Project description:Background The discrimination between active tuberculosis (ATB) and latent tuberculosis infection (LTBI) remains challenging. The present study aims to investigate the value of diagnostic models established by machine learning based on multiple laboratory data for distinguishing Mycobacterium tuberculosis (Mtb) infection status. Methods T-SPOT, lymphocyte characteristic detection, and routine laboratory tests were performed on participants. Diagnostic models were built according to various algorithms. Results A total of 892 participants (468 ATB and 424 LTBI) and another 263 participants (125 ATB and 138 LTBI), were respectively enrolled at Tongji Hospital (discovery cohort) and Sino-French New City Hospital (validation cohort). Receiver operating characteristic (ROC) curve analysis showed that the value of individual indicator for differentiating ATB from LTBI was limited (area under the ROC curve (AUC) < 0.8). A total of 28 models were successfully established using machine learning. Among them, the AUCs of 25 models were more than 0.9 in test set. It was found that conditional random forests (cforest) model, based on the implementation of the random forest and bagging ensemble algorithms utilizing conditional inference trees as base learners, presented best discriminative power in segregating ATB from LTBI. Specially, cforest model presented an AUC of 0.978, with the sensitivity of 93.39% and the specificity of 91.18%. Mtb-specific response represented by early secreted antigenic target 6 (ESAT-6) and culture filtrate protein 10 (CFP-10) spot-forming cell (SFC) in T-SPOT assay, as well as global adaptive immunity assessed by CD4 cell IFN-γ secretion, CD8 cell IFN-γ secretion, and CD4 cell number, were found to contribute greatly to the cforest model. Superior performance obtained in the discovery cohort was further confirmed in the validation cohort. The sensitivity and specificity of cforest model in validation set were 92.80% and 89.86%, respectively. Conclusions Cforest model developed upon machine learning could serve as a valuable and prospective tool for identifying Mtb infection status. The present study provided a novel and viable idea for realizing the clinical diagnostic application of the combination of machine learning and laboratory findings. Supplementary Information The online version contains supplementary material available at 10.1186/s12879-022-07954-7. Highlights The first study to establish 28 models using machine learning for TB diagnosis. The first TB diagnostic model based on routine, TB-specific and non-specific tests. Cforest model presented excellent performance in discriminating ATB from LTBI. Supplementary Information The online version contains supplementary material available at 10.1186/s12879-022-07954-7.

Project description:This series regroups different datasets (training set, test set, validation set, longitudinal set, separated cell set) to identify and characterise a specific transcriptional signature for patients with active TB, distinct from patients with latent TB and healthy controls. The training set dataset was used to identify a whole blood transcriptional signature for active TB patients in London, across a range of ethnicity. This signature was then validated in an independent cohort of patients, also recruited in London (the test set), and then further confirmed in an additional independent cohort recruited in Cape Town, South Africa (validation set), in order to confirm that the defined signature was present in both high (Cape Town, South Africa) and medium incidence regions (London, UK). The longitudinal dataset was then used to explore how successful TB treatment modifies this transcriptional signature. The separated cell set compares the transcriptional profiles in purified cell subsets (neutrophils, monocytes and T cells) to assess which cell types are contributing to the whole blood signature, and in what way. These studies may ultimately help to improve the diagnosis of active tuberculosis which normally relies on culture of the bacilli, which can take up to 6 weeks, and sometimes the bacilli cannot be obtained from sputum thus requiring invasive techniques such as bronchoalveolar lavage (BAL). In some cases (30%) the bacill cannot be grown from sputum or BAL. Any diagnostic tool would need to be valid across a range of ethnicities, and be valid in both high and low incidence countries. A further aim was to determine whether latent TB patients have a distinct homogeneous or heterogeneous signature, since it is not currently possible to determine using present tests (Tuberculin skin test - TST - or MTb antigen responsiveness of blood cells to produce IFN-gamma - IGRA assay) whether the mycobacteria have been cleared, are still present but are controlled by an active immune response, or to predict which patients will develop active TB. Defining heterogeneity in the latent TB patients would be an important step in developing diagnostics which could detect those most at risk of developing active TB, and thus enable targeted preventive therapy. The latter situation may be determined if Latent patients have a blood transcriptional signature similar to that in Active patients. The transcriptional signature in whole blood and cell subsets from Active TB patients may also provide information as to the factors leading to immunopathogenesis, thus possibly identifying therapeutic targets. The transcriptional profile in latent TB may give information regarding protective factors controlling the infection, important for vaccine development. Finally, definition of a transcriptional signature which responds to therapy could facilitate the development of surrogate biomarkers for drug or vaccine studies. Since any active TB signature may reflect common inflammatory responses evoked during many diseases, we also performed analysis of significance, comparing transcriptional profiles from patients with TB to those from patients with other bacterial and inflammatory diseases to identify a TB specific signature. The resulting signature was then tested against patients normalized to their own controls from 7 independent datasets: TB (Training and Validation Sets), Staphylococcus infection, Group A Streptococcus infection, Still's disease, and adult and pediatric SLE. This SuperSeries is composed of the SubSeries listed below.

