Fig 1.
Positive (red X) and negative (blue ♦s) training antibody data.
(a) N plotted against RBD, (b) the ELISA-based total IgG is added as the vertical axis. The green boxes in (a) and (b) are the negative sample mean plus 3σ confidence intervals. S1 Fig shows an animation of (b).
Fig 2.
3D probability model plotted along with the training data.
Positive samples are indicated by red Xs and negatives with blue ♦s. S2 Fig shows an animation.
Fig 3.
Optimal classification domains for the training (a) and test (b) data.
Positive samples are indicated by red Xs and negatives with blue ♦s. The green boxes drawn in (a) and (b) are the negative sample mean plus 3σ confidence intervals. S3 and S4 Figs show animations.
Fig 4.
Optimal classification domains for the training (a) and test (b) data with a holdout region (white; magenta and cyan markers are indeterminate samples).
Positive samples are indicated by red Xs and negatives with blue ♦s. The green boxes drawn in (a) and (b) are the negative sample mean plus 3σ confidence intervals. S5 and S6 Figs show animations.
Table 1.
Summary information about the SARS-CoV-2 datasets with sensitivities, specificities, and classification accuracies for training and test data with and without allowing an indeterminate class.
Model and 3σ (relative to negative sample mean) confidence interval results are shown for all samples; the original analysis [9] was conducted on all seven antibody targets and the ELISA-based total IgG without indeterminate samples.