I've run RNAseq and qPCR on a set of genes, and while the log2 expression values are consistent between the tests for most of the genes in the set, there are a handful that appear to be unregulated according to the RNAseq results and down regulated according to the qPCR results (and vice versa). Is there any possible reason that could explain this, other than just human error?
In NGS, read counts have somehow to be mapped to genes. Reads are not always matching to only a unique gene, so it partially depends on the mapping algorithm how many "reads" are attributed to a gene.
Further, genes may be expressed in different variants, and the presence or abscence of an exon may impact the number of reads attributed to that gene.
For genes that are not regulated very strongly, the method used for normalization can impact the eventually observed direction of regulation.
qPCR is sensitive to the assay performance and the stability of the chosen reference genes. If amplification efficiencies are not ideal, results may change depending on the absolute Ct values of the gene. However, these are eventually "human errors" of not properly validating the qPCR assays and reference genes.
These are possible reasons that come into my mind. There may be other reasons.
Hi Charles
2 different techniques, two may different biases between them for the same targets.
did you use the same samples. I mean whole RNA for qPCR and mRNA for NGS?
fred
This can be explained by many reasons as other posters have pointed out. For example, if the data has outliers then they can effect the fold change estimation - depending on which model was used to fit the rna-seq data and qpcr data
I'm a bit late to the party but thought I would be able to chip in.
- You haven't mentioned whether you're trying to validate genes using identical samples that the RNA-seq was performed on or an independent set of samples. I'll assume the latter was done to validate biology behind the differential expression, not the technology (in case you run qPCR on the same samples).
- You provide little detail on how RNA-seq was done. E.g., if RNA-seq was done using a small sample size (n
The discussion of this topic included reads, mapping algorithms, reference genes and normalization, sample status, number of samples, and two different technologies. Here is an article saying "Do we need this?"
Article Do results obtained with RNA-sequencing require independent ...
Audrys G. Pauža , I don't understand your caclulation of the precision of the qPCR. The SD of the measurements of a gene between biol. replicates is usually not very interesting, as the samples all differ in cell numbers, efficiency of RNA isolation, cDNA synthesis and purification efficiency. This variation is typically addressed by measuring dCt values.
But let's assume you mean that under ideally reproduced conditions, the pure biological variance alone (plus technical variation of the qPCR itself) would result in an SD of Ct values of 0.3. The SD of a dCt would then be sqrt(2) * 0.3 = 0.424. The standard error of a triplicate dCt (3 biol. replicates) is then 0.424 / sqrt(3) = 0.245. The standard error of the difference between to such mean dCts (=ddCt) is then sqrt(2) * 0.245 = 0.346. The half-width of a 95% confidence interval is then 2.78 * 0.346 = 0.96, so about 1 cycle (the factor is the (1-0.025)-quantile of the t-distribution with 4 d.f.). This means that log2-fold changes can be determined with a precision of about plusminus 1 (for triplicates).
That's identical to your result :)
PS: To get a precision of the ddCt of 0.5 cycles, 12 biol. reps would be needed.
Dear Jochen Wilhelm, I’m glad we arrived to the same result. Mine was based on variance of a housekeeper across 16 samples. True that dCt is there to correct for variation but only on assumption that an optimal housekeeping gene is used. In my study, I searched for the best housekeeper based on the level of expression in my target tissue and variation between samples across conditions in my RNA-seq dataset, and then tested for variation between samples across conditions in an independent set of samples using qPCR. I compared several genes by measuring variance just to confirm that some commonly used housekeepers like Actb or B2m were real poor performers (at least for my samples). I got an SD of 0.3 Ct for my best performing housekeeper, so I used this as a reference that I’m unlikely to detect changes smaller that the variance of my most consistently expressed gene. Well at least in theory…
When I was typing this I realised that dCT values would not necessarily behave the same way… but your calculation seems to corroborate the LFC>1 detection limit of qPCR. I remember checking some RNA-seq hits with LFC
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation