different phenotypes using RNA-seq and microarray analysis

Question

[Nil Desperandum]

Hi,

We have gene knockdowns (12 gene targets). We have both microarray (HTA2.0) and RNA-seq (100 bp, SE, illumina NEB ultra low library prep kit) data (don't ask me why we did both, we just did). The thing is, the data is concordant with each other except rom one particular knockdown.

We took this geneX and did independent knockdown experiments, and hybridised to another array type (U219 affymetrix) and also repeated the RNA-seq on a similar knockdown samples. The result is the same: a massive transcriptional phenotype with the array, but no phenotype with the RNA-seq. We also know that the knockdown of a gene has a physical phenotype in the cell (a differentiation phenotype). So perhaps the array is right.

Assuming there are no technical mess-ups- what can be causing this difference? Could it be that this geneX effects transcription in a global manner such that its effects are picked up by the array-based normalisation techniques, but not seen by the RNA-seq analysis tools?

Can you think of any reasons?

[Christian]

Interesting. Hard to answer without knowing specifics of how you call DEGs, but my hunch is it has to do with normalization. In case you figure out, could be an illustrative example to publish on its own to understand the limits of current RNA-seq protocols!

Answer 1

[igor]

The array is sampling specific regions of the gene, but RNA-seq is sampling the entire gene. What exactly is happing depends on how your knockdown works.