Analysing the transcriptome (RNA-Seq, microarrays, qPCR etc) is probably the most widely used technology to assess gene expression and dynamic cellular processes. The results are then extrapolated (functional analyses, GO etc) to infer biological cellular consequences. RNAs are really only the middle-man. Presence of an mRNA doesn't necessarily mean that it will become a protein. It could be degraded by microRNAs etc. before translation. It is eventually the protein that are actually causal to a biological change.

So, my question is why not just go directly to the proteins. Why not just extract the proteins and sequence them, quantify them, analyse and draw conclusions? What are the compelling arguments to study the transcriptome to understand biological processes?

  • Another thing that is not directly related to your cellular process premise, but is nonetheless a positive of RNA-Seq: it also gives you coding and protein sequences at a cost far less than proteomics or genome sequencing, so it's useful for phylogenetics/phylogenomics – NatWH Sep 13 at 17:10
  • 1
    @NatWH, A good point. We also need sequence databases to search mass spectrometry-based proteomics data. In the absence of genomic data many groups use RNA-seq data to build this sequence database so they can then use MS. – Michael_A Sep 14 at 20:16
up vote 4 down vote accepted

Q: Why not just extract the proteins…

A: It’s not just a question of extracting the proteins, you would need to separate them and then isolate each of them. There is currently no practical way to do this for say, 10,000 proteins, which in any case may have multiple forms. The beauty of the RNAseq method is that it does not require physical separation or isolation of transcripts.

An approach more likely to be useful is to identify and quantify the proteins in situ, but this is still some way off having the comprehensiveness and sensitivity required.

Q: …sequence them…

A: Protein sequencing is slow and difficult. It cannot be applied on a mass scale, unlike sequencing of DNA copies of RNA transcripts. Other methods of identification based on physical behaviour (migration or elution position) are more promising, and the direction that is likely to be followed.

Q: …quantify them…

A: Even that is difficult, when you are eluting proteins from gels, for example. In contast, RNAseq methodology just counts transcripts. (But the likely direction this will go is quantitation by peak height, without separation or isolation, as with metabolomics, or by the intensity of some signal from the protein immobilized in the gel, staining of something more sophisticated.)

So, in general, you do protein analysis on particular individual proteins of interest to you, whereas you can do RNAseq analysis on all the transcripts in a biological sample, and then go fishing (among the results, and afterwards on the lake if that’s your thing).

  • There are no methods that will separate, say, 10,000 proteins gel electrophoresis can easily separate this number of proteins. Liquid chromatography can too. Detecting intact proteins is currently low-throughput but LC-MS on peptides can provide quantitative data for 1000's of proteins, the ability to detect PTMs is a feature not a bug. – Michael_A Sep 14 at 21:05
  • @Michael_A — technically you may be right (or partly so), but in the context of this question from a naive list member I think your comment is more likely to confuse than enlighten. I'll rewrite my answer in a way that deals with your point without diluting the explanation. But it will have to wait another day. – David Sep 17 at 20:51
  • @Michael_A — I have modified my answer to address the points you made. At some time I may modify it further with a consideration of how technological breakthroughs influence science, and the references to the current state of proteomics. – David Sep 18 at 13:01

While you are correct that 'the end output' is always protein and that functional analysis on the mRNA level can ignore things translational control, there good reasons to analyse the transcriptome:

1) Like you said a lot can happen with mRNA: it can be degraded, translated or also be repressed for a time until it's activated again. Most of these control processes have pretty strong effect on the 'final protein output' but are much easier to study on the mRNA/transcriptome level (since that is where they happen).

2) Analysing the transcriptome on a genome wide level is technically relatively easy: Sequencing technology is pretty advanced by now and there are many different protocols established to look not only at expression levels of individual mRNAs but also other things like stability, processing (i.e. splicing or polyadenylation) and also translation of mRNA. All of this can also be done with relatively low material input.
Protein analysis on the other requires the use of mass-spectrometry which is not necessarily as powerful or sensitive as sequencing.

  • 2
    I think it's useful to bear in mind that the transcriptome does consist of non-protein coding transcripts such as miRNA, and that while protein is largely regulated, the end product is not always protein. ncbi.nlm.nih.gov/pmc/articles/PMC4177046 – CKM Sep 13 at 1:41

Why not just extract the proteins and sequence them, quantify them, analyse and draw conclusions?

Because proteomics is really hard. To provide one example, there is no protein equivalent of PCR. That means there is no simple way to select and amplify a particular protein from a complex mixture. The invention of PCR was central to the rapid advances in genomics over the last 40 years.

Mass spectrometry is the primary tool for high throughput proteomics and mass spec analyses are still very difficult to run, and even more difficult to interpret.

What are the compelling arguments to study the transcriptome to understand biological processes?

There are compelling arguments to study the transcriptome but they depend on the research question. Research that focuses on the activation of transcription factors should use transcriptome data as should research into RNA processing events (splicing, pre-mRNA processing, stability). You also want transcript data when studying RNA transcripts that don't encode proteins.

I don't think there are compelling biological arguments for studying the transcriptome when you want to know about the expression of protein-coding genes in tissues or cells. You should focus on proteins as they are the final step in the expression of protein-coding genes. The correspondence between the levels of protein-coding transcripts and their protein can be excellent but may also be very poor. Transcript data can't segregate these groups. In global estimates, transcript levels only explain around half of the variation in protein levels. I think that level of inaccuracy unacceptable when there's a viable alternative.

Why not just extract the proteins and quantify them?

There are physical and historic limitations to global quantification of proteins. I've compiled a couple limitations that I consider to be most important. Mass spectrometry (MS) is the current method of choice for high-throughput protein quantification. Methods such as Edman sequencing or immunological detection have much lower through-put.

There are physical limitations that currently stop MS from performing quantitative global measurements using intact proteins. That's very unfortunate. Global measurements are currently performed by quantifying peptides from digested proteins using a liquid chromatograpy system to feed the MS with peptides (LC-MS). The peptides are identified and then assigned to a protein or group of proteins. The end results are reliable and most will resolve to individual protein identifications. Some peptides will only resolve to identify a group of proteins. This is a minor limitation in most instances.

Making global measurements of protein quantities has only become possible in the last decade while global measurements have only become routine with the advent of ultra-high performance LC paired to the newest mass spectrometers. Many academic researchers will have access to this technology via their local mass spectrometry facility but most won't realise it. This newness is a big part of the reason it's less common.

Summary

I think LC-MS quantification of proteins will be used in place of transcript data more often over the next couple of years. This is due to the ability of as current LC-MS instruments provide better answers in biological systems that require protein-level information. Reviewers will start to demand it. I think within the next decade high-performance LC-MS could become cheap enough to even replace lower throughput immunological techniques such as immunoblotting and immunofluorescence.

There is one aspect that has not been detailed: a transcriptome is way cheaper than a proteome. Analyse one single protein in LC-MS/MS can cost as much as a transcriptome. And if you belong to those small underfunded groups that work on non-model organisms which are really interesting but not really useful for curing cancer, then you have no choice.

Your Answer

 
discard

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.