Design of colorimetric assay for LAM detection

To address the challenge of analyzing urinary LAM below 10 pg/mL for point-of-care TB detection^{34,37,40,41,42}, we applied magnetic force-based enrichment and a colorimetric signal-transducing method. This approach enhances sensitivity and cost-effectiveness by eliminating the need for bulky instrumentation. The assay steps for LAM detection are illustrated in Fig. 2a and Supplementary Fig. 1. The magnetic beads (MB) were conjugated with a LAM antibody (Ab) by activating the amide bond between the carboxyl group on the MB surface and the amine group on Ab using ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC). We evaluated the antibody’s conjugation efficiency to the magnetic bead surface to maximize target capture (Supplementary Fig. 2). The efficiency gradually increased with the treated Ab concentration and became saturated above 0.4 mg/mL. Therefore, we determined the amounts of Ab treated on MB surface as 0.4 mg/mL. Salmon sperm DNA (SSD) was utilized as a blocking agent for preventing nonspecific absorption and 1 mg/mL of the blocker was selected for minimizing background signal that could interrupting detection with naked eye (Supplementary Fig. 3). Subsequently, LAM and an HRP–aptamer conjugate (H-apt) were introduced into the SSD–Ab–MB suspended matrix and a sandwich complex was generated by capturing LAM between SSD–Ab–MB and H-apt. Different amounts of H-apt deriving TMB reaction were explored after the formation of the sandwich complex (Supplementary Fig. 4). The absorbance at a wavelength of 280 nm was gradually increased after the binding of apt on LAM–SSD–Ab–MB complex because of the HRP and the highest value at 200 ng/mL was observed. Therefore, the H-apt of 200 ng/mL was utilized for the detection of LAM. Moreover, 0.5 % Tween 20 in 0.1 M phosphate-buffered saline (PBS) was employed as a washing buffer for effective removal of unbounded H-apt and impurities (Supplementary Fig. 5). After removing the supernatant, the complex was suspended in the TMB solution. In the presence of HRP, TMB was oxidized to oxTMB, resulting in a color change from transparent to blue. The optimal reaction time was determined as 5 min generating the largest difference of color (Supplementary Fig. 6). Finally, the reaction was stopped by isolating the complex using magnetic force from the TMB solution and the LAM concentrations was semi-quantitatively measured by analyzing the degree of color change and absorbance of 650 nm wavelength.

a Schematic representation for generation of MB complex and TMB reaction for LAM detection. b Characterization of different assay steps of MB surface via UV–vis spectra absorbance at 260 nm, 280 nm, 650 nm. TEM image of (c) MB, (d) SSD–Ab–MB and (e) H-apt–LAM–SSD–Ab–MB using Ab and apt labeled with AuNPs with sizes of 60 and 20 nm, respectively (inset shows the corresponding high-magnification TEM images). f Calibration curves of UV-vis spectra at different concentrations of LAM in 0.1 M PBS (pH 7.4) and mixed urine (n = 20) at 650 nm, respectively. The LAM concentrations detected ranged from 10⁻² to 10³pg/mL. g Specificity of the assay for LAM detection. Concentrations of 1.0 × 10²pg/mL for ESAT-6, IFN-γ, IL-2, TNF-α and 1.0 × 10⁻¹pg/mL LAM were utilized. h Selectivity of the assay for LAM detection. Concentrations of C. albicans, S. aureus, B. cereus, E. coli, S. enterica, P. putida, S. aeruginosa lysates are 1.0 × 10⁷pg/mL and 1.0 × 10²pg/mL M.tb lysate were utilized. i Specificity of the assay against cross-reactive. Concentration of 0.1 mg/mL for IgG, 3 mg/mL for glucose, 3 mg/mL for BSA, 0.3 mg/mL for albumin and 0.01 mg/mL for hemoglobin were utilized. j Specificity of the assay against endogenous interference. Concentration of 0.5 mg/mL for uric acid, 15 mg/mL for urea, 3 mg/mL for creatinine, 2.3 mg/mL for KCl, 1.26 mg/mL for NH₄Cl, 1.75 mg/mL for NaCl, 1.75 mg/mL for CaCl₂ were utilized. k Specificity of the assay against exogenous interference. Concentration of isoniazid, rifampicin and paracetamol were 30 µg/mL, while the ibuprofen was 5 µg/mL. A concentration of 1.0 × 10¹pg/mL LAM was used in all interference tests (i–k). All data points and error bars correspond to average values and standard deviation obtained from four independent measurements. Data are presented as means ± standard errors of the mean. In the graph, asterisks indicate statistical significance as follows: ∗p  < 0.05, ∗∗p  < 0.01, and ∗∗∗p  < 0.001. Statistical analysis was conducted using a paired t-test.

