Integrated horizontal convective PCR system for clinical diagnostics

Abstract

Rapid and accurate disease diagnosis in resource-limited settings is critical for optimizing medical resources and controlling epidemic outbreaks. However, developing a point-of-care detection system that is user-friendly, cost-effective, highly accurate, and exhibits strong specificity remains a formidable challenge. Here, we report an integrated horizontal convection PCR system (IHCS) for clinical point-of-care testing, capable of real-time fluorescence detection. IHCS leverages horizontal thermal convection within a capillary to drive sample flow and temperature cycling for nucleic acid amplification, making it suitable for both home use and on-site rapid detection. Capillary action enables swift sample loading, thereby simplifying the overall procedure, and the horizontal thermal convection overcomes the volume limitations inherent in vertical systems. Validation with 130 clinical samples demonstrated that the system successfully detected HBV, SARS-CoV-2, and influenza A virus within 40 min, achieving excellent sensitivity (>96%) and specificity (100%). Overall, the IHCS integrates microfluidics, thermal engineering, and optical sensing and offers a paradigm-shifting solution for point-of-care diagnostics.

An integrated horizontal convection PCR system enables fast sample loading and real-time fluorescence amplification.

INTRODUCTION

Since its invention in 1985 (1, 2), the polymerase chain reaction (PCR) has gradually become the gold standard for viral diagnosis (3). However, the accuracy of this method relies on large analytical instruments and skilled personnel, which hinders its application in point-of-care testing (POCT) (4, 5). Therefore, there is an urgent need to develop a POCT platform that is user-friendly, rapid, cost-effective, highly sensitive, and specific (6–8). As alternatives to conventional PCR, various nucleic acid amplification methods have been used for disease diagnosis, such as photothermal PCR (9–11), digital PCR (12–14), and continuous-flow PCR (15–17). Digital PCR is essentially a miniaturized version of conventional PCR. However, the complexity of microchannels and photolithography limits its large-scale production, and cumbersome sample loading and thermal cycling control further restrict its miniaturization (18, 19). To address these limitations, continuous-flow PCR systems using simplified “quasi-isothermal” heating strategies have been developed, leveraging thermally driven buoyancy convection for automated thermal cycling (20, 21).

In contrast to conventional thermal cyclers reliant on external temperature modulation, convective PCR (cPCR) leverages thermally driven fluid circulation through spatially defined denaturation (≈95°C), annealing, and extension (50° to 70°C) zones. The self-sustaining flow field uniformly traverses the reaction chamber with optimized flow velocity to ensure complete temperature-phase reactions (22, 23). While dual-heater configurations establish stable Rayleigh-Bénard convection for thermal gradient maintenance (24), recent advances demonstrate that single-heater systems can achieve comparable performance through ambient heat dissipation (25–27). These systems typically use capillary-based reactors where bottom-initiated heating induces upward fluid motion for DNA denaturation, followed by ambient cooling during ascent to enable annealing/extension at the capillary apex, ultimately completing the convective cycle via gravitational return (27, 28). However, prevailing vertical capillary orientations exhibit gravitational dependence, rendering them susceptible to flow instability under mobile operation (e.g., vehicular or outdoor testing) (29). In addition, to facilitate the onset of convection, larger capillary diameters are typically required, which in turn demands greater reagent volumes (26, 27). In contrast, the horizontal convection architecture overcomes the volume constraints of conventional vertical convection systems, representing a critical advancement for practical implementation of POCT.

Although capillary-based continuous-flow PCR offers many advantages, several aspects still require improvement for POCT applications (23). To achieve a user-friendly platform, it is imperative to avoid complex sample-loading procedures. Because of the small inner diameter of capillaries, complete filling using a pipette is challenging (30). Air bubbles are inevitably introduced during sample loading in capillaries sealed at one end. Upon heating, these bubbles expand rapidly, thereby compromising the stability of the thermal cycling process. Hence, developing a rapid and stable sample-loading method is critical. Exploiting capillary forces can facilitate more efficient filling of the capillary (31), thereby enhancing loading reliability. Another challenge in developing POCT devices lies in the quantitative detection of amplified products. For amplified samples, analysis is typically performed by gel electrophoresis (32, 33), capillary electrophoresis (21), colorimetry (34), or fluorescence detection (35, 36). Electrophoresis, as an electric field–based separation method, primarily enables DNA fragment qualification and differentiation, while colorimetry is usually used to determine the results of isothermal amplification detection. In comparison, fluorescence detection offers broader applicability (37, 38), supporting both quantitative and qualitative analyses through flexible detection modes while enabling easy integration to minimize operational steps.

