Table 5. Summary of the performance of methods using haplotypes or intensities to infer the -α -3.7I genotypes from features/SNPs within the -α -3.7I deletion region.
| Methods | Principle | Directly
typed (N) |
Sensitivity *** | Specificity *** | Positive Predicative
Value *** |
Overall
Association (OR [95%CI] p-value) |
Model | Genotypic Association (OR)
adjusted |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Norm | Het | Homo | Norm | Het | Homo | Norm | Het | Homo | Norm
(ref) |
Het | Homo | |||||
| Direct
genotyping |
PCR
genotyping |
3036 | NA | NA | NA | 0.78
(0.70-0.87) (p=1.1x10 -5) |
Additive | 1 | 0.79
[0.67-0.93] (p=0.005) |
0.60
[0.47-0.76] (p=2.43x10 -5) |
||||||
| Intensity
distributions (4 features) * |
Intensity
thresholds (AF3;XX) |
0 | 24.1 | 83.1 | 93.1 | 91.3 | 44.7 | 95.8 | 81.4 | 61.8 | 78.2 | 0.76
[0.64-0.9] (p=0.0013) |
Homozygous
or Heterozygous |
1 | 0.74
[0.62-0.88] (p=7.2x10 -4) |
0.83
[0.66-1.04] (p=1.11x10 -11) |
| to | to | to | to | to | to | to | to | to | to | to | to | |||||
| 63.2 | 94.8 | 95.2 | 98.8 | 72.9 | 98.6 | 92.6 | 74.4 | 91.8 | 1.74
[1.42-2.13] (p=5.8x10 -8) |
1.87
[1.43-2.44] (p=4x10 -6) |
1.49
[1.06-1.98] (p=2.1x10 -9) |
|||||
|
Hierarchical
clustering (5 features) |
Hierachical
clustering (AF3;TT) |
0 | 56 | 74 | 96 | 95 | 71 | 86 | 80 | 73 | 88 | 1.43
[1.28-1.69] (p=1.3x10 -10) |
Additive | 1 | 1.12
[0.91-1.37] (p=0.05) |
2.02
[1.6-2.55] (2.2x10 -9) |
| Hierarchical
clustering (4 features) |
Hierarchical
clustering (AF3;TT) |
0 | 41 | 74 | 96 | 93 | 61 | 87 | 73 | 67 | 88 | 1.86
[1.56-2.22] (p=4.98x10 -12) |
Recessive | 1 | 1.24
[1.02-1.49] (p=0.02) |
2.02
[1.62-2.51] (p=2.5x10 -10) |
| MRM * | Regression
(AF3;VV) |
100 | 58 | 75 | 79.8 | 83.1 | 53.4 | 96.6 | 70.1 | 61.7 | 71.6 | 0.79
[0.68-0.92] (p=0.0025) |
Heterozygous
or Dominant |
1 | 0.8
[0.67-0.95] (p=0.01) |
0.81
[0.65-1.02] (p=0.08) |
| to | to | to | to | to | to | to | To | to | to | To | to | |||||
| 73 | 86 | 85 | 89.6 | 75.3 | 98.6 | 79.2 | 76.7 | 90.1 | 1.67
[1.45-1.94] (p=3.7x10 -12) |
1.63
[1.37-1.95] (p=3.5x10 -8) |
2.61
[2.05-3.33] (p=6.8x10 -15) |
|||||
| CART * | Hierarchical
clustering and regression (AF3;UU) |
100 | 68.1 | 45.5 | 68.6 | 76.7 | 61.3 | 96.7 | 62.4 | 61.4 | 67.6 | 0.799
[0.69-0.93] (p=0.004) |
Heterozygous,
or Dominant |
1 | 0.79
[0.67-0.93] (p=0.01) |
0.81
[0.64-1.02] (p=0.08) |
| to | to | to | to | to | to | To | To | to | to | To | to | |||||
| 78.8 | 86.7 | 97.2 | 97.7 | 81.6 | 99.1 | 84.4 | 81.4 | 92.9 | 1.70
[1.47-1.97] (p=1.19x10 -12) |
1.72
[1.44-2.04] (p=9.8x10 -9) |
1.9
[0.85-1.39] (0.47) |
|||||
| IMPUTE2
**
70% threshold (means ± SD) |
Imputation
(1000 runs) (AF3;AG/AI) |
100 | 77 | 77 | 75 | 93 | 88 | 97 | 87 | 83 | 94 | 0.74 ± 0.07
(p =1.71x10 -3 ± 3x10 -3) |
Additive,
dominant or recessive |
0.79 ± 0.04
(p=0.02 ± 0.04) |
0.61 ± 0.04
(p=0.009 ± 0.003) |
|
| ± | ± | ± | ± | ± | ± | ± | ± | ± | 1 | |||||||
| 5 | 5 | 9 | 2 | 2 | 1 | 2 | 3 | 1 | ||||||||
| IMPUTE2
**
90% threshold (means ± SD) |
Imputation
(1000 runs) (AF3;AH/AJ) |
100 | 68 | 66 | 62 | 95 | 92 | 98 | 85 | 79 | 93 | 0.76 ± 0.08
(p = 0.005 ± 0.010) |
Additive,
dominant or recessive |
0.77 ± 0.05
(p=0.03 ± 0.06) |
0.61 ± 0.05
(p=0.004 ± 0.013) |
|
| ± | ± | ± | ± | ± | ± | ± | ± | ± | 1 | |||||||
| 6 | 8 | 14 | 2 | 2 | 1 | 2 | 4 | 2 | ||||||||
Further details of methods and results can be found in the Extended Data.
* Where several SNPs (as indicated) were tested, then the ranges of results are shown.
** Data are mean ± SD for 1,000 individual runs of IMPUTE2 (70% and 90% refer to the threshold used to assign a genotype).
*** Data are shown as percentages.
MRM, Multiple-Regression Model; CART, Classification and Regression Trees.