skewed MAplot - DESeq

[ekimmike]

Hi,

As seen below my MAplots for RNAseq from DESeq are a bit skewed;
To me 2 and 3 are acceptable while no. 1 show clear slope suggesting that there is more upregulation of genes with lower expression and etc.
is this a normalization artifact?

I would appreciate for any suggestion, I tried to estimate dispersions in different ways, also different combination between replicates, still the same...

1.pdf

[Luyi Tian]

If only look at the MA plot, it is obvious that the condition 1 are skewed , suggesting upregulation.

RNA-seq is better than microarray in low-abundance quantification. So I think the result in condition is not due to noise or detection limit. But different library construction could lead to protocol-dependent biases due to GC content and transcript length as well as stereotypic heterogeneity in coverage across transcripts correlated with position relative to RNA termini and priming sequence bias.[1] Maybe you should take a look at these biases.



[1]Synthetic spike-in standards for RNA-seq experiments

[ekimmike]

could this be a problem of mapping instead?

If only look at the MA plot, it is obvious that the condition 1 are skewed , suggesting upregulation.

I could send some more plots for diagnostics

different library construction could lead to protocol-dependent biases due to GC content and transcript length as well as stereotypic heterogeneity in coverage across transcripts correlated with position relative to RNA termini and priming sequence bias.

I tried to separate replicates and calculated DE for each case; sample which is more GC biased looks very similar

Synthetic spike-in standards for RNA-seq experiments

Actually, the size factors assigned based on 50 spike-ins are much worse than those calculated from DESeq based on library size

[Luyi Tian]

I don't know.
But personally speaking, I think bias caused by protocol or data processing method won't cause such obvious difference in two samples.

[glados]

I have the same problem, though my plots look much more skewed than that. If it's not library construction, then what could it be? Can there be a biological reason? My groups are very different from each other. I'm very interested in hearing your thoughts and advice.

[glados]

ekimmike, did you ever find out why you plot looked like that?