Characterization of assay steps by exploring the different states of MB surface

After establishing the optimal assay conditions, we characterized each step of the assay by investigating the UV-vis absorption spectra (Supplementary Fig. 7, and Supplementary Fig. 8) and quantifying the absorbance at wavelength of 260, 280, and 650 nm, indicating the amount of nucleic acid, protein and oxTMB, respectively. The sequential treatment with Ab, SSD and H-apt led to increase of the absorbance at 260 and 280 nm owing to an increase in the number of resonance structures of purine, pyrimidine bases and aromatic chains on the amino acids such as tryptophan and tyrosine residues, respectively^43,44 (Fig. 2b). The blue color (absorbance peak at 650 nm) was shown due to the oxTMB generated from the catalytic reaction⁴⁵. The UV–vis spectra indicated that the colorimetric assay for LAM detection was successfully achieved.

Furthermore, transmission electron microscopy (TEM) images were obtained to identify the immobilization of Ab on the MB surface and the formation of the sandwich complex. Due to the significant size disparity between the MBs (1 – 4 μm) and the bioreceptors (Ab: ~10 nm, aptamer: ~2 nm), distinguishing the bioreceptors on the MB surface in TEM images was challenging (Fig. 2c). To address this visualization challenge, gold nanoparticles (AuNPs) with sizes of 60 nm and 20 nm were conjugated to Ab and apt, respectively, for use as labels (Supplementary Fig. 9). The labeled bioreceptors were sequentially treated with bare MBs and LAM–SSD–Ab–MB complex. TEM images show the immobilization of Ab on the bead (Fig. 2d) and the formation of the complex (Fig. 2e) were successfully accomplished based on the presence of AuNPs on the bead surface. Additionally, the conjugated MB complexes were evaluated using dynamic light scattering (DLS) and zeta potential analysis to verify their structural stability (Supplementary Fig. 10).

