In order to avoid the high false-positive rates inherent to most paired-read approaches, and to better localize the breakpoints, as is needed for orthogonal validation of the breakpoints by PCR, some algorithms for structural variation detection make use of split reads, in which a single read contains spans a breakpoint between two distant genomic regions. One indication of the presence of split reads in aligned sequence data is the existence of clusters of soft-clipped reads. Soft clips are produced by some alignment software (including Novoalign and BWA) when one end of a paired-end read maps uniquely and entirely to the genome but the other end does not. If the second end maps only partially, but in the correct orientation and has an [insert size] within the normal range, the remainder of the sequence is represented as a “soft clip” (29). These soft-clipped reads often indicate reads with split mappings and so can be used to localize translocations with single-base accuracy. CREST and ClipCrop are two algorithms that make use of soft-clipping information to detect split reads (30,31). SLOPE is another
