Fig. 4.

CytoDx detects latent CMV infection using high-dimensional CyTOF data. We used CytoDx to detect latent CMV infection using CyTOF samples from the ImmPort SDY478 dataset. (A) A bar graph showing the feature importance of each marker in detecting latent CMV infection. We scanned across a range of regularization strength (λ) to generate 100 candidate models. Cross-validation was used to select the optimal predictive model. Red bars represent markers included in the optimal model. For each marker, feature importance is estimated using the percent of candidate models that contain the marker. (B) We applied the optimal model on a test dataset of 19 samples. The performance is visualized by the receiver operator curve (ROC) and measured by area under the ROC curve (AUC). P values were calculated using Wilcoxon tests (Color version of this figure is available at Bioinformatics online.)