What Do You Think About This Correlation Between Qpcr And Rna-Seq?

Hi community,

I would like to know your opinions about the correlation I obtained between qPCR and RNA-seq for 8 genes, being two of them reference genes. The r2 is good but I worry that qPCR FC are not as high as in RNAseq. It could be:

Normalisation bias in RNAseq analysis: normalisation was performed because ex-vivo sample had 40% of eukaryotic RNA.
Bias in library preparation, PCR step
Other

Here are the correlation analysis and the technical analysis result:

https://drive.google.com/file/d/0B8-ZAuZe8jldQnEzcThkOXlkNWc/edit?usp=sharing https://drive.google.com/file/d/0B8-ZAuZe8jldYmZaSnZxMFRuVWs/edit?usp=sharing

Thanks for your help. Best, Bernardo

rnaseq

• 4,828 views • link

updated 12.5 years ago by Charles Warden

• written 12.5 years ago by biotech

1 answer

I think your results look quite good.

I would typically expect ~75% validation rate for microarray data (and presumably something similar with RNA-Seq data). This means, you check for differential expression between conditions for each gene in the qPCR results and do something like a t-test to check for a p-value < 0.05. It doesn't look like this is exactly what you did, but it is probably worth adding (although I don't see replicates in the detailed table).

This is the sort of thing I would look for when doing "validation." When a correlation is calculated in addition to that, I would say you'll doing well when you get a r > 0.70. You had an r = 0.95. That is is clearly good. I should just recommend checking more samples per group (if I am right about the lack of replicates).

• 0 views • link

written 12.5 years ago by Charles Warden

Thanks cwarden45, I had three technical replicates/sample in the qPCR but no biological ones neither in qPCR or RNAseq. I worry that qPCR FC are not as high as in RNAseq.

• 0 views • link

written 12.5 years ago by biotech

Microarray fold-change values tend to be lower than RNA-Seq and qPCR fold-change values. I haven't seen a consistent trend for qPCR versus RNA-Seq: I think it depends upon your primer design.

I would use IGV to visualize the read coverage to make sure that the areas overlapping your primers have decent coverage.

Other than that, I don't know what to suggest beyond using replicates to tell which genes significantly vary between groups. I would typically use |fold-change| > 1.5 and FDR < 0.05 to identify differential expressed RNA-Seq genes (for qPCR, you can use p-value instead of FDR if you just look at a handful of genes to validate). It only looks like there is one gene that fails to meet the fold-change criteria in both RNA-Seq and qPCR, and validating 5 out of 6 genes would be good. If the results were always identical, you wouldn't need to do qPCR validation in the first place.

• 0 views • link

written 12.5 years ago by Charles Warden