standard deviation in DiffBind Consensus peaks

Question

[Dimitri]

Could someone offer an explanation to how DiffBind calculates its p- values.

Being that p-values consider variance, how does DiffBind, dba.analyze () measure variance between the peaks in replicates.

Ive reached this question because the Volcano Plot of "log10FDR v log2diff binding" did not have much scatter. It seems as if FDR and log2diff between my conditions have a linear relationship.

Answer 1

[Rory Stark]

DiffBind relies on underlying GLM-bases analysis packages (DESeq2 and/or edgeR) for the statistical analysis (p-values, FDR, and estimation of fold changes) based on fitting a negative binomial distribution.

If the variance in the replicates is consistent, there is generally a fairly linear relationship between the magnitude of the fold change and -log10(FDR), which gives the plot its typical "volcano" form.