Hi,

We are trying to use diffBind and are puzzled by the scale on our correlation matrices. In one of our datasets the correlation goes between 0.9-1, while in the other from 0.7-1. Is there a way to modify it such that both are from 0-1 or 0.5-1? Thank you,

Elisheva J

Hello-

The main heatmap plot shows correlation values. DiffBind uses heatmap.2 in ggplots for the underlying plot, which by default scales to the range of values. Hence if everything is highly correlated the scale may only include values above 0.9, while in another case there can be anti-correlation and a much different scale between -1 to +1.  heatmap.2 doesn't offer a way to specify the scale.

What you can do is re-scale the values using z-scores, either by row or by column. For the correlation heatmaps, the rows and columns are the same (symmetrical square matrix) so it doesn't matter which you choose, but you will always get a scale between -2 and +2. An example:

> plot(myDBA, scale="row")

(equivalent to dba.plotHeatmap(myDBA, scale="row"))

Cheers-

Rory

You are correct, even the Z-score transformation doesn't fix the scale (based on the degree of variance). It would be nice to be able to do this, I'll keep looking into it.

-R

Thank you