3.5 years ago by
The short answer is: there is no standard threshold.
Because these types of experiments are performed on a population of cells at one time point, the data represents a single snapshot result of a very complex system. In theory, chromatin should only exist in two states: open or closed. But in practice, there are lots of sites which are half-open. This could mean that they are open in only half the cells in the population, or that they are bound by a transcription factors with different binding affinities, or a whole host of other unknown factors.
I've personally found that the best way to validate a set of regions is to do it experimentally. For example, generate a list of DNAse HS sites, sort by descending tag-counts and experimentally test points on the scale until I find the threshold where validation begins to fail.
The quick and dirty way would be to rank by tag-counts and set an arbitrary cut-off value that pleases you. This is often a sliding scale which is context-dependent. It's a widely accepted practice.