Project description:This series regroups different datasets (training set, test set, validation set, longitudinal set, separated cell set) to identify and characterise a specific transcriptional signature for patients with active TB, distinct from patients with latent TB and healthy controls. The training set dataset was used to identify a whole blood transcriptional signature for active TB patients in London, across a range of ethnicity. This signature was then validated in an independent cohort of patients, also recruited in London (the test set), and then further confirmed in an additional independent cohort recruited in Cape Town, South Africa (validation set), in order to confirm that the defined signature was present in both high (Cape Town, South Africa) and medium incidence regions (London, UK). The longitudinal dataset was then used to explore how successful TB treatment modifies this transcriptional signature. The separated cell set compares the transcriptional profiles in purified cell subsets (neutrophils, monocytes and T cells) to assess which cell types are contributing to the whole blood signature, and in what way. These studies may ultimately help to improve the diagnosis of active tuberculosis which normally relies on culture of the bacilli, which can take up to 6 weeks, and sometimes the bacilli cannot be obtained from sputum thus requiring invasive techniques such as bronchoalveolar lavage (BAL). In some cases (30%) the bacill cannot be grown from sputum or BAL. Any diagnostic tool would need to be valid across a range of ethnicities, and be valid in both high and low incidence countries. A further aim was to determine whether latent TB patients have a distinct homogeneous or heterogeneous signature, since it is not currently possible to determine using present tests (Tuberculin skin test - TST - or MTb antigen responsiveness of blood cells to produce IFN-gamma - IGRA assay) whether the mycobacteria have been cleared, are still present but are controlled by an active immune response, or to predict which patients will develop active TB. Defining heterogeneity in the latent TB patients would be an important step in developing diagnostics which could detect those most at risk of developing active TB, and thus enable targeted preventive therapy. The latter situation may be determined if Latent patients have a blood transcriptional signature similar to that in Active patients. The transcriptional signature in whole blood and cell subsets from Active TB patients may also provide information as to the factors leading to immunopathogenesis, thus possibly identifying therapeutic targets. The transcriptional profile in latent TB may give information regarding protective factors controlling the infection, important for vaccine development. Finally, definition of a transcriptional signature which responds to therapy could facilitate the development of surrogate biomarkers for drug or vaccine studies. Since any active TB signature may reflect common inflammatory responses evoked during many diseases, we also performed analysis of significance, comparing transcriptional profiles from patients with TB to those from patients with other bacterial and inflammatory diseases to identify a TB specific signature. The resulting signature was then tested against patients normalized to their own controls from 7 independent datasets: TB (Training and Validation Sets), Staphylococcus infection, Group A Streptococcus infection, Still's disease, and adult and pediatric SLE. This SuperSeries is composed of the following subset Series: GSE19435: Transcriptional profiles in Blood of patients with Tuberculosis - Longitudinal Study GSE19439: Blood Transcriptional Profiles in Active and Latent Tuberculosis UK (Training Set) GSE19442: Blood Transcriptional Profiles of TB in South Africa GSE19443: Blood Transcriptional Profiles of Active TB (UK Test Set Separated) GSE19444: Blood Transcriptional Profiles of Active and Latent TB (UK Test Set) GSE22098: Whole blood transcriptional profiles of patients with active tuberculosis (TB) and other inflammatory and infectious diseases Active Pulmonary TB: PTB - All patients were confirmed by isolation of Mycobacterium Tuberculosis on culture of sputum or bronchoalvelolar lavage fluid. Latent TB: LTB - All patients were screened at a tuberculosis clinic, being either new entrants to the UK from endemic countries or being household contacts of infectious cases, or in the case of the validation set recruited in South Africa, were residents of a high incidence country. All UK patients were positive by tuberculin skin test (>14mm if BCG vaccinated, >5mm if not vaccinated) and were also positive by Interferon-Gamma Release assay(IGRA); specifically Quantiferon Gold In-Tube Assay (Cellestis, Australia). The South African latent TB patients were all positive by Interferon-Gamma Release assay (IGRA); specifically Quantiferon Gold In-Tube Assay. Latent patients had no clinical, radiological or microbiological evidence of active infection and were asymptomatic. Healthy controls - these were volunteers without exposure to TB who were negative by both tuberculin skin test (<15mm if BCG vaccinated, <6mm if unvaccinated); and IGRA (as described above). Experimental variables : Patient group: Active PTB; Latent TB, Healthy controls (BCG vaccinated and unvaccinated). Ethnicity - a wide range of ethnic groups is represented. The active PTB group incorporates a range of smear positive and smear negative disease and a spectrum of disease extent/severity. Experimental methods: Whole blood was collected into Tempus tubes (Applied Biosystems, Foster City, CA, USA) and stored between -20degrees Celsius and -80 degrees Celsius before RNA extraction. For the training set cohort, and the active TB patients baseline samples in the longitudinal cohort, total RNA was isolated from whole blood using the PerfectPure RNA Blood kit (5 PRIME Inc, Gaithersburg, MD, USA). For the separated cell samples, total RNA was isolated using the Qiagen RNeasy Mini Kit. For all other cohorts Total RNA was isolated from whole blood using the MagMAX 96 well RNA isolation kit (Applied Biosystems, Foster City, CA, USA). Isolated total RNA was then globin reduced using the GLOBINclear 96-well format kit (Ambion, Austin, TX, USA) according to the manufacturer's instructions. Total and globin-reduced RNA integrity was assessed using an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA). Biotinylated, amplified RNA targets (cRNA) were then prepared from the globin-reduced RNA using the Illumina CustomPrep RNA amplification kit (Ambion, Austin, TX, USA). Labeled cRNA was hybridized overnight to Sentrix HT12 V3 BeadChip arrays (>48,000 probes, Illumina Inc, San Diego, CA, USA), washed, blocked, stained and scanned on an Illumina BeadStation 500 following the manufacturer's protocols. Illumina's BeadStudio version 2 software was used to generate signal intensity values from the scans, substract background, and scale each microarray to the median average intensity for all samples (per-chip normalisation). This normalised data was used for all subsequent data analysis.

