Dataset Information

Multinomial modelling of TB/HIV co-infection yields a robust predictive signature and generates hypotheses about the HIV+TB+ disease state.

ABSTRACT: BACKGROUND:Current diagnostics are inadequate to meet the challenges presented by co-infection with Mycobacterium tuberculosis (Mtb) and HIV, the leading cause of death for HIV-infected individuals. Improved characterization of Mtb/HIV coinfection as a distinct disease state may lead to better identification and treatment of affected individuals. METHODS:Four previously-published TB and HIV co-infection related datasets were used to train and validate multinomial machine learning classifiers that simultaneously predict TB and HIV status. Classifier predictive performance was measured using leave-one-out cross validation on the training set and blind predictive performance on multiple test sets using area under the ROC curve (AUC) as the performance metric. Linear modelling of signature gene expression was applied to systematically classify genes as TB-only, HIV-only or combined TB/HIV. RESULTS:The optimal signature discovered was a 10-gene random forest multinomial signature that robustly discriminated active tuberculosis (TB) from other non-TB disease states with improved performance compared with previously published signatures (AUC: 0.87), and specifically discriminated active TB/HIV co-infection from all other conditions (AUC: 0.88). Signature genes exhibited a variety of transcriptional patterns including both TB-only and HIV-only response genes and genes with expression patterns driven by interactions between HIV and TB infection states, including the CD8+ T-cell receptor LAG3 and the apoptosis-related gene CERKL. CONCLUSIONS:By explicitly including distinct disease states within the machine learning analysis framework, we developed a compact and highly diagnostic signature that simultaneously discriminates multiple disease states associated with Mtb/HIV co-infection. Examination of the expression patterns of signature genes suggests mechanisms underlying the unique inflammatory conditions associated with active TB in the presence of HIV. In particular, we observed that dysregulation of CD8+ effector T-cell and NK-cell associated genes may be an important feature of Mtb/HIV co-infection.

SUBMITTER: Duffy FJ

PROVIDER: S-EPMC6629068 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Multinomial modelling of TB/HIV co-infection yields a robust predictive signature and generates hypotheses about the HIV+TB+ disease state.

Duffy Fergal J FJ Thompson Ethan G EG Scriba Thomas J TJ Zak Daniel E DE

PloS one 20190715 7

<h4>Background</h4>Current diagnostics are inadequate to meet the challenges presented by co-infection with Mycobacterium tuberculosis (Mtb) and HIV, the leading cause of death for HIV-infected individuals. Improved characterization of Mtb/HIV coinfection as a distinct disease state may lead to better identification and treatment of affected individuals.<h4>Methods</h4>Four previously-published TB and HIV co-infection related datasets were used to train and validate multinomial machine learning ...[more]

PMID: 31306460

Similar Datasets

Project description:BACKGROUND:Gut microbiota interacts with the human gut in multiple ways. Microbiota composition is altered in inflamed gut conditions. Likewise, certain microbial fermentation products as well as the lipopolysaccharides of the outer membrane are examples of microbial products with opposing influences on gut epithelium inflammation status. This system of intricate interactions is known to play a core role in human gut inflammatory diseases. Here, we present and analyse a simplified model of bidirectional interaction between the microbiota and the host: in focus is butyrate as an example for a bacterial fermentation product with anti-inflammatory properties. RESULTS:We build a dynamical model based on an existing model of inflammatory regulation in gut epithelial cells. Our model introduces both butyrate as a bacterial product which counteracts inflammation, as well as bacterial LPS as a pro-inflammatory bacterial product. Moreover, we propose an extension of this model that also includes a feedback interaction towards bacterial composition. The analysis of these dynamical models shows robust bi-stability driven by butyrate concentrations in the gut. The extended model hints towards a further possible enforcement of the observed bi-stability via alteration of gut bacterial composition. A theoretical perspective on the stability of the described switch-like character is discussed. CONCLUSIONS:Interpreting the results of this qualitative model allows formulating hypotheses about the switch-like character of inflammatory regulation in the gut epithelium, involving bacterial products as constitutive parts of the system. We also speculate about possible explanations for observed bimodal distributions in bacterial compositions in the human gut. The switch-like behaviour of the system proved to be mostly independent of parameter choices. Further implications of the qualitative character of our modeling approach for the robustness of the proposed hypotheses are discussed, as well as the pronounced role of butyrate compared to other inflammatory regulators, especially LPS, NF- κB and cytokines.

