Substitutions Cas9 leaves behind, which no single sample can measure: sequencing substitutes bases on its own at a rate that varies by position and by run. So this takes two samples — CRISPR-treated and mock-treated — and reports the positions where the ratio between them is larger than sequencing noise explains.
Substitutions are counted over R bp either side of the cleavage site.
Zero by default here: a substitution seen once is exactly what this tool is looking for, so discarding rare sequences would discard the signal.
The default counts reads containing N separately instead of as wild-type length, and does not treat a missing histogram bucket as a zero. v1 is the published arithmetic, kept so a figure from the paper can be reproduced.