Skip to main content

Table 2 Improvement of breakpoint resolution using local de novo assembly and breakpoint refinement in SVMerge

From: Enhanced structural variant and breakpoint detection using SVMerge by integration of multiple detection methods and local assembly

  Raw Refined
SV type Called Correct Mean distance Breakpoints detected Correct Mean distance
Homozygous       
   Deletion (random) 99 9 +5/-3 99 77 -1/-1
   Deletion (repeat) 99 4 +11/-8 99 89 0/0
   Inversion 100 0 -169/175 85 46 51/24
   Insertion 99 0 0/205 97 60 -1/1
Heterozygous       
   Deletion (random) 96 2 +6/-4 96 40 -35/+18
   Deletion (repeat) 94 0 +19/-15 91 35 0/2
   Inversion 99 0 -166/+165 73 30 -58/+287
   Insertion 96 0 +1/+202 18 18 0/0
  1. To evaluate the performance of the local assembly and breakpoint refinement step in SVMerge, structural variants (SVs) were generated in human chromosome 20. For each category, 100 SVs were generated by random selection of location and size. Repeat deletions were selected from a list of LINEs and SINEs on chromosome 20. The raw, unfiltered calls are from BreakDancer raw output, except insertion calls, which are from SECluster raw output. 'Called' is the total number of SVs out of the 100 simulated SVs that were found in the raw output; 'Breakpoints detected' is the number of SVs, out of the total called, for which the SVMerge pipeline was able to detect breakpoints with the local assembly, and contig alignment and analysis steps; 'Correct' is the number of predictions that had matches to the actual breakpoint coordinates; 'Mean distance' is the mean distance from the actual breakpoints, where the numbers represent the 5'/3' breakpoints. The '+' indicates the mean distance was upstream of the actual breakpoint, and '-' indicates the mean distance was downstream. Raw and refined breakpoints were considered 'correct' if the direction and deviation at both the 5' and 3' breakpoints were equal.