Hi,

As you know, because of various levels of post-transcriptionnal and post-translational regulation, the amount of transcript does not always correlate with the amount of protein produces. In addition, there is a huge variability in this correlation for different genes, some will correlate better than others. Finally, please keep in mind that qRT-PCR as indicate by it's name is "quantitative", while western-blotting, even using a infrared or fluorescent detection method, will be at best semi-quantitative. So it's going to be hard to compare them directly too.

In conclusion, it all depends what your question is ? Ultimately if you care about the protein levels, and you have available antibodies, then you should look at the protein level. If no antibody are available, and you don't want to make them neither, then there is no other option than transcript levels. Just keep in mind that in specific situations, the correlation between RNA and protein levels will not be that straightforward.

Good luck,

best

Hi,

As you know, because of various levels of post-transcriptionnal and post-translational regulation, the amount of transcript does not always correlate with the amount of protein produces. In addition, there is a huge variability in this correlation for different genes, some will correlate better than others. Finally, please keep in mind that qRT-PCR as indicate by it's name is "quantitative", while western-blotting, even using a infrared or fluorescent detection method, will be at best semi-quantitative. So it's going to be hard to compare them directly too.

In conclusion, it all depends what your question is ? Ultimately if you care about the protein levels, and you have available antibodies, then you should look at the protein level. If no antibody are available, and you don't want to make them neither, then there is no other option than transcript levels. Just keep in mind that in specific situations, the correlation between RNA and protein levels will not be that straightforward.

Good luck,

best

I am agree with Ogla, its hard to correlate directly at protein level

You can compare relative levels. For some proteins the correlation is very good. For some- it is bad. And for some - it depends on what you are detecting and when. I believe Mathias Mann's group had a global correlation coefficient around 0.5, but details elude me at the moment.

I would also say that antibodies should be last resort except in the case of highly validated assays. Mass-spec approaches like SILAC would be much better though they have their own "gotchas".

Comparing qRT results with protein levels can be completely misleading because of the reasons mentioned above plus the fact that qRT-PCR calculations are based on the expression level of stable reference gene . At the most it can be used to compare two different situations/treatment/sources, where the relative values of qRT-PCR might be reflected in protein expression.

Not all genes/proteins are the same. I have generated some results and came across papers that show same pattern of gene and protein expression in the same sample. Some time there is a time difference where you collect the sample at certain time point when your treatment has effect on the gene expression but not on the protein level. In other cases if the protein you are looking for is regulated at the posttranslational level be ubiquitination or other way then they will not be correlated. The same applies if the mRNA expression undergo post-transcriptional modification.

The question Swati raised applies to my situation, especially considering that there are kits these days that are robust enough to allow for the extraction of RNA, DNA and protein from the same tissue (Biorad, Qiagene). In my case, I am trying to establish a trend in both protein level (by Elisa) and the mRNA transcripts (by qPCR), under the same reaction conditions and based on different treatments. My expectation is that the trend established in the qPCR should be replicated in the proteins (not necessarily quantitatively), across based on treatment. My sampling would be at the same time and in the same tissue, with all parameters standardized. If I understand Olga correctly, then that means there may still be differences in the pattern for mRNA expression compared to that for the protein levels?

If this happens, wouldn't that be a contradiction of the central dogma of Molecular Biology (DNA transcribed into mRNA, which is translated into proteins)? This is considering that I am monitoring the same protein, coded for by the same gene and both measured in the same tissue and under the same standardized conditions.

Thank you

Increaseed mRNA levels may not equal increased protein levels especially when microRNA is at play. MicroRNAs can inhibit translation of your transcripts, not a contradiction to the dogma but an extra level of regulation.

Correlation between mRNA and protein expression is variable and may not go straight like central dogma. One has to consider the temporal changes in gene and protein expression. Performing Microarray analysis is best way to start with.

As with all the answers above, there are lots of caveats about trying to make a correlation between mRNA & proteins. The degradation rate of the mRNA will also affect the relationship. If you need protein levels then it might be best to measure the protein, even if the method is only semi-quantitative. Timecourse experiments may help overcome that shortcoming.

I am agree with Olga but we can can quantify the level of expression by qPCR at mRNA level. In this context we just find out any house keeping gene in the host or vector and compare the expression level of taget gene with refrence gene i relative way or absolute way.

I feel that Ewan and Pinakin have understated their comments. Chasing after the direct correlation between transcript levels and protein abundances will lead you to great headaches because the dynamics of gene regulation, kinetics of protein degration let alone post-translational modifications as articulated here by, among others, Olga, are not linear. Nonetheless, you stand a better chance of understanding this correlation if you design a study that examines the temporal profiles that arise from the fluctuations of each mRNA and protein as a function of time or some other ordinal variable such as dose. See for instance Jayapal et. al., PLoS One (2008) v.3 e2097. This is my two cents and I credit Ewan and Pinakin for starting the conversation

To a very good extent

There is no correlation between protein levels and gene expression measured by QPCR. I remember an amazing talk of Pr. M Kirschner where he described how protein levels are, in some cases, different to measured transcript. So, the best way to analize a protein, and the relationship with gene expression, is by immunostaining.

