Skip to main content

Table 2 Accuracy of variant calling across genomic contexts and variant types in patient samples

From: Whole-genome sequencing as an investigational device for return of hereditary disease risk and pharmacogenomic results as part of the All of Us Research Program

 

Panel+a/WGS–b

Panel–/WGS+

Panel+/WGS+

Panel–/WGS–

PPA [95% CI]

NPA [95% CI]

All compiled

76c

201

29,475

45,847,148

99.74% [99.68–99.81%]

100% [100–100%]

P/LP variants only

0

0

145

28,474,503

100% [100–100%]

100% [100–100%]

SNVs only

23

6

28,710

45,848,108

99.92% [99.87–99.97%]

100% [100–100%]

Insertions only

9

8

225

29,887,204

96.20% [94.30–98.00%]

100% [100–100%]

Deletions only

33

12

540

42,170,158

94.24% [92.65–95.83%]

100% [100–100%]

Segmental duplications

12

2

2119

38,466,267

99.44% [99.10–99.78%]

100% [100–100%]

Known pseudogenes

9

74

1075

45,676,596

99.17% [98.60–99.73%]

100% [100–100%]

Low mappability regions

18

12

1069

36,912,107

98.34% [97.84–98.85%]

100% [100–100%]

Low complexity regions

24

69

956

38,072,735

97.55% [96.35–98.75%]

100% [100–100%]

Low GC regions

6

12

197

14,792,575

97.04% [95.87–98.22%]

100% [100–100%]

High GC regions

0

0

23

3,891,553

100% [100–100%]

100% [100–100%]

Heterozygous variants only

75

147

18,295

45,858,382

99.59% [98.48–99.70%]

100% [100–100%]

Homozygous variants only

1

83

11,180

45,865,561

99.99% [99.97–100%]

100% [100–100%]

  1. a+ indicates that a variant was present on the panel or the corresponding whole genome sequencing (WGS) sample. b– indicates that a variant was not present on the panel or the corresponding WGS sample. cSubsequent to the investigational device exemption (IDE) submission, additional review indicated that this number should in fact be a total of 21 panel positive, but WGS negative variants distributed across the various subcategories (some variants were missannotated in the original tables). However, to reflect what was actually submitted as part of the IDE, the original number is listed here. The impact on overall performance measures is small. PPA positive percent agreement, CI confidence interval, NPA negative percent agreement, P/LP pathogenic/likely pathogenic, SNVs single-nucleotide variants