Project description:While blood transcriptional profiling has improved diagnosis and understanding of disease pathogenesis of adult tuberculosis (TB), no studies applying gene expression profiling of children with TB have been described so far. In this study, we have compared whole blood gene expression in childhood TB patients, as well as in healthy latently infected (LTBI) and uninfected (HC) children in a cohort of Warao Amerindians in the Delta Amacuro in Venezuela. We identified a 116-gene signature set by means of random forest analysis that showed an average prediction error of 11% for TB vs. LTBI and for TB vs. LTBI vs. HC in our dataset. Furthermore, a minimal set of only 9 genes showed a significant predictive value for all previously published adult studies using whole blood gene expression, with average prediction errors between 17% and 23%. Additionally, a minimal gene set of 42 genes with a comparable predictive value to the 116-gene set in both our dataset and the previously published literature cohorts for the comparsion of TB vs. LTBI vs. HC was identified. In order to identify a robust representative gene set that would hold stand among different ethnic populations, we selected ten genes that were highly discriminative between TB, LTBI and HC in all literature datasets as well as in our dataset. Functional annotation of these ten genes highlights a possible role for genes involved in calcium signaling and calcium metabolism as biomarkers for active TB. These ten genes were validated by quantitative real-time polymerase chain reaction in an additional cohort of 54 Warao Amerindian children with LTBI, HC and non-TB pneumonia. Decision tree analysis indicated that five of the ten genes were sufficient to diagnose 78% of the TB cases correctly with 100% specificity. We conclude that our data justify the further exploration of our signature set as biomarkers to diagnose childhood TB. Furthermore, as the identification of different biomarkers in ethnically distinct cohorts is apparent, it is important to cross-validate newly identified markers in all available cohorts. In this study, 27 children 1 to 15 years of age with TB (n=9), LTBI (n=9) and HC (n=9) were recruited between May 2010 and December 2010. Tuberculin skin test (TST) and QuantiFERON-TB Gold In-Tube assay (QFT-GIT) were performed on all children. A sputum sample was collected from all children with expectoration and a gastric aspirate was taken from all children under 6 years of age. Children with active TB were diagnosed based on culture of M. tuberculosis (n=2) or on the basis of clinical, epidemiological and radiological features (n=7). The latter group were children with a TST = 10 mm or a positive QFT-GIT result who presented all of: persistent fever >38°C objectively recorded daily for at least two weeks, persistent cough for more than three weeks, weight loss (>5% reduction in weight compared with the highest weight recorded in last three months) or failure to thrive (documented crossing of percentile lines in the preceding three months), persistent lethargy or decrease in playfulness/activity reported by the parent and absence of clinical response on broad-spectrum antibiotics. Standard antero-posterior and lateral chest radiographs (CXRs) were taken from all children. Two independent experts, blinded to all clinical information, evaluated the CXRs and documented their findings on a standard report form. Where the two objective experts disagreed, a third expert was consulted and final consensus was achieved. A diagnosis of TB was only made when the CXR was consistent with TB9 and the child showed a positive clinical response to anti-TB treatment. Children were followed up clinically, radiologically and, in case of a negative TST at inclusion, by means of TST at six and 12 months after inclusion. LTBI was defined as a TST = 10mm and a positive QFT-GIT with a negative culture result on inclusion in the absence of radiological and clinical evidence of TB disease on inclusion as well as on t=6 and t=12 months. HC were children with a TST = 0 mm at inclusion and at t=6 and t=12 months. The HC had a negative QFT-GIT and a negative culture result at inclusion without radiological or clinical evidence of TB disease on inclusion nor on t=6 and t=12 months. TB patients were sampled before initiation of anti-TB treatment. Of three of the nine TB patients, a follow-up sample was taken when the patient was in anti-TB treatment for five months. All children were HIV-negative.

Dataset Information

Multinomial modelling of TB/HIV co-infection yields a robust predictive signature and generates hypotheses about the HIV+TB+ disease state.

Publications

Multinomial modelling of TB/HIV co-infection yields a robust predictive signature and generates hypotheses about the HIV+TB+ disease state.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets