Figure 8

Sensitivity and specificity analysis of the strategy. 18-mer probes were assigned to SNP (+) if the sequence of the test channel probe of a specific spot contained at least one single nucleotide different from the sequence A*0101. The other probes were considered SNP(-). ROC (Relative Operating Characteristic) analysis [21] of curves for 3 different allelic combinations. The red line represents a ROC curve generated from 448 spots in 7 b,b homozygosity and b,c heterozygosity experiments with 269 SNP(-) and 179 SNP(+) spots. The blue line represents ROC curves generated from 265 spots in a,b heterozygosity experiments with 169 SNP(-) and 87 SNP(+) spots. The green line represents a ROC curve from 704 spots in all experiments with a total of 438 SNP (-) and 266 SNP (+) spots. The dots on the curves stand for optimal threshold in each group. Threshold for _LogRatio were based on ROC analysis to balance sensitivity and specificity. For the group with b,b and b,c heterozygosity (red line) experiments the optimal threshold was a _LogRatio ≤ -0.91. The sensitivity using this threshold was 82% and specificity 96%. The optimal threshold for the a,b heterozygosity (blue line) was _LogRatio ≤ -0.43 with a sensitivity of 82% but a specificity drop to 82%. For the analysis combined all the data (green line) the optimal threshold was _LogRatio ≤ -0.62 with a sensitivity at 82% and specificity at 89%. This last ROC exemplifies the most common experimental condition where the relationship between test and reference sample is not known and, therefore, all possible allelic combinations should be expected.