Figure 4.
(A) Taking read density distribution accounts for ALL identified APA isoforms, unlike every other methodology that ignores APA changes involving intermediate C/PASs. (B) Taking C/PAS distribution accounts for non-equidistant 3’UTR length changes, unlike every other methodology that assumes C/PASs to be uniformly distributed. (C) Modified vector projection-based engine that takes both the distribution of C/PASs along a gene and the read density underlaying each C/PAS into account.