Hi,
I have been doing Reverse Transcriptase qPCR on some liver mRNA samples. I am targeting liver inflammation genes like Col1a1, alpha SMA, IL6, IL1B and etc... Even though mRNA concentrations are good and primers are working, some samples are not working (difference between two samples are higher than 0.5) for a few runs in a row. After I started using triplicates of repeat samples instead of duplicates, I am getting more successful results. For analysis, I am considering the closest two and taking the average of these two samples. Does it sound appropriate in terms of data analysis ? Has anyone been doing this ?
If you do triplicate runs, use the average of the triplicates. Making a general policy of discarding replicate data is a bad idea.
In practice, this comes with some (user-defined) case-by-case cut-offs.
If I got 22.2, 22.3 and 26.8 for triplicates, I would look at all the other samples, and if they were all hovering around the 21-23 ballpark, I would absolutely drop the 26.8 as a "John did some crap pipetting" error. If I got 22.2, 22.3 and 23.1, I would be more inclined to use all three, or repeat the run.
The issue really comes down to how much you can trust your data: with variable duplicates, either result could be the valid one. With variable triplicates, at least you usually have two that agree and one that doesn't. And if that isn't the case (the dreaded 22.1, 23.2, 24.5 result), then you re-run those samples (alongside trustworthy data so you can back-calibrate).
I would always recommend using triplicate wells, reserving duplicate wells only for genes that you have run countless times and found to be incredibly consistent. When you consider the economics, duplicate wells you have to repeat end up costing more than triplicates you don't.
I know others may argue in favour of a more rigid approach, but it's molecular biology: an innately messy process. Be informed by the actual scenario, rather than hard rules. A gene that gives 22.3, 22.5, 22.1 in untreated samples and 26.7, 25.4, 26.9 in treated samples is a gene that is clearly dropping in expression, whether you decide to trust that 25.4 or not, and (equally) the extent to which it is dropping is pretty similar either way. The biological readout is clear.
Oguz Ozler I agree with John Hildyard In addition, I would suggest you should go through The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments (DOI: 10.1373/clinchem.2008.112797). For qPCR generally, data collected for mRNA prepared from samples collected at two Different points (biological replicate) and each experiment performed in triplicate (technical replicates)should be use for analysis and interpretation.
I largely agree with John Hildyard , and like to make one addition:
Running tech reps is ok if the technical variance is large (larger than the biological variance) or if you need a precise estimate for one particular specimen/individual. Actually, the assay should be sufficiently optimized and validated before running the actual qPCRs so that high technical variance shouldnot be a (major) problem, and generating precise estimates for a single infividual is rarely the aim in research. Thus, it it usually better to measure as many different biol reps as possible. If you can substitute a tech rep by a biol rep, then do this. It's usually better to spend all the money and time on more biol reps than on tech reps.
If you measure a couple of different genes, you don't need many samples to fill a plate, and running tech reps will likely require running several plates, what will make plate calibration neccesary and in any case adds another source of variance. This can be a bad deal.
- Badges
- Science method