Skip to main content

Table 3 Comparison of prediction accuracies for HLA predictive models built with and without imputed SNPs

From: Empirical evaluations of analytical issues arising from predicting HLA alleles using multiple SNPs

  

Without Imputed SNPs

With Imputed SNPs

 

HLA-

CT = 0

CT = 0.5

CT = 0.9

CT = 0

CT = 0.5

CT = 0.9

Intermediate Resolution

A

97

97(100)*

99(87)

98

98(100)

99(89)

 

B

95

95(99)

98(86)

96

96(100)

98(88)

 

C

98

98(100)

98(97)

99

99(99)

99(97)

 

DRB1

93

93(99)

97(78)

93

93(100)

98(78)

 

DQB1

96

97(99)

98(95)

97

97(100)

97(94)

High Resolution

A

95

95(100)

97(86)

97

97(100)

98(86)

 

B

93

93(98)

96(78)

93

94(97)

96(72)

 

C

97

97(99)

98(95)

98

98(100)

98(93)

 

DRB1

83

87(85)

94(48)

87

88(95)

95(59)

 

DQB1

94

95(100)

95(94)

95

95(100)

96(96)

  1. * Prediction accuracy % (call rate %)
  2. Using the Caucasian samples genotyped on the Affy 5.0 array in the FHCRC cohort, the accuracies of the predictive models built with the training set (N = 633) were
  3. evaluated on the validation set (N = 627) for HLA-A, -B, -C, -DRB1 and -DQB1 at intermediate and high resolution, with CT = 0, 0.5 and 0.9.