SEACR without control – what is the difference between relaxed and stringent mode?

[teemuronkko]

I have been using SEACR without providing a control sample, and I am confused about the difference of "relaxed" vs "stringent" when no control is given.

I tested SEACR on a small bedgraph file containing 312 301 "signal blocks"/windows. I calculated this by merging the adjacent windows in the original bedgraph using "bedtools merge" to see how many "signal blocks" there would be in total. Before merging, the bedgraph had 550 181 "windows". Then, I ran SEACR on the original bedgraph without control using 0.05 as the threshold and trying both stringent and relaxed modes. With stringent, SEACR called 15 147 peaks, and with relaxed mode, SEACR called 303 851 peaks. The amount of peaks with the stringent mode made sense to me, as it is approximately 5% of the total 312 301 signal blocks. However, I don't understand how the relaxed mode can even find over 300 000 peaks if given that I wanted to have only 5% of the top peaks.

Based on the SEACR article, I thought that the distinction between relaxed vs. stringent mode is not relevant when a control sample is not available, since the stringent/relaxed threshold is calculated using the ratio of signal blocks in input sample and the total amount of signal blocks (blocks in input and control). However, if you don't have a control sample to begin with, how does SEACR find this threshold for relaxed or stringent mode? I thought that when no control is available, the algorithm just sorts the signal blocks by total signal and takes the n % of the top signal blocks by total signal as the significant peaks.