Here, we propose an integrated horizontal cPCR system (IHCS) for clinical diagnostics. The IHCS uses horizontal thermal convection within a capillary to drive both sample flow and temperature cycling for nucleic acid amplification, while using sealed consumables to minimize contamination risk, rendering it suitable for both home use and on-site rapid testing. Despite a detection limit of just 10 copies/μl, IHCS demonstrates superior performance compared to other methods, in particular in detection throughput, specificity, and compact device design (table S1). The IHCS incorporates all essential characteristics required for a POCT system and complies with the World Health Organization (WHO) guidelines for POCT (table S2). The system is characterized by: (i) the use of horizontal thermal convection, which reduces environmental interference without sacrificing amplification efficiency; (ii) rapid and reliable sample loading and dual-end sealing via capillary action, which prevents bubble formation and reduces contamination risk; (iii) a single-end heating approach that simplifies device complexity; and (iv) a modular design that enables flexible assembly to enhance throughput. To facilitate both amplification and detection, we developed an integrated device capable of precise temperature control and real-time fluorescence monitoring. Using hepatitis B virus (HBV) DNA, we characterized the IHCS’s fundamental properties. Validation with 70 clinical serum samples (hepatitis B) and 60 nasopharyngeal swab samples (COVID-19 and influenza A) demonstrated successful detection within 40 min, achieving >96% accuracy and 100% specificity. This cPCR system combines operational simplicity with low cost, demonstrating strong potential for POCT deployment.

RESULTS

Structure and working principle of IHCS

When one end of a horizontally oriented capillary tube is heated, the reduced density of the heated solution drives upward flow. Upon reaching the opposite end, the fluid cools, increases in density, and descends, establishing a counterclockwise Rayleigh-Bénard thermal convection loop (Fig. 1A). Nucleic acid molecules within the sample are transported through distinct temperature zones via this flow, undergoing denaturation, annealing, and extension in sequence. Each complete convective cycle corresponds to one PCR thermal cycle, enabling precise control of amplification cycles by regulating heating duration.

Fig. 1. Schematic illustrations and working principle of IHCS.

(A) Schematic diagram of horizontal cPCR. The sample solution completes one amplification cycle each time it circulates through the capillary. (B) Operational procedure of IHCS. The lysed sample is loaded into the IHCS for reverse transcription, amplification, and detection. (C) Heating component of IHCS. The system enables simultaneous horizontal cPCR for 10 capillaries. (D) The key-sized device. (E) Real-time fluorescence curves of different samples. The sample ID (sID) is derived from the code on the sampling tube used during clinical sample collection. Created in BioRender. Guo, W. (2025) https://BioRender.com/nikb117.

Figure 1B illustrates the operational workflow of the IHCS, which is designed for universal detection of both DNA and RNA viruses from serum or nasopharyngeal swab specimens using a standardized protocol. A straightforward operation enables rapid sample loading and dual-end sealing, ensuring that the liquid remains fixed within the capillary while preventing bubble formation during heating, which could otherwise interfere with the experimental results. The detection process comprises two main steps: reverse transcription (for RNA viruses) and convective amplification. For high-throughput applications, the expanded heating module enables simultaneous amplification of 10 samples (Fig. 1C). The thermal module employs precisely aligned upper and lower heating blocks that maintain stable temperature conditions upon lid closure, ensuring consistent performance for both reverse transcription and subsequent convective amplification. The amplification process commences as soon as the capillary is mounted onto the heater, with real-time results displayed on an integrated screen. Through highly integrated design, the device can be miniaturized to a keychain-compatible size (Fig. 1D). Unlike conventional endpoint fluorescence quantification, which can be unstable due to factors such as sample variability, uneven excitation light distribution, or external vibrations (22), this real-time fluorescence readout minimizes environmental interference, allowing for stable and accurate quantification of the target (Fig. 1E).

