Fig. 3: Detection of rearrangement by (a) paired-end reads and (b) junction reads.

A rearrangement within a sequencing segment can be detected when the distance between the mapped positions of paired-end reads, R1 and R2, is longer than expected or pointing toward the same direction (a). Paired reads sharing locations in close vicinity are collected to determine the presence and position of rearrangement (a′). The rearrangement can be detected as a junction read when one side of the read matches one position in the genome, and the other side matches a different position (b). Junction reads sharing the same boundary are collected as a junction cluster, and the consensus of mismatch sequence is searched on genome to locate the rearrange position (b’). Based on the relative positions and sequences around them, rearrangements are classified into five categories (c). (1) Deletion (del), (2) Tandem repeat insertion (ins), (3) Rearrangements caused by microhomology (mhr), (4) Rearrangements without sequence similarity (unk), and (5) U-turn like rearrangement (pal).