Skip to main content

Table 4 Best performing SV tool for each size class of insertion and deletion using normal and less-ideal datasets of Arabidopsis and Human Chromsome 21

From: SV-AUTOPILOT: optimized, automated construction of structural variation discovery and benchmarking pipelines

A. Precision
  Data type Std Dev Length 20–49 P (rel/str) Length 50–99 P (rel/str) Length 100–249 P (rel/str) Length 250–999 P (rel/str) Length 1 K-50 K P (rel/str)
Insertion
  Human chr. 21 15 Clever Pindel Clever Clever n/a
88.3/77.8 85.7/68.8 88.5/45.8 100.0/50
Tair v.9 15 Pindel Clever/ Pindel Clever Clever n/a
95.0/93.2 94.3/47.4 68.1/23 66.7/55.6
  Human chr. 21 50 Pindel Clever/Pindel Clever n/a n/a
87.9/75.2 83.3/55.6 66.7/17.9
Tair v.9 50 Pindel Clever/Pindel Clever Clever n/a
94.9/92.7 94.6/46.2 73.9/8.3 60.0/20
Deletion       
Human chr. 21 15 Clever/Pindel Pindel Clever/Pindel Pindel Breakdancer/Clever
92.8/89 92.3/76.9 84.6/42.9 100/100 100/33.3
Tair v.9 15 Pindel Delly Pindel Pindel Clever
  94.6/94.2 100/100 89.2/90.2 87.9/88.7 68.3/58.1
Human chr. 21 50 Clever/Pindel Pindel Clever/Pindel Pindel Breakdancer/Clever
90.9/88.2 100/90 59.1/42.9 81.8/81.8 100/50
Tair v.9 50 Pindel Delly/Pindel Pindel Pindel Clever/Pindel
94.5/94.3 100/90.6 86.5/86.3 89.4/89.4 69.6/61
B. Recall
  Data type Std Dev Length 20–49 R (rel/str) Length 50–99 R (rel/str) Length 100–249 R (rel/str) Length 250–999 R (rel/str) Length 1 K-50 K R (rel/str)
Insertion       
Human chr. 21 15 Clever Clever Clever Clever n/a
72.8 83.3 60 5.3 n/a
Tair v.9 15 Clever Clever Clever Clever n/a
56.4 81.2 92.7 15.9 n/a
Human chr. 21 50 Pindel Clever Clever Clever n/a
  61.8 48.6 76.7 10.5 n/a
Tair v.9 50 Pindel Pindel Clever Breakdancer n/a
34.6 43 95.1 51.2 n/a
Deletion       
Human chr. 21 15 Prism Clever Clever Delly Breakdancer
88.1 90.9 52.6 73.7 50
Tair v.9 15 Clever Clever Clever Delly SVDetect
65 82.1 89.8 93 97
Human chr. 21 50 Prism Prism Clever Breakdancer Breakdancer
85.6 72.7 63.2 73.7 50
Tair v.9 50 Pindel Clever Clever Delly SVDetect
94.5 52.9 91.9 94.9 96.2
  1. For this work, a standard deviation of 15 is considered normal while a standard deviation of 50 is considered less-ideal. Recall and Precision were two measures used to evaluate the ability of a tool to accurately predict SVs. Here the winner for each length class is provided along with the tools winning value for that category. (P = Precision; R = Recall; Std. Dev = Standard Deviation of the Insert size; n/a = no call was made by any tools tested). The Additional file 1 contains all tool performance statistics in the PDF reports. In Table 4a both ‘relaxed’ and ‘strict’ criteria (REL/STR) (see Methods) are provided for the precision measurements which indicates how accurate the tools are at making their calls. In Table 4b the scores of tool recall demonstrate how much of the SVs the tools are able to discover.