Project description:AbstractTuberculosis (TB) is one of the leading causes of childhood morbidity and death globally. Lack of rapid, effective non-sputum diagnosis and prediction methods for TB in children are some of the challenges currently faced. In recent years, blood transcriptional profiling has provided a fresh perspective on the diagnosis and predicting the progression of tuberculosis. Meanwhile, combined with bioinformatics analysis can help to identify the differentially expressed genes (DEGs) and functional pathways involved in the different clinical stages of TB. Therefore, this study investigated potential diagnostic markers for use in distinguishing between latent tuberculosis infection (LTBI) and active TB using children's blood transcriptome data.From the Gene Expression Omnibus database, we downloaded two gene expression profile datasets (GSE39939 and GSE39940) of whole blood-derived RNA sequencing samples, reflecting transcriptional signatures between latent and active tuberculosis in children. GEO2R tool was used to screen for DEGs in LTBI and active TB in children. Database for Annotation, Visualization and Integrated Discovery tools were used to perform Gene Ontology enrichment and Kyoto Encyclopedia of Genes and Genomes pathway analysis. STRING and Cytoscape analyzed the protein-protein interaction network and the top 15 hub genes respectively. Receiver operating characteristics curve was used to estimate the diagnostic value of the hub genes.A total of 265 DEGs were identified, including 79 upregulated and 186 downregulated DEGs. Further, 15 core genes were picked and enrichment analysis revealed that they were highly correlated with neutrophil activation and degranulation, neutrophil-mediated immunity and in defense response. Among them TLR2, FPR2, MMP9, MPO, CEACAM8, ELANE, FCGR1A, SELP, ARG1, GNG10, HP, LCN2, LTF, ADCY3 had significant discriminatory power between LTBI and active TB, with area under the curves of 0.84, 0.84, 0.84, 0.80, 0.87, 0.78, 0.88, 0.84, 0.86, 0.82, 0.85, 0.85, 0.79, and 0.88 respectively.Our research provided several genes with high potential to be candidate gene markers for developing non-sputum diagnostic tools for childhood Tuberculosis.

