How to test the quality of RNA-seq gene expression data ?

Question

[jack]

Hi,

I'm looking for the quality control test of RNA-seq gene expression data. I have just expression values, not the sequence. which kind of test I can think of about it to have feeling about the quality of data?

I found the similar thing for microarray gene expression data.http://res.illumina.com/documents/products/technotes/technote_gene_expression_data_quality_control.pdf

Answer 1

[mikhail.shugay]

If you have several biological replicates - build scatterplots among them and compute correlation. You can also consider analyzing gene expression value distribution for "abnormalities" (e.g. with Q-Q plot), in this case you need some sample that is made using relatively same techniques and use the same expression value calculation scheme (i.e. TopHat -> FPKM -> log10).

Answer 2

[Devon Ryan]

A PCA or hierarchical clustering are pretty much your only options.

[GouthamAtla]

Is a correlation plot generated between biological replicates would not be enough?

[Devon Ryan]

If you have a lot of samples, then the correlation plots end up getting rather large and difficult to interpret, that's when PCA plots and such become a bit simpler.

[jack]

Interesting, but what PCA plot will tell me about the quality of data ? For PCA, my features will be the biological replicates?

[Devon Ryan]

It'll tell you about the underlying variance structure. If you have samples clustering in odd ways, then there are likely batch effects on sets of genes.