Interference inhibition in assay of urine samples

We encounter the issue of derivation of high non-specific signal on the LAM assay in the case of using real urine samples. To address the issue, the materials that used in this study potentially leading to non-specific signal were investigated. Each step of MB complex treated with urine were monitored for the effect on non-specific signal. The color change occurred exclusively after the apt treatment step, confirming that it was due to the H-apt. (Supplementary Fig. 11). Consequently, to investigate the selectivity of Ab for substances other than LAM and the nonspecific adsorption of HRP to the bead surface due to urine components, we measured the changes in absorbance over time. (Supplementary Fig. 12). The difference in the rate of transition from TMB to oxTMB for PBS and urine indicates that the specificity of Ab is not a major cause of nonspecific signal (Supplementary Fig. 12a). Furthermore, the HRP non-conjugated with apt did not adsorbed on MB complex (Supplementary Fig. 12b). Therefore, we suspect that a certain substance in urine boosts the non-specific binding of apt to the bead surface. Next, we focused on the component of LAM with analogs. According to previous reports, LAM is a lipopolysaccharide that is primarily composed of sugar compounds, including arabinogalactan, mannose, and fatty acids⁴⁶. The utilized Ab is reported to bind to the branched sections of LAM, specifically with Ara4 and Ara6 capped with Manp, and apt binds to mannan area^47,48. As observed in previous experimental results, the H-apt nonspecifically binds to the surface of bead complex when actual urine is applied, leading to increased background noise. Based on this observation, it was deduced that certain substances in urine bind to the bead surface and also, the H-apt cross-reacts with these specific substances in urine. Compounds such as Mannan, N-acetyl-D (+)-glucosamine, and lactose, displaying glycan motifs similar to ManLAM, are known to participate in competition during the binding process of the apt⁴⁹. Furthermore, the aforementioned sugar components are recognized to be present in urine in the following concentrations: 10–200 ng/mL glucosamine, 500–700 ng/mL mannose, 670–710 mg/L lactose, and 100–300 mg/L glucose^{50,51,52,53,54}. Until now, no blocking agents have been reported to prevent the nonspecific adsorption of sugar compounds, leading us to select Concanavalin A (ConA) as a candidate to inhibit interference. The protein is known to bind specifically to the D-mannosyl and D-glucosyl groups of sugar, glycoprotein, and glycolipids⁴⁸. Furthermore, the dissociation constant for each bioreceptor and ConA towards LAM showed that the two bioreceptors were 7–9 times more capable of effectively binding to the target substance compared with ConA, indicating their potential utility as inhibitors (Supplementary Table. 3). Based on this knowledge, we expect that ConA can prevent the non-specific adsorption of interferents and also hinder the binding of the H-apt (Supplementary Fig. 13a). To evaluate the feasibility of ConA as an inhibitor, LAM was added to pooled urine samples from TB-negative patients (n = 20) with various concentrations of ConA. The absorbance of the blank and the LAM of 1 ng/mL spiked in urine with ConA exhibited in Supplementary Fig. 13b. The blank and the target signal could be distinguished due to the ConA as an inhibitor in the range of 0.1–100 ng/mL. The discrepancy mostly expanded after treatment with 10 ng/mL of ConA, displaying that non-specific signals were effectively suppressed. Therefore, the ConA of 100 ng/mL was utilized as an inhibitor for quantifying LAM in a real urine sample.

Evaluation of assay performance

In order to eliminate the interference of non-specific signal in urine and maximize the assay performance of LAM in urine, the adjustment of MB concentration, various blocker, buffer for H-apt with various pH were conducted and explore the effect of pH on LAM detection (Supplementary Fig. 14, and Supplementary Fig. 15, Supplementary Fig. 16). As a result, the MB of 1 mg/mL (5.0 × 10⁷ particles/mL), SSD and carbonate-bicarbonate buffer (pH 9.6) shows large signal difference between blank and the target. Subsequently, the assay performance was evaluated by conducting absorbance measurement on serially diluted LAM in pooled urine added with 100 ng/mL of ConA in 0.1 M PBS. In addition, storage at 4 °C maintained the functional activity of the magnetic bead complex (SSD-Ab-MB) and H-aptamer for up to 14 days (Supplementary Fig. 17). Under the optimized experimental conditions, the analytical performance of the designed assay was evaluated by conducting absorbance measurement on serially diluted LAM in 0.1 M PBS and actual urine matrix. The absorbance peak obtained from the UV–vis spectra gradually decrease with declining concentrations of the LAM (Supplementary Fig. 18). Figure 2f shows the absorbance change is linearly related to the logarithms of each LAM concentration in 0.1 M PBS (R² = 0.99) and pooled urine (R² = 0.98). The limits of detection (LOD) of the constructed assay in 0.1 M PBS was estimated as 2.3 × 10⁻³pg/mL and in actual urine matrix was 3.7 × 10⁻³pg/mL with respect to the sum of the change in the current of the blank sample and three standard deviations. The LOD value implies that trace amounts of LAM can be detected in the urine of HIV-negative/TB-positive patients. In addition, as shown in Supplementary Fig. 19, the absorbance at 650 nm increased proportionally with LAM concentration from 1.0 × 10⁻² to 1.0 × 10³ pg/mL. Beyond this concentration, the signal reached a plateau, indicating saturation of the colorimetric response. Therefore, the detection range in this study was defined as 1.0 × 10⁻² to 1.0 × 10³ pg/mL, ensuring accurate quantification within the linear response region.

