RNA-seq and GC content bias

[federete]

Dear all,

I am analyzing a dataset of RNA-seq (from Illumina Genome Analyser IIx), and have found some interesting expression pattern trends related to the GC content of genes. After reading the literature, I have noticed that RNA-seq has some biases related to the GC content. Hence, I am not sure of the validity of my inferences.

So, my question, or questions are:
-How can I determine if this is a biologically meaningful trend or rather so a bias of the methodology?
-To what extent does the GC content biases the results of RNA-seq experiments? (the "trends" that I observe are strong)
-I am working with published RPKM values, and was wondering if I need to recalculate all these values considering GC content biases or if I could just test how much bias is in the data.

Any clue with any of these questions is highly appreciated!
Federico

[dpryan]

Have a look at the conditional quantile normalization paper from Hansen et al. (http://biostatistics.oxfordjournals..../13/2/204.full). I expect that has much of what you're looking for.

[federete]

Many thanks, dpryan

I am not sure whether I could guess from the paper the extent of the influence of the GC content, but will do some tests with my data to see what can I see

Fede

[amcloon]

This was also a concern of mine since I work with a GC-rich species, and it is clear from Aird et al, 2011 (Genome Biology 12:R18) that PCR conditions during library prep can non-uniformly amplify transcripts with different GC contents. If you are really seeing systematic biases, I would figure out what you did/had done during library construction (polymerase, buffers, etc.), to know if you can anticipate that you had biased amplification.

I am not sure this will help your analysis any, but at least if your preparation didn't take this into account, you can write off some of the variability that way.

Anna