Project description:BackgroundGiven that there is no rapid and effective method for distinguishing active tuberculosis (ATB) from latent tuberculosis infection (LTBI), the discrimination between these two statuses remains challenging. This study sought to investigate the value of nutritional indexes and tuberculosis-specific antigen/phytohemagglutinin ratio (TBAg/PHA ratio) for distinguishing ATB from LTBI.MethodsParticipants were consecutively recruited based on positive T-SPOT.TB results between January 2018 and January 2020. ATB was diagnosed by positive mycobacterial culture and/or positive GeneXpert MTB/RIF, with clinical symptoms and radiological characteristics suggestive of ATB. Individuals with positive T-SPOT.TB but without the evidence of ATB were defined as LTBI. Patients younger than 17 years and undergoing anti-TB treatment were excluded.ResultsA total of 709 (312 ATB and 397 LTBI) and another 309 (120 ATB and 189 LTBI) subjects were respectively recruited from Tongji Hospital (Qiaokou cohort) and Sino-French New City Hospital (Caidian cohort). The level of prealbumin was significantly lower in ATB than in LTBI. With a cut-off value of 139 mg/L, the sensitivity and specificity of prealbumin in distinguishing ATB from LTBI were 50.96% (45.41%-56.51%) and 91.69% (88.97%-94.40%). Meanwhile, TBAg/PHA ratio was found statistically higher in ATB compared with LTBI. If using the threshold of 0.29, the sensitivity and specificity of TBAg/PHA ratio were 65.71% (60.44%-70.97%) and 90.93% (88.11%-93.76%), respectively. Moreover, the combination of prealbumin and TBAg/PHA ratio (obtaining by diagnostic model) yielded better specificity (90.18%, [87.25%-93.10%]) and sensitivity (87.18%, [83.47%-90.89%]), while the clinical utility index (CUI) positive and CUI negative were respectively 0.76 and 0.81. After anti-TB treatment, TBAg/PHA ratio was declined while the level of prealbumin was restored (Wilcoxon test, P < 0.001). Furthermore, the performance of diagnostic model obtained in Qiaokou cohort was confirmed in Caidian cohort.ConclusionsThe diagnostic model based on combination of prealbumin and TBAg/PHA ratio is a rapid and accurate tool for discriminating ATB from LTBI.

