Hi I had a question about the following example in the diffbind tutorial:

Plot effect of a range of filter values and then apply filter

data(tamoxifen_counts)
rate.max <- dba.count(tamoxifen, peaks=NULL, filter=0:250)
rate.sum <- dba.count(tamoxifen, peaks=NULL, filter=0:250,filterFun=sum)
plot(0:250,rate.max/rate.max[1],type='l',xlab="Filter Value",ylab="Proportion Retained Sites")
lines(0:250,rate.sum/rate.sum[1],col=2)
tamoxifen <- dba.count(tamoxifen,peaks=NULL,filter=125,filterFun=sum)
tamoxifen

I understand everything done here except for why a filter of 125 was applied as the final step, how is this filter value obtained as the optimal from the proportion of retained sites? Thank you.

I don't think there is any deep principle underlying the choice of 125, this is just an example in the man page to show how to look at the impact of different filters then how to apply a filter.

In my own work I am more likely to use a max filter with a much lower threshold (10 or 20), so I filter sites that don't have at least one sample with a reasonable count.