Evaluation of specificity and selectivity

To evaluate assay specificity, the signals for LAM at 1 × 10^-1pg/mL were compared with responses for four TB-related biomarkers at 100-fold excess concentrations (1 × 10²pg/mL)^{48,51,52,53,54,55,56,57,58}. These interferent biomarkers, even at 100-fold higher concentrations, produced signals nearly identical to the blank (Fig. 2g). Furthermore, LAM noticeable signal alterations of higher magnitudes, effectively separating it from interference. The signal differences were statistically significant (***p < 0.001) highlighting the excellent specificity of the assay for detecting LAM even in the presence of potential interference. To assess the selectivity of the assay in urine, the signals corresponding to 1.0 × 10²pg/mL of M. tb lysate were compared with those obtained from 10⁵-fold excess concentrations (1.0 × 10⁷pg/mL) of lysates from seven bacteria commonly known for infecting human. The lysates of the bacteria were added to pooled urine and then utilized in the test. In pooled urine, bacterial lysates generated signals close to blank, whereas M. tb lysate induced a marked signal increase (Fig. 2h). This significant distinction (***p < 0.001) demonstrates the assay’s capability to selectively detect M.tb in urine, even with high concentrations of potentially interfering bacteria.

Furthermore, to comprehensively evaluate cross-reactivity and assay robustness, abundant urinary proteins and metabolites such as IgG, glucose, BSA, albumin, and hemoglobin were tested (Fig. 2i). Although these molecules are not structurally related to LAM, they are frequently present in urine and may cause nonspecific binding or affect the colorimetric readout. To further assess assay robustness, we evaluated potential interference from endogenous substances (Fig. 2j) (uric acid, urea, creatinine, KCl, NaCl, NH₄Cl, CaCl₂) and exogenous compounds (Fig. 2k) (isoniazid, rifampicin, ibuprofen, and paracetamol), which may be present in urine due to metabolic activity or pharmacological treatment. All substances were spiked into artificial urine samples, and no significant interference with the assay signal was observed under standard assay conditions.

Design and operation of LIC

The device was consisted with two main modules. One is assay operation module sequentially transporting the sample and reagent at a certain point of time. The other is power control module for revealing sequential transportation of liquid at desired time by inducing rotation of reaction cylinder (Fig. 3a). We incorporated the steps of the assay into a handheld device to realize the sample-to-answer detection of LAM in a short time (Fig. 3b). The combination of these two modules in this system could realize handheld type POC based detection of LAM and finally, we developed LIC. The magnetic-force-based enrichment and direct visual readout from the catalytic reactions of TMB with HRP were followed by the rotation of the reaction cylinder (Fig. 3c, and Supplementary Fig. 20, Supplementary Fig. 21)). After a user introduced the reaction mixture containing a urine sample into the sample chamber, the cartridge was activated by winding the switch on the device. The gear system delivered constant power for rotating the reaction cylinder. During the rotation of the cylinder, the reaction mixture containing the MB sandwich complex with LAM was introduced into the reaction channel under gravity, and the complex was enriched. Subsequently, the washing buffer flushed out unbound H-apt and other irrelevant materials. Next, the TMB solution was added to the MB complex attached to the channel, and the HRP on the complex triggered the oxidation of TMB to oxTMB. Finally, the reactant was separated from the complex by transferring the solution into the detection chamber (Supplementary Fig. 22, and Supplementary Table. 4, Supplementary Video). We detected the presence of LAM by comparing the color changes in the detection and control chambers through visual observation. Furthermore, we minimized the unit cost of the device by making the assay operation module disposable and detachable from the combination of the converter and power control module. Therefore, the unit cost for test could be minimize facilitating the on-site detection of TB, particularly in underdeveloped regions (Supplementary Fig. 23, Supplementary Fig. 24).

