Fig. 3.

Clustering split clones to detect inversions. a We first identify clone locations that are shorter than the expected clone size, but when paired with another short clone found in the same pool, the total length sums up to a full clone length. We refer to such clones as “split clones”. b We then cluster pairs of split clones that are mapped to approximately the same breakpoints. Note that due to read mapping errors and our clone reconstruction heuristics, a split clone may be identified as spanning a breakpoint. c Finally we cluster multiples of split clones from different pools if they agree on breakpoint location and the size of the inversion. gap: size of the region between the start and end locations of split clones from different pools. overlap: size of the overlapping region of split clones from different pools