I did western blotting and mRNA quantification via in situ hybridization and had similar results.

Here is the review about these very differences: doi:10.1038/nrg3185

Due to increased role of microRNAs (as Wah Chi Boon has indicated) and the fact that some non viral RNA sequences act as templates for cellular DNA synthesis, the central dogma is no longer central to understanding the direction of flow of genetic information.

So considering that we do not have micro array in our laboratory, which of the two (protein level or qPCR of mRNA) would anyone advise to focus on, to make a confident and justified argument? I only wanted to also quantify the protein level to corroborate the observation of up and down regulation of the transcripts. I may have been overly too ambitious.....right?

Hi Swati, the correlation can be from a few % to 100% depending on the levels of post-transcriptional and post-translational regulation. Only a Western of your protein will tell this.

Recent studies have demonstrated that mRNA abundance can explain only about 27% of the variance whereas the coding sequence, 5’UTR and 3’UTR sequences account for almost 40% of variance in the protein abundance data (Vogel et al., 2010, Mol Systems Biol., 6:400). It has become evident that the relationship between mRNA and protein is quite complex (Licatalosi & Darnell, 2010, Nature Reviews, 11:75; Rogers et al., 2008, Bioinformatics, 24:2894; Vogel et al., 2010, Mol Systems Biol., 6:400). There are a multitude of factors, in addition to the regulation at post-transcriptional and post-translational levels, including differences in the methods used to obtain mRNA and protein abundance data that may influence the lack of correlation between the mRNA and protein abundance data (Waters et al., 2006, Briefings Funct. Genomic & Proteomics, 5:261; Vogel et al., 2010, Mol Systems Biol., 6:400). These observations re-enforce the suggestions that the use of the data at the multiple levels, such as mRNA and protein abundance, miRNA abundance, sequence variations etc., would improve our ability to understand mechanisms involved in the regulation of expression of these genes (at mRNA and protein level).

The term "correlate" can have two entirely different meanings in this context, and the answer is entirely different for each:

1) The correlation between mRNA and protein level of different genes in the same sample. This is very poor and there is nothing surprising or mysterious about it. Given the differences in translational efficiency of mRNAs, in post-translational processing, targeting and degradation rates of different proteins (not to mention the very poor precision in quantifying most proteins), it is impressive that even 27% of the protein-level variance can be explained by mRNA abundance. However, this is an uninteresting question, rarely asked by researchers.

2) The correlation that most of us are interested in, is how mRNA level predicts protein expression of the same gene in different samples (e.g. the given gene product of interest in treated vs. untreated cells, CD4 vs. CD8 lymphocytes, brain vs. liver, etc.). This correlation is much better and much of the discrepancies are likely due to the uncertainty in our methodology for protein quantification.

The technical difficulties in quantifying proteins should not be underestimated. A recent search for genetic loci determining gene expression levels of both RNA and protein among recombinant mouse strains using large-scale proteomics, found large numbers of eQTLs both at the RNA and protein level. There was very little overlap (not surprisingly, per my point #1). More interestingly, phenotypic trait loci aligned much better with the mRNA eQTLs than with the protein ones. Unless we are willing to accept that mRNAs have biological effects unrelated to encoding protein, the conclusion is inevitable: protein quantificaiton technology has a long way to go to match our ability to quantify nucleic acids.

Forgot to cite the reference for the proteomics paper. It is Ghazalpour et al., 2011

http://www.ncbi.nlm.nih.gov/pubmed?term=39.%09Ghazalpour%20A%2C%20Bennett%20B%2C%20Petyuk

In a first, simple assumption you can correlate your qPCR with protein levels.

If you can express and purify your protein, you can use this as a standad in a Westernblot.

If you have an enzymatic activity or physological function and an assay for this enzyme or function, you can quantify the protein expression.

Also binding of radioactive labeled agonist or antagonist can be used for protein quantification.

Theo

Hello, It is difficult to correlate the amount of transcrip to the protein level, since the transcription and translation is regulated. So I advice to always measure your protein level after you have check the mRNA level. Hope I have answered to your qestion. Thanks. Simon

Clearly it is case by case.

But I would like to point to this great article written by Schwanhausser et al. (Nature 473: 337-342). Using SILAC and RNAseq, they compared the turn over rates of mRNA and proteinprote, and subsequent effect on mRNA and protein abundance. They estimate that they could predict around 40 % of protein abundance from mRNA abundance. So, mRNA abundance has some importance. Must read!

yes you are right mRNA has some importance but you cannot correlate it 100%. you can get variation in your mRNA and protein expression. Its not necessary that if mRNA is expressed more then protein will also express more.

But we can go for correlation at protein level to give relevance to the results obtained at mRNA level. This will add more value to the work.

