Figure 1.
The figure illustrates the definitions of the patient subgroups and the routes by which they were recruited into the study.
Table 1.
Characteristics of study patients.
Figure 2.
Heat map of crude plasma spectral data from active TB and symptomatic controls.
Each vertical line represents an active TB patient or symptomatic control. Each horizontal line represents a protein with a particular molecular mass. Areas where a protein is present in high abundance are seen in red and low abundance in green.
Figure 3.
Mass spectra comparing 11.5 kDa and 5.8 kDa peaks in active TB and symptomatic controls.
Mass spectra from 5 kDa to 12 kDa of four active TB and four symptomatic controls individuals. Intensity in µA is plotted in y-axis.
Figure 4.
Clustering of patients with active TB and symptomatic controls with or without latent TB using principal component analysis.
a. Crude plasma spectra; b. Fractionated plasma spectra. Each sphere represents an individual patient spectrum plotted in 3D space defined by the first three principal components. Purple = active TB; Blue = symptomatic controls with latent TB; Green = symptomatic controls without latent TB.
Figure 5.
Diagnostic performance of proteomic fingerprints.
The diagnostic performance of classifiers based on proteomic fingerprints are shown using Receiver Operator Characteristic Curves (ROC). (a,b) active TB vs. all symptomatic controls using crude or pre-fractionated plasma respectively; (c,d) active TB vs. symptomatic controls with latent TB using crude or pre-fractionated plasma respectively; (e,f) active TB vs. symptomatic controls without latent TB using crude or pre-fractionated plasma respectively. The ROCs are derived from 1000 random train/test re-samplings of the data. Error bars show standard deviations. The Area Under the Curve (AUC) is shown in the centre of each plot.
Table 2.
Discrimination of active from latent tuberculosis in symptomatic patients.
Table 3.
Number of mass/charge (m/z) clusters derived from crude and pre-fractionated plasma.