a The main components of LIC and (b) dimensions of the cartridge. c Schematic illustration of workflow for the detection conducted sequentially in assay operation module. The LAM detection could be started by adding reaction mixture into sample chamber from collection tube containing urine sample. The cartridge operation is initiated by winding up the switch and the gear system deliver constant power for rotating reaction cylinder. The rotation of the cylinder induces the enrichment of MB complex at the bottom of the channel located in the cylinder. Subsequently, the wastes and unbound apts are flushed out by using washing buffer transported from gravity. Next, the TMB reaction is derived by delivering TMB solution in the channel and it trigger oxidation of TMB. The blue color deepens as the number of oxTMB molecules increases. After the reaction cylinder is rotated for 5 min, the TMB solution is transported to the detection chamber to separate it from the complex. d Rack gear force at different angles of main gear. Each data point represents the mean ± SD of five independent technical replicates (n = 5), with each replicate corresponding to a distinct device. Error bars indicate standard deviation (SD). Measurements were collected at multiple angles, with each device as the independent experimental unit. e Angular velocity at different angles of the reaction cylinder. Data points show mean ± SD from five independent replicates (n = 5), each performed on a distinct device. Error bars indicate SD. Measurements were acquired at multiple time points, with each device as the independent experimental unit. f Leakage rate and contamination rate for operating cartridge by using various sealing methods. Data points show mean ± SD from five independent replicates (n = 5), each corresponding to an independent sealing assembly. Error bars indicate SD.

Investigation of fluid movement in LIC

To maximize the washing effect in the channel of reaction cylinder, we employed a two-stage cleaning strategy and investigated the flow velocity driven by gravity at channel in reaction cylinder by using computational fluid dynamics simulations (Supplementary Fig. 25, and Supplementary Fig. 26). The enrichment of MB complex takes place at the bottom of the channel where the magnet is located, leading to a stronger washing effect at the site of enrichment as the velocity of the fluid passing through the channel increases. Therefore, the flow velocity distribution across different channel widths was investigated, and the simulated result showed that wider channels lead to weaker forces of surface tension acting on the flow of the fluid, thereby accelerating the flow velocity. Furthermore, to enhance slipperiness by increasing the hydrophobicity of the channel walls, silane was applied at varying concentrations, and the contact angle was analyzed (Supplementary Fig. 27). The contact angle became saturated at 0.5 % silane treatment, also it led to remove the fluid retention rate on channel wall (Supplementary Fig. 28). Additionally, to evaluate the mechanical functionality of the engineered POC tool, we investigated the consistency of rack gear force and angular velocity in the reaction cylinder (Fig. 3d, e) and confirming both the rack gear force and angular velocity displayed gradients within the permissible tolerance ranges at various angles. Besides, in order to prevent fluid leakage and cross-contamination of reagents between chambers during the operation of LIC, four types of sealing methods were compared (Fig. 3f, and Supplementary Fig. 29, Supplementary Fig. 30). Reagent leakage did not occur during the operation when capillary slots and polydimethylsiloxane (PDMS) seals were used. However, only PDMS seals showed no leakage and contamination rates owing to their elasticity, which enabled them to fill the small gaps between the surfaces.

