Figure 4
From: Alternative splicing tends to avoid partial removals of protein-protein interaction sites
![Figure 4](http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F1471-2164-14-379/MediaObjects/12864_2012_Article_5130_Fig4_HTML.jpg)
Toy examples of control creation for a single-isoform test (A and B) and for an all-isoforms test (C). (A) The isoform of an interacting protein for which controls are created is represented as a two-bit string, in which 0 describes a conserved position (i.e. not removed by a splicing event) and 1 a non-conserved position. A stretch of 1s is called “missing stretch” (M.S.) and is coloured in red. In this example the isoform has one missing stretch. Controls are created by randomly changing the number of residues (size) of the missing stretch according to a Poisson distribution with mean equal to the real size of that stretch and randomly selecting the starting point of the stretch. (B) The isoform has two missing stretches and the region between them is called “non-missing stretch” and is coloured in blue. The region that encompasses all the missing and non-missing stretches is called variable region. Again, controls are created changing the size of the missing stretches, but also of the non-missing stretches (N-M.S.) and by randomly selecting the starting point of the variable region. (C) A set of isoforms forming an overlap group, i.e. a group in which, for each isoform, the missing stretches partially or totally overlap with those of one or more isoforms from the same group. All the variable regions in the overlap group are fused into an “overlapping variable region” divided into segments by the junctions between missing stretches and non-missing stretches or regions outside variable regions, represented as pipes in the figure. Each segment's size is randomly changed according to a Poisson distribution with mean equal to the segment's size, then the starting point of the newly formed overlapping variable region is randomly selected.