Reason for NA's as p-values and q-values

[Shanta]

Hi,

I recently generated 5kb resolution outputs using chromosome size inputs from my Homer processed data, and noticed a decent number of the interactions with p-values and q-values as 'NA'. There does not seem to be nearly as many NAs for p-values and q-values when I input 10kb resolution processed data. I wanted to ask what the NAs could mean in the data, and if filtering them out would make sense.

Thanks,

Shanta

[Ferhat Ay]

It is likely because of the bias value filtering filters out more loci at 5kb resolution. For the interactions where either one end or the other end (or both) is not within the desired bias value range, the p-values will be NAs. You can certainly filter them out.

If they are too many, I would say look at the distribution (histogram) of your bias values and make sure they look like a nice Gaussian centered on bias value = 1. If not there may be a problem with your normalization