Analytical performance of LIC

The internalization and automated operation of the developed assay using the LIC are demonstrated by employing pooled urine samples. To enhance the applicability of the LIC for decentralized testing, pooled urine samples were initially pretreated through syringe filtration, which allowed for the removal of excessive interference. The preprocessed samples were incubated for 10 min with reagents containing H-apt, functionalized MB, and ConA. Subsequently, they were injected into the cartridge, and when the device operation was completed, we assessed the results shown on the target chamber in the detection part. As observed in the detection part of the LIC, the amount of oxTMB transferred to the target chamber increased with the LAM concentration, resulting in a deeper blue color (Fig. 4a). The color intensity in the control chamber was engineered such that the threshold value of the absorbance obtained from the assay of urine samples was 0.16 OD. In addition, the analysis of the reactant absorbance in the target chamber showed a linear range. Notably, the assay for LAM detection showed lower standard deviation values compared with manual assays (Fig. 4b). This indicates that the capacity of the device for automatic detection could reduce signal variability, which probably occurred during manual testing, while improving accuracy (Fig. 4c, d). Intra- and inter-assay variations were examined to determine the reproducibility of the biosensor. Under optimal conditions, the relative standard deviation (RSD) was evaluated four times using LAM of 1.0 × 10⁻²pg/mL. The RSDs of the intra- and inter-assay variations were estimated to be 4.60 % (n = 4) and 3.84 % (n = 4), respectively (Fig. 4e). To further confirm the robustness and consistency of the LIC system, Supplementary Fig. 32 presents a lot‑to‑lot variability assessment. To evaluate batch consistency, three independent lots were tested using LAM samples at concentrations of 0.01, 1, and 100 pg/mL (n = 10 per lot). The absorbance values showed no significant differences among the batches, with coefficients of variation below 5%, indicating the robust and reproducible performance of the LIC. Hence, the obtained results endorse the reliable repeatability and reproducibility of the proposed strategy.

a Schematic drawing and image of detection part at the end of LIC operation for different concentrations of LAM (1.0 × 10⁻² – 1.0 × 10³pg/mL) in pooled urine (n = 20). The generated color of control chamber designed for representing threshold value of the blank (0.16 OD). b Corresponding calibration plot of oxTMB absorbance extracted from LIC. Dashed line represents a blank value. c Comparison for repeatability and (d) reproducibility of manual detection and utilization of LIC by using 1.0 × 10⁻² pg/mL of LAM. Boxplots show absorbance values (650 nm) measured by manual assays and LIC. Medians are indicated by center lines; boxes denote interquartile ranges (IQR, 25th–75th percentile), and whiskers extend to the most extreme values within 1.5 × IQR. Each data point represents one independent replicate measurement. The left panel illustrates variability across five test batches, while the right panel shows measurements from five consecutive assay days. e Inter-, intra- analysis of the assay for LAM detection by using LIC. All the data points and error bars correspond to average values and standard deviation obtained from four independent measurements.

