TABLE 2

Impact of advances in systems biology fields on the management of patients with tuberculosis (TB)

Clinical questionField of systems biologyClinical relevance of the methodMethod in clinical practice
Yes/no		Quality of evidence (GRADE) [112]
StudyBody of evidence
Who is susceptible to TB?Human geneticsThe clinical utility of LTA4H polymorphism for TB management is in phase 3 randomised clinical trial [55]NoModerateModerate
Patients with certain genetic aetiologies of MSMD and paediatric TB are treated with adjuvant IFN-γ therapy [49]YesModerate
Who will develop TB when latently infected?Human transcriptional profileRISK6 signature predictive of progression to active TB within 1 (AUC 87.6%, 95% CI 82.8–92.4) and 2 years (AUC 74.0%, 95% CI 66.0–82.0) [63]NoVery lowVery low
11-gene version of 16-gene COR signature predicts TB 1 year before the onset of disease (sensitivity 71%, specificity 84%) [66]NoVery low
ProteomicsTRM5 and 3PR plasma protein signatures predict disease progression to active TB within 1 year (sensitivity 46–49%, specificity 75%) [86]NoVery low
MetabolomicsMetabolic TB biosignature predictive of progression to active TB (sensitivity 69%, specificity 75% within 5 months of diagnosis) [94]NoVery low
ImmunophenotypingDifferent immune profiles between healthy controls, individuals with LTBI and with active TB [75]NoVery low
Identification of TB-specific endotype for application of host-directed therapy [98, 99]NoVery low
Who has active TB?Human transcriptional profileRISK6 signature detected active TB (sensitivity 90%, specificity 93.4% in HIV-uninfected and 72.5% in HIV-infected persons) [63]NoVery lowVery low
3-gene signature discriminates active TB from healthy controls (AUC 0.90, 95% CI 0.85–0.95), latent TB (AUC 0.88, 95% CI 0.84–0.92), and other diseases (0.84, 95% CI 0.80–0.95) [69]NoVery low
Bacterial transcriptomicsHighly sensitive 16S rRNA MBLA accurately quantifies M. tuberculosis viable bacillary load to as low as 10 CFU per mL [25]NoModerate
ProteomicsPlasma CC and CXC chemokines as markers of disease severity, predicting increased bacterial burden and delayed culture conversion [73]NoVery low
Serum 6-protein signature discriminates TB from other respiratory diseases (sensitivity 90%, specificity 80%) [80]NoVery low
A set of three inflammatory cytokines discriminates TB from other respiratory diseases (sensitivity 85%, specificity 96%) [82]NoVery low
8-protein signature in stimulated whole blood culture discriminates between TB and nonTB, including healthy controls, LTBI and nonTB pneumonia (sensitivity 75%, specificity 84%) [81]NoVery low
3-protein signature distinguishes between TB and other respiratory diseases (sensitivity 72.2%, specificity 75%) in cohort of children [83]NoVery low
ImmunophenotypingSingle-positive TNF-α M. tuberculosis-specific CD4+ T-cell response assay discriminates between LTBI and active disease (sensitivity 67%, specificity 92%) [95]NoVery low
Is the specific strain of M. tuberculosis susceptible to anti-TB drugs?Bacterial genomicsGenotypic prediction of phenotypic drug susceptibility by NGS technologies using validated mutation catalogues [8]YesHighHigh
How severe is TB disease?Human transcriptional profileRESPONSE5 signature correlates with pulmonary inflammation predicting Week 24 PET-CT status at baseline, Week 1 and Week 4 (AUC 0.72–0.74, p<0.02) [72]NoVery lowVery low
LipidomicsElevated plasma eicosanoid LXA4 and 15-epiLXA4 levels are associated with disease severity defined by extended lung pathology and bacterial burden [88]NoVery low
How is the patient responding to anti-TB treatment?Bacterial transcriptomics16S rRNA MBLA detects M. tuberculosis viable count in real time that correlates with culture result and can be used for treatment monitoring [25, 26, 28]NoModerateVery low
ProteomicsSerum proteomic signature predicts 8-week culture status (sensitivity 95%, specificity 90%) [76]NoVery low
MetabolomicsChanges in SLC1G level in urine correlate with different treatment response outcomes [78]NoVery low
Systems pharmacologyPK/PD models assess penetration of anti-TB drugs in cavitary lesions [100]NoVery low
Drug–drug interaction models can help set dosing for drugs by predicting the change in the efficacy of a given dose when used together with other drugs [109–111]NoVery low
What will be the outcome of anti-TB treatment?Human transcriptional profileRISK6 signature predicts treatment failure before the treatment initiation (AUC 77.1, 95% CI 52.9–100) and at the end of treatment (AUC 95.2, 95% CI 87.5–100) [63]NoVery lowVery low
9-gene DISEASE signature predicts treatment failure 1 week (AUC 0.70, p=0.04) and 4 weeks (AUC 0.72, p=0.03) after treatment initiation [72]NoVery low
MetabolomicsCerebral tryptophan metabolism can predict the outcome of TB meningitis being 9-times lower in survivors than nonsurvivors (p<0.001) [79]NoVery low
Systems pharmacologyPK/PD models identify patients at risk of treatment failure due to suboptimal drug concentrations in TB lesions [100]NoVery low
Why does a patient develop unwanted drug reactions?Microbiome profilingThe microbiota accounts for elevated levels of low-density lipoprotein cholesterol and total cholesterol that might be associated with metabolic disorders [46]NoVery lowVery low
Will the patient experience TB relapse?ImmunophenotypingSerum biosignature combining 4 immunological with 2 clinical parameters measured at diagnosis predicts TB relapse (sensitivity 75% and specificity 85% in the discovery cohort; sensitivity 83% and specificity 61% in the validation cohort) [97]NoVery lowVery low

GRADE: Grading of Recommendations, Assessment, Development and Evaluation; MSMD: Mendelian susceptibility to mycobacterial disease; IFN-γ: interferon-γ; AUC: area under curve; LTBI: latent tuberculosis infection; MBLA: molecular bacterial load assay; CFU: colony forming unit; TNF: tumour necrosis factor; NGS: next generation sequencing; PET-CT: positron emission tomography-computed tomography; PK/PD: pharmacokinetic/pharmacodynamic.