Table 3 Model performance comparison of the 10-SNP set Bayes Classifier by Maroñas et al. (2014) and the 36-SNP set prediction model from the present study using the independent “model comparison set” of 194 individuals from 17 populations not previously used for marker discovery by applying the same phenotyping method previously employed by Maroñas et al. (2014) to allow direct comparison of the two prediction approaches

From: Global skin colour prediction from DNA

  AUC Sensitivity Specificity PPV NPV
Bayes classifier 10-SNP model Maroñas et al. (2014)      
 White 0.79 0.97 0.62 0.84 0.91
 Int 0.63 0.37 0.88 0.47 0.83
 Black 0.64 0.30 0.98 0.67 0.92
36-SNP set model current study      
 White 0.82 0.99 0.65 0.86 0.98
 Int 0.62 0.26 0.98 0.79 0.82
 Black 0.92 0.90 0.94 0.64 0.99
36-SNP set model current study—Fitzpatrick scale*      
 Light 0.92 0.99 0.85 0.95 0.98
 Dark 0.74 0.50 0.99 0.86 0.93
 Dark-Black 0.94 0.92 0.96 0.79 0.99
  1. * The 36-SNP set model performance assessment using Fitzpatrick scale phenotypes as the observed phenotype