Clinical validation of LIC

After confirming the LIC’s utility for applying LAM detection in urine, we explored the practical feasibility of the LIC using clinical urine samples obtained from 60 TB-positive and 40 TB-negative patients. All the patients were HIV-negative, and the urine samples of the TB patients were collected while they underwent microbiological examinations. The LIC exhibits the fastest detection rate from urine samples to answer than other conventional tuberculosis detection methods (Fig. 5a). Figure 5b shows the quantitative results of the absorbance for these clinical samples using the developed device. Although both methods were based on TMB chromogenic reactions, the enzyme-linked immunosorbent assay (ELISA) was only able to yield positive results ( > 0.1 OD) in two out of 60 positive samples and exhibited low absorbance values, whereas significantly stronger signals were observed with LIC (Supplementary Fig. 33). The difference in signals indicated that when compared to the surface area available for capturing biomolecules on ELISA plates, the area of the MB-based assay used inside the LIC is about 20 times larger, leading in more amount of target capture. Additionally, the polymer form of HRP used further induced more oxTMB transitions, thereby secondarily amplifying the signal and contributing to the observed difference in signal intensity. The results from the LIC were verified by comparison with the multiple clinical diagnostic results using ELISA for targeting LAM, mycobacterial cultures, AFB smear result (Supplementary Table. 5). The cut-off of absorbance values was determined as 0.16 OD for distinguishing between non-TB patients and TB patients in order to enhance diagnostic accuracy and specificity and also, this value was almost similar with LOD of LIC. Employing the established cut-off value, LIC was able to distinguish 35 of 40 non-TB patients and 55 of 60 TB patients, with LAM results from 90/100 samples aligning with the clinical diagnosis representing that the LIC play a significant role in auxiliary diagnosis. The clinical samples were further examined with naked eye observing the detection chamber of LIC. The signal difference was observed in urine samples between TB-positive and TB-negative patients (Supplementary Fig. 35). The degree of color changes of urine samples from non-TB patients were lower compared to those of urine samples from TB patients and also lower than the control chamber. These results confirmed that the semi-quantitative differentiation method based on naked-eye observation can also be effectively utilized for the analysis of LAM in clinical urine samples. Furthermore, the positive or negative results of LAM detection by using LIC and ELISA were compared to conventional methods with AFB smear, culture, and RT-PCR results from 40 TB-positive patients and 20 TB-negative patients uninfected from HIV (Fig. 5c). Only 2 positive cases could be discerned by ELISA. However, RT-PCR distinguish all 60 number of TB-positive, negative patients. Also, the observed result of AFB smear and culture shows 31/40 and 38/40 positive, respectively, and also the 20/20 negative result. Furthermore, LIC could determine 38/40 positive and 18/20 negative results presenting the utility of the proposed monitoring tool for TB testing. Moreover, the diagnostic value and application potential of LIC were further investigated in 23 clinical urine samples from 3 HIV, TB co-infected patients, 10 HIV-negative/TB-positive patients, and 10 HIV-negative/TB-negative patients. The results are shown in Fig. 5d and suggest that those who are co-infected had higher levels of LAM detected. Although more number of samples investigations are required for HIV-positive/TB-positive patients due to the low case of samples, the observed results highlight the significance of identifying LAM at low concentrations and throughout a wide range for patients with both HIV-positive/TB-positive and HIV-positive/TB-negative patients. As for the clinical diagnostic efficacy, the result for LAM detection of all the enrolled cases are shown in Fig. 5e. Our LAM detection method shows sensitivity of 92 % (55/60) for confirmed TB (culture-positive) with in non- TB patients of 88 % specificity (35/40) with a cut-off value of 0.16 OD. The receiver operating characteristic curve (ROC) curves are shown in Fig. 5e. The area under the curve (AUC) was 0.91 when the controls were only non-TB patients. Compared to other methods, the method presented in this study was found to possess the highest sensitivity, fulfilling criteria recommended by the WHO’s END TB strategy (sensitivity > 90%, specificity > 70%) (Fig. 5f).

a The detection rates compared between LIC and other conventional detection methods. Created in BioRender. Jung, H. (https://BioRender.com/8rz9zwn.) b Absorbance value of LIC and ELISA. c Heatmap for distinguishing result of ELISA, AFB smear, culture and LIC. d Boxplots comparing the results of LAM test enrolled clinical HIV-positive/TB-positive, HIV-negative/TB-positive, and HIV-negative/TB-negative patients. Medians are indicated by center lines; boxes denote interquartile ranges (IQR, 25th–75th percentile), and whiskers extend to the most extreme values within 1.5 × IQR. Statistical comparisons were performed using two-tailed t-tests. Sample sizes were 3 patients for the HIV-positive/TB-positive group and 10 patients each for the HIV-negative/TB-positive and HIV-negative/TB-negative groups. Each data point represents one patient. The HIV-negative/TB-negative group served as the negative control. e Receiver operating characteristic curve (ROC) of LAM detection and comparison of diagnostic values obtained using ELISA, AFB smear, culture and LIC. f Performance summary of various methods for detecting M.tb.