Now going by the richness of the conversation which this interesting topic has generated, and after reading most of the articles referred by different commentators; I am inclined to assume that;

1. It may be on a case by case basis like Madoka suggested, since it looks like experts in proteomics are advancing arguments based on their experiences while genomics experts are doing the same.

2. The meaning of correlation in my case is what Constantin explained in the 2nd context; and it looks like mRNA abundance remains the most reliable predictor of effects on gene regulation.

However, I would still check the trend of the protein abundance and find out if the trend correlates with what mRNA abundance suggests.

Thank you all for the wonderful contributions.

In qualitative terms the mRNA and protein levels correlate fairly well. i.e. if under two conditions the mRNA is over expressed, the protein will be increased. However the fold change and the timing of the increase or decreased may not. These never correlates as 1) lag in time of transcription vs translation 2) Sensitivity of methods (personally I believe the latter is more likely), I am yet to see very neat reliable papers where they have shown that although under experiential conditions mRNA has increased but protein has decreased. Mostly those who have reported are experimental artifacts like lower fold change in mRNA.

Hi Paul,

I would like to know when you say lack of correlation is it in terms of fold change or in terms of overexpression or underexpression. I mean, if a particular mRNA is overexpressed in my samples as compared to controls, will the protein also increase or not ? I am at this point not worried about the fold change.

Waiting for your response

Hi,

Following on from previous answers, the work in our lab as revealed less than a 5% correlation between mRNA and protein levels. Such correlations can vary and there are an increasing number of research articles highlighting this.

Sorry paul. I yet did not get what do u mean by correlation ??

Hi Deepak. Over a developmental time course the transcript profile of a set of genes matched thier protein abundance profiles in less than 5% of the total genes studied. Therefore 95% did not match. Hope that's a bit clearer.

Yes that's clear now. However I am not surprised on this As 1) the technologies that we use to compare the two are different and it will be the dynamic range and the sensitivity of the two methods used. For e.g we do not get absolute match in real time vs microarrays of the same gene. 2) There is always a lag in therms of transcription vs translation 3) Also we need to account for the degradation of misfolded proteins during synthesis. So for protein vs mRNA it is expected.

However I am keen to know that in any given experiemnt, for e.g over a developmental time frame, what percent of mRNA that ever got up-regulated or down regulated by lets say 5 or 10 fold never showed any change at protein level or showed a reverse trend at a subsequent time point.

Hello Deepak,

your last question is exactly my point.

Is it likely that a 5-10 fold up-regulation in the mRNA would also show me higher concentration of the protein (not necessarily 5-10 fold)? If the answer is YES, I am happy. I can then conveniently discuss my observations using the protein concentration as additional evidence to boost confidence in my results.

Could that translate into being overly too optimistic? What else would I need to show?

For most experiments, people will be happy with up vs. down in RNA reflecting a change in protein in the same direction. If you wish to know the exact quantitative effect on the protein, you must pay attention to your method of quantifying the protein, a huge issue. RNAseq can now give us RNA changes with digital precision, regardless of what gene you are looking at. A method of comparable precision just doesn't exist for the vast majority of proteins. Precise immunometric assays have been developed for a tiny fraction of all proteins in a handful of species. Doing so for each individual protein is time- and money-consuming. Not for hypothesis-free, proteome-wide quantification. You need to be lucky enough for an assay for your protein to exist already, or be prepared to spend months (years!) developing, troubleshooting and optimizing one yourself.

But, take heart, quantitative proteomics is improving. Selected Reaction Monitoring (SRM) allows you to quantify any protein you choose with precision that is beginning to approach that of RNA methodologies. It was declared Method of the Year by Nature Methods ( http://www.nature.com/nmeth/focus/moy2012/index.html ). Still, it is not high-throughput in the order of magnitude that RNAseq is. You need to decide which proteins you like and have peptides synthesized for each.

As I have written before, when one expects apples to reflect apples and oranges to reflect oranges (i.e compare THE SAME protein between, e.g., control and experimentally manipulated cells), RNA is an excellent substitute for quantifying protein (except in the, unusual I would imagine, situation in which your experiment changes transcription and translation - turnover in opposite directions).

Most of the literature cited in response to the question sounds disappointing because it shows that apples are not very good at reflecting oranges: Relative RNA levels of DIFFERENT genes are very imperfect predictors of the protein-product's level in the cell. If you are surprised or disappointed by this, you should read the literature that shows order-of-magnitude differences in translational efficiency of different RNAs and in cellular half-lives of different proteins.

Very clear Constantin. What more can I add? This painstakingly lengthy explanation has really cleared all the clogs in my thinking. Thank you sir!

Awesome Constantin .... to the point and very precise explanation. People are expecting apples to reflect oranges which will NOT and the blame game begins !! Thanks for the clarifications !!

can mass spec be used to quantitate a large number of proteins in a sample? Or their relative abundances?