Volcano Plot Of Significantly Differential Exp Genes

Question

[l0o0]

I did the protocol as Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks from the nature protocol. The online pdf is here

I have a question when i see Figure 8, as below,

I find the boundary between sig and no sig is around 2.5 on the left axis, its p-value is about 0.003. But i think it should be 2(p-value 0.01) or 1.3(p-value 0.05) on the axis.
I am confused by this figure and need some help.

[matted]

Maybe the threshold is corrected for the multiple hypotheses being tested?

[Bioch'Ti]

From a biological point of view, I'm wondering what is the interpretation of such a volcano plot? I mean, I find 'suspicious' that the DE genes show very low fold changes... (even on a log2 scale). Do you expect such pattern?

[Neilfws]

"Significant" in this case is based solely on the alpha (p) value. It's not suspicious, just completely arbitrary :)

[l0o0]

Do you mean "Significant" is arbitrary? But i think they have use the wrong alpha value.

[Neilfws]

Yes to both of those :)

Answer 1

[Neilfws]

The article provides the R code used to generate the plot:

csVolcano(genes(cuff_data), 'C1', 'C2')

According to the documentation for csVolcano() the parameter alpha:

alpha: Provide an alpha cutoff for visualizing significant genes

takes a default value of 0.05, which corresponds to 1.3 on the axis in this figure.

However, as you correctly note, the cutoff in the figure is closer to 2.5, which would correspond to alpha = 0.003.

So either:

• in older versions of cummeRbund, the default alpha value was around 0.003
• there is an error in the article and the code does not exactly match the figure

I suspect the latter is more likely.

[l0o0]

I think i should ask the author for help. Thank you for your reply.

[mfahim]

Hi Neil,

Could you update the thread.. did you get help on this?

[Neilfws]

Not me asking for help.