Skip to main content

Table 2 Average accuracy (error rate) and average size of the gene panel (number of selected genes) selected by automated predictors of infected state (class) for different viral challenges (data from DEE2/DEE5, DEE3/DEE4 and HRV-UVA/HRV-Duke were pooled and designated as H3N2, H1N1, and HRV in table). Shown are the reference-aided predictor (w/ baseline reference), the standard predictor (w/o baseline reference), the standard predictor with constraints to have the similar complexity in terms of the number of genes as the reference-aided predictor (constrained standard predictor), and the differential predictor. Across all viruses the inclusion of the baseline reference results in a decrease in error rate and a reduction in the number of genes used by the predictor. The reported accuracy and size of panel were computed by cross-validation of the predictors using leave-one-subject-out resampling to partition the data into training and target samples

From: An individualized predictor of health and disease using paired reference and target samples

Virus H3N2 H1N1 HRV RSV
Classes 2345 2345 2345 234
w/ baseline reference     
Error rate 0.386 0.480 0.483 0.635
Number of selected genes 200.34 287.54 287.55 238.53
w/o baseline reference     
Error rate 0.448 0.508 0.526 0.728
Number of selected genes 327.90 420.25 358.48 593.82
Constrained standard predictor     
Error rate 0.458 0.517 0.544 0.737
Number of selected genes 207.38 271.83 298.00 243.29
Differential predictor     
Error rate 0.415 0.492 0.571 0.678
Number of selected genes 481.48 1209.46 464.39 503.29