuge etc. But maybe one can accept equal loading of total RNA amount before rRNA depletion at some point due to the stability of rRNA amount.

And in this case, maybe spike-in equal amount of random fragmentized distant species' mRNAs would be a good choice? I'd prefer to have just one exogenous mRNA such as luciferase, but I am not sure how unique it's sequences fragments would be.

Also since we are cutting between 28-34nt based on Ingolia's protocol, may be just a mixture of the two markers could serve the function if we just zoom in on that region for downstream analysis?

[First citizen of GWIPS]

Thank you for the good words Probably you are right. if normalization is done after removal of rRNA, it should not affect x:z ratio.

"My anxiety is at which step do we spike in the controls." - I think the suggestion was to mix cells before the lysis. When else could this be done?

[plutopluto]

So would you start with same cell number between each conditions or matching with the same A260 value or something?

(19-Nov-2014 04:44 PM)First citizen of GWIPS Wrote:  Thank you for the good words Probably you are right. if normalization is done after removal of rRNA, it should not affect x:z ratio.

"My anxiety is at which step do we spike in the controls." - I think the suggestion was to mix cells before the lysis. When else could this be done?

[First citizen of GWIPS]

I think the number of cells.

[djrizzz]

"My anxiety is at which step do we spike in the controls." - I think the suggestion was to mix cells before the lysis. When else could this be done?