Figure 4: Frequency of analytes in the top 20 GDA predictive models that most accurately classified study participants as TB disease or no TB. The columns represent the number of times each analyte was included in the top 20 discriminatory models. (a) Frequency of analytes in the models generated from the host markers detected in saliva, (b) frequency of analytes in models generated when the data obtained from saliva were combined with the data obtained from serum samples.