Project description:BackgroundTuberculosis (TB) still is a major worldwide health problem, with 10.4 million new cases in 2016. Only 5-15% of people infected with M. tuberculosis develop TB disease while others remain latently infected (LTBI) during their lifetime. Thus, the absence of tests able to distinguish between latent infection and active tuberculosis is one of the major limits of currently available diagnostic tools.MethodsA total of 215 patients were included in the study as active TB cases (n = 73), LTBI subjects (n = 88) and healthy persons (n = 54). Peripheral blood mononuclear cells (PBMCs) were isolated from each patient and the LIOSpot® TB anti-human IL-2 ELISpot assay was performed to test their proliferative response to M. tuberculosis antigens ESAT-6, CFP-10 and Ala-DH. Statistical analysis was performed to define the sensitivity and the specificity of the LIOSpot® TB kit for each antigen used and to set the best cut off value that enables discrimination between subjects with active TB or latent TB infection.ResultsComparing the LIOSpot® TB results for each tested antigen between uninfected and infected subjects and between people with latent infection and active TB disease, the differences were significant for each antigen (p< 0.0001) but the ROC analysis demonstrated a high accuracy for the Ala-DH test only, with a cut off value of 12.5 SFC per million PBMCs and the Ala-DH ROC curve conferred a 96% sensitivity and 100% specificity to the test. For the ESAT-6 antigen, with a best cut off value of 71.25 SFC per million PBMCs, a sensitivity of 86% and specificity of 36% was obtained. Finally, the best cut off value for CFP-10 was 231.25 SFC per million PBMCs, with a sensitivity of 80% and a specificity of 54%. Thus, as for IGRA assays such as Quantiferon and T-Spot TB tests, ESAT-6 and CFP-10 are unable to distinguish LTBI from active TB when IL-2 is measured. On the contrary, the IL-2 production induced by Ala-DH, measured by LIOSpot® TB kit, shows high sensitivity and specificity for active TB disease.ConclusionsThis study demonstrates that the LIOSpot® TB test is a highly useful diagnostic tool to discriminate between latent TB infection and active tuberculosis in adults patients.

Project description:The aim of the project was to define transcriptional signatures in whole blood of TB patients (before drug treatment) and healthy controls to distinguish the signature of Latent and Active TB patients from each other and from healthy controls. This will help in diagnosis of active tuberculosis which normally relies on culture of the bacilli, which can take up to 6 wks, and sometimes the bacilli cannot be obtained from sputum thus requiring invasive techniques obtaining bronchoalveolar lavage (BAL). In some cases the bacill cannot be grown from sputum or BAL. Secondly the aim was to determine whether Latent patients have a homogeneous or heterogeneous signature, one may expect the latter since it is not possible to determine by the present tests (Tuberculin skin test - TST - or MTb antigen responsiveness of blood cells to produce IFN-gamma - IGRA assay) whether the mycobacteria has been cleared, is still present but is controlled, or if patients are recently infected or reactivated and will develop active TB. The latter situation may be determined if Latent patients have a blood transcriptional signature similar to that in Active patients. The transcriptional signature in Active TB patients may also provide information as to the factors leading to immunopathogenesis, thus possibly identifying therapeutic targets. The transcriptional profile in Latent TB may give information as to the protective factors controlling the infection, thus important for monitoring vaccine development. A further aim was to examine the transcriptional blood profile of active TB patients at the time of recruitment (before drug treatment) and then subsequently at specific time points after drug treatment to determine whether the signature is extingsuihed with treatment and when. London is a site of intermediate burden of TB - the study was initiated in London, across a broad range of ethnicities to obtain a robust signature that could be used in different developing countries where there is a high burden TB disease.

Dataset Information

Combination of Blood Routine Examination and T-SPOT.TB Assay for Distinguishing Between Active Tuberculosis and Latent Tuberculosis Infection.

Background

Methods

Results

Conclusions

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets