Strategy | Calls (C) | Valid (v) | PPV ∗ | (SE, n) | Sensitivity ∗ | (SE, n=223) |
---|
IR | 127 | 64 | 0.50 | (0.044, 127) | 0.83 | (0.025) |
MG | 106 | 41 ∗ | 0.39 | (0.056, 76) | 0.53 | (0.033) |
PM | 199 | 71 ∗ | 0.36 | (0.043, 124) | 0.92 | (0.018) |
VS | 381 | 26 ∗ | 0.07 | (0.030, 72) | 0.34 | (0.031) |
Any one | 637 | 77 ∗ | 0.12 | (0.022, 223) | 1.00 | (NA) |
IR ∩ MG | 42 | 40 | 0.95 | (NA, 42) | 0.52 | (0.033) |
IR ∩ PM | 94 | 58 | 0.62 | (0.050, 94) | 0.75 | (0.029) |
IR ∩ VS | 27 | 26 | 0.96 | (NA, 27) | 0.34 | (0.031) |
MG ∩ PM | 40 | 39 | 0.97 | (NA, 40) | 0.51 | (0.033) |
MG ∩ VS | 31 | 22 | 0.71 | (0.082, 31) | 0.29 | (0.030) |
PM ∩ VS | 29 | 26 | 0.90 | (NA, 29) | 0.34 | (0.031) |
Any two | 111 | 61 | 0.55 | (0.047, 111) | 0.79 | (0.027) |
IR ∩ MG ∩ PM | 38 | 38 | 1.00 | (NA, 38) | 0.49 | (0.033) |
IR ∩ MG ∩ VS | 22 | 22 | 1.00 | (NA, 22) | 0.29 | (0.030) |
IR ∩ PM ∩ VS | 27 | 26 | 0.96 | (NA, 27) | 0.34 | (0.031) |
MG ∩ PM ∩ VS | 22 | 22 | 1.00 | (NA, 22) | 0.29 | (0.030) |
Any three | 43 | 42 | 0.98 | (NA, 43) | 0.55 | (0.033) |
IR ∩ MG ∩ PM ∩ VS | 22 | 22 | 1.00 | (NA, 22) | 0.29 | (0.030) |
IR ∪ PM | 232 | 77 ∗ | 0.33 | (0.026, 157) | 1.00 | (NA) |
(IR ∪ PM) ∩ (MG ∪ VS) | 51 | 45 | 0.88 | (0.046, 51) | 0.58 | (0.033) |
- Variant calls made and calls validated by visual inspection are given for each software package, as well as their intersections, excluding those variants identified by ANNOVAR as intronic or intergenic, which leaves 223 relevant SNPs collectively identified. The column labelled PPV shows Positive Predictive Value. Sensitivity and PPV were calculated as Sensitivity=v/V and PPV=v/C, respectively, where numerator v is the number of valid variants correctly called by a given caller or set of callers, denominator V is the extrapolated set of 77 valid variants (including 68 examined variants that we classified as valid plus 9 additional “valid” variants estimated via extrapolation), and denominator C is the total number variants called by all callers. The strategies described as “Any one,” “Any two,” and “Any three” each require the variant site to be identified by at least that many packages, which is effectively a union of sites identified by all the combinations in that section of the table. Asterisks (*) indicate values that were extrapolated. Standard error (SE) for PPV and Sensitivity is provided when n∗min(p,1−p)≥5, where n is the number of variants examined rather than the number called, because the latter is based on an extrapolation and does not contribute to estimated precision