| Analysis | The best placements of paired-ends were used for identifying several different categories of SV: (i) deletions (size sd=3 kb) were identified from two or more overlapping discordant paired-ends with paired-end span >cutoff (with the condition that both putative breakpoints are spanned); (ii) simple insertions (3 kb > ssi > 2 kb) were identified from two or more overlapping discordant paired-ends with paired-end span < cutoff; (iii) mated insertions were identified from two unpaired SVs that lie in nearby (i.e. 6 kb) genomic regions and had =2 paired-ends linking to a common, distant genomic region <100 kb; mated insertions may involve tandem duplications or events related to transpositions. (iv) Inversions were called when =2 paired-ends matched different strands. (v) Unmated insertions were predicted from =2 paired ends that support a rearrangement of a genomic region in which loci change relative order without changing the relative orientation (i.e., the strand). (These events are similar to mated insertions; however, unmated insertions have only one assigned breakpoint.) In each case we required at least two paired-ends to support a predicted SV. Furthermore, at least one paired-end had to match the human reference genome at sequence identity =97%. In addition, ends were required to yield best-scoring sequence alignments genome-wide to their respective region as assessed by Blat. |
