From immunology to artificial intelligence: revolutionizing latent tuberculosis infection diagnosis with machine learning

Abstract

Latent tuberculosis infection (LTBI) has become a major source of active tuberculosis (ATB). Although the tuberculin skin test and interferon-gamma release assay can be used to diagnose LTBI, these methods can only differentiate infected individuals from healthy ones but cannot discriminate between LTBI and ATB. Thus, the diagnosis of LTBI faces many challenges, such as the lack of effective biomarkers from Mycobacterium tuberculosis (MTB) for distinguishing LTBI, the low diagnostic efficacy of biomarkers derived from the human host, and the absence of a gold standard to differentiate between LTBI and ATB. Sputum culture, as the gold standard for diagnosing tuberculosis, is time-consuming and cannot distinguish between ATB and LTBI. In this article, we review the pathogenesis of MTB and the immune mechanisms of the host in LTBI, including the innate and adaptive immune responses, multiple immune evasion mechanisms of MTB, and epigenetic regulation. Based on this knowledge, we summarize the current status and challenges in diagnosing LTBI and present the application of machine learning (ML) in LTBI diagnosis, as well as the advantages and limitations of ML in this context. Finally, we discuss the future development directions of ML applied to LTBI diagnosis.

Keywords: Biomarkers; Differential diagnosis; Latent tuberculosis infection (LTBI); Machine learning (ML); Tuberculosis (TB).

Fig. 1

Bibliometric analysis of studies involved in latent tuberculosis infection (LTBI). In the Web of Science database, the search formula “{[TS = (tuberculosis)] OR [TS = (TB)]} AND {[TS = (latent tuberculosis infection)] OR [TS = (LTBI)]}” was used to retrieve and export the full record results (n = 2724). In addition, CiteSpace 6.2.R2 (64-bit) Basic (https://citespace.podia.com) was used to perform citation-based visualization of the data derived from Web of Science, including: a the research progress map with three years as time slices, b literature clustering based on keywords, and c the distribution of clusters on the timeline. IP-10 interferon protein-10, TS topic, TB tuberculosis

Fig. 2

Overview of the mechanisms of innate and adaptive immunity in response to invasion by MTB in humans. Following MTB recognition by APCs, such as macrophages, antigens are presented to CD4⁺ T lymphocytes via MHC II molecules, and the activated CD4⁺ T lymphocytes will differentiate into Th1 (microenvironment with IFN-γ and IL-12), Th2 (IL-2 and IL-4), and Th17 (TGF-β and IL-23) cells. Th1 cells secrete IFN-γ and facilitate the clearance of MTB, while Th2 inhibits the action of Th1 and can also stimulate the production of antibodies by B cells to kill MTB. Th17 secretes cytokines that recruit neutrophils, macrophages, etc., to play an anti-inflammatory effect. NK cells promote the maturation of APCs, such as DCs, and can activate other immune cells, including macrophages and CTLs. Polarization of macrophages in different cytokine environments results in different immune effects. The complexities of host immunity to MTB highlight the need for further research to better understand the underlying mechanisms of host defense response. APCs antigen-presenting cells, CTL cytotoxic T lymphocyte, MHC II histocompatibility complex II, IFN-γ interferon-γ, IL interleukin, Mφ macrophage, PRR pattern recognition receptor, Th cells helper T cells, TGF transforming growth factor, DC dendritic cell, NK cell natural kill cell, Fas-FasL Fas and Fas ligand, TNF tumor necrosis factor, MTB Mycobacterium tuberculosis, GM-CSF granulocyte–macrophage colony-stimulating factor, M1 type I macrophage, M2 type II macrophage

Fig. 3

Immune signaling pathways involved in MTB infection in vivo. MTB infection triggers immune responses by activating various Toll-like receptors (TLRs) through binding to a range of lipoproteins and lipopolysaccharides. MTB secretes specific antigens (Rv0577, Rv2660c, Rv3875, Rv3628, Rv2873, and Rv1808) that are recognized by TLR2, leading to dendritic cell maturation and the induction of Th1/Th17 response in tuberculosis immunity and inflammatory reactions. Similar to TLR2, TLR4 recognizes MTB antigens (Rv3478, Rv3417c, Rv0440, Rv0652, Rv0475, Rv1009, and 38 kD glycoprotein) in conjunction with dectin-1, resulting in apoptosis of MTB-infected macrophages and the production of IL-17A. Additionally, TLR9 recognizes MTB’s CpG DNA, promoting IFN-α production and regulating the Th1/Th2 balance. TBK1 TANK-binding kinase 1, TIRAP Toll/interleukin 1 receptor domain-containing adaptor protein, TRAF3 tumor necrosis factor receptor factor 3, IRF7 interferon regulatory factor 7, type I IFN type I interferon, TRIF Toll/interleukin 1 receptor-domain-containing adapter-inducing interferon-β, TAK1 transforming growth factor β activated kinase 1, IKKs inhibitor of nuclear factor κB kinases, NF-κB nuclear factor κB, IFN-γ interferon-γ, TNF-α tumor necrosis factor-α, IL interleukins

Fig. 4

Schematic representation of different outcomes and states after MTB infection of the host. The first outcome is active tuberculosis (ATB), where granulomas rupture allowing MTB to multiply in large numbers and enter the alveoli and surrounding tissues, causing the development of ATB. This condition commonly occurs in individuals with a weakened immune system, such as those with HIV infection or receiving immunosuppressive therapy, or in people with impaired immune function due to other reasons. The second outcome is TB elimination, which occurs when the immune response is sufficient to clear the MTB infection. The third outcome is an intermediate state, where MTB becomes dormant and stops replicating when the host can restrain its virulence or when MTB loads are low, leading to an LTBI, incipient TB, or subclinical TB, that may reactivate when the immune system becomes impaired. The upper part of this figure is a modification of Fig. 1 by Drain et al. [128], 2018. MTB Mycobacterium tuberculosis, TB tuberculosis, LTBI latent tuberculosis infection, ALF airway lining fluid

Fig. 5

Evasion of autophagic-lysosomal and phagocytic-lysosomal killing by MTB. The bactericidal process of autophagic-lysosomal degradation involves the formation of autophagic precursors that engulf the infected cells to create autophagosomes which then fuse with lysosomes. This results in the hydrolysis of infected cells by lysosomal enzymes. However, in the presence of toxic MTB, the formation of autophagic precursors is inhibited through the regulation of cytokine production. Additionally, MTB’s lipoproteins LprE can delay the fusion of phagocytic lysosomes by regulating cytokines production, leading to the evasion of phagocytic-lysosomal killing. The phagocytic process involves the engulfment of MTB vesicles by lysosomes containing acid hydrolases that can kill MTB. MTB evades phagocytic-lysosomal killing in various ways. MTB Mycobacterium tuberculosis, IL interleukin, KefB a potassium/proton antiporter in MTB (Rv3236c), aprABC an MTB complex-specific locus, MAPK mitogen-activated protein kinase, CYP27B1 1 alpha-hydroxylase, VDR vitamin D receptor, LC microtubule-associated protein light chain, ATG8 autophagy associated proteins 8

Fig. 6

Schematic representation of machine learning classification. AC autoencoder, CA cluster analysis, DL deep learning, DQN Deep Q-Network, DT decision tree, LIR linear regression, LOR logistic regression, PCA principal component analysis, RF random forest, SARSA State-Action-Reward-State-Action, SVM support vector machine