base composition variation in Illumina runsH

[chrisbala]

Hi everyone,

We've recently done a set of mRNA-seq runs using the Illumina platform, 75bp reads. The library was random primed, and polyA selected.

We notice that the GC content of the first ~12 bases fluctuates rather dramatically, suggesting a pretty strong bias in the transcripts that are being sequenced. We do not believe that these fluctuations are caused by adaptors. I've been told that random priming is not exactly random, and that this type of bias is not atypical. Attached are plots of base compostion (%A, %G etc.) for each position across all reads.



I just wanted to confirm with a broader audience that other people see things like this in their data. Or whether I should be terribly depressed about the condition of our data. Note that these images are from 5 different runs (each 1 lane) sequenced in two different batches (months apart) and that the pattern of the fluctuations is consistent in all 5.

Also, FYI, I used fastQC, a very handy little software to make these ( and a number of other) QC plots.

Thanks for your input!

Chris

[pzumbo]

*not* seeing those patterns following illumina's poly-a, random-primed mrna-seq protocol would be disconcerting.

i don't think this should lessen, however, being terribly depressed about the conditions of the data!

[kmcarr]

Check out this paper:

Hansen et al. Biases in Illumina transcriptome sequencing caused by random hexamer priming. Nucleic Acids Res (2010) (link)

They analyze in depth the biases in the Illumina random hexamer priming method.

[chrisbala]

bias correction

THanks!

Any thought about/experience with the bias correction proposed in the paper above? I'll give it a shot and see what happens...

[chrisbala]

base composition differences

This is not exactly related to my posts above - but does anyone know of any methods to correct for base composition differences among runs?

Here I'm not talking about differences between sites within reads as above, here I am talking about differences among libraries in gc content.