Setting Peak Cutoff Using Minimum # of Reads

Question

[Andrew Siladi]

Is there standard way to set a cutoff for minimum number of reads to consider something an H3K4Me3 peak? The issue we face is that peaks with very few reads often have the greatest fold change in control vs treatment. However, when examining these peaks in IGV sometimes they look real with very few reads in the vicinity, but other times they look like background.

Thank you! Andrew

[venu]

That's a usual problem with differential analyses. Along with fold change, use normalized read count (e.g. RPM) as the cut off, e.g. min RPM must be >5.