Convection performance and parameter optimization

Since most prior studies have relied on vertical thermally driven convection for amplification reactions, we began by comparing Rayleigh-Bénard convection in horizontally and vertically oriented capillaries. Before this comparison, we confirmed mesh independence using volume-averaged flow velocity as the evaluation metric (fig. S1) (39). The simulation model and mesh are illustrated in Fig. 2A. Our analysis revealed that horizontal convection overcomes the volume constraints of vertical capillaries, enabling efficient nucleic acid amplification with substantially reduced reagent volumes (text S1). Subsequently, we used a capillary with a diameter of 1.3 mm as a representative example to compare the flow and temperature field distributions at various tilt angles (Fig. 2, B and C). The results showed that convective strength progressively weakened as the tilt angle increased, with convection nearly disappearing in the vertical orientation. Notably, at tilt angles below 30°, the convective pattern remained largely unaffected, indicating that the IHCS can sustain effective convection even when placed on imperfectly level surfaces.

Fig. 2. Numerical simulation and experimental results of thermal convection.

(A) Structural model (i) and mesh generation (ii) used in the numerical simulation. (B) Flow field distributions at different inclination angles. (C) Temperature field distributions at different inclination angles. (D) Influence of solution length (L) on the flow velocity. The legend denotes different capillary diameters. (E) The relationship between the solution length and the amplification results. (F) Impact of embedding depth on amplification performance. (G) The variation in NA concentration within the capillary after parameter optimization. (H) Changes in dsDNA concentration in capillary when placed horizontally and vertically. (I) The LOD measured using HBV quality control samples.

To optimize the capillary parameters for nucleic acid amplification, we performed numerical simulations to evaluate the average flow velocity within capillaries of various diameters. Figure 2D demonstrates that under constant solution length conditions, average flow velocity increases with capillary diameter. Moreover, for each capillary diameter, there exists a specific liquid length that generates the maximum average flow velocity, which also corresponds to the peak amplification performance. Furthermore, we calculated the time-dependent evolution of the flow and temperature fields (fig. S5). The results show that convection within the capillary stabilizes within approximately 200 s, which has a negligible impact on the overall amplification outcome (40 min). Experimental validation using HBV quality control samples (350 ng/ml) in three glass capillaries of different inner diameters confirmed this trend (Fig. 2E). Postamplification measurements showed nucleic acid concentrations peaking then declining with increasing solution volume, mirroring simulation predictions. Similarly, we calculated the average flow velocities for different liquid volumes and measured the final concentrations, obtaining consistent experimental results (fig. S6). The horizontal convection system we used exhibits relatively slow flow, with a maximum velocity not exceeding 0.4 mm/s. Through experimental observations, we found that the amplification efficiency increases with the capillary diameter. However, when the diameter increased from 1.3 to 1.8 mm, the DNA concentration improved by only 8%, while the required reagent volume increased by 131%. Therefore, we selected capillaries with a 1.3 mm inner diameter and filled them with 10 mm of solution for convection and amplification. We then optimized the embedding depth, defined as the solution column length inserted into the heating block (fig. S7). As shown in Fig. 2F, variations in embedding depth result in changes in the temperature at the end of the capillary, which in turn affect amplification performance (40). Both experimental (initial DNA was 350 ng/ml) and simulation (1 mol/m³) conditions showed peak performance at 5 mm embedding depth, despite differences in concentration units. On the basis of these results, we standardized subsequent experiments using 1.3-mm capillaries with 13-μl reaction solution (10-mm column) and 5-mm embedding depth.

Furthermore, we simulated the concentration profiles of various DNA components during the amplification process (Fig. 2G), with subpanels displaying reaction rate distributions across different regions of the capillary. Combined with the temperature field, it is evident that the reaction predominantly occurs in three distinct regions, corresponding to the three temperature zones involved in denaturation, annealing, and extension. Although the simulation assumed infinite primer availability (precluding plateau phase representation), it effectively described early-stage DNA concentration changes. The reaction proceeds through three characteristic phases: (i) denaturation of double-stranded DNA (dsDNA) into single-stranded DNA (ssDNA), (ii) primer annealing to form ancient DNA (aDNA), and (iii) extension yielding new dsDNA (fig. S8). Accordingly, initial conditions showed negligible ssDNA and aDNA concentrations. As the reaction proceeds, the concentration of dsDNA gradually decreases, while those of ssDNA and aDNA increase. Only after aDNA accumulates to a sufficient level does the concentration of newly synthesized dsDNA begin to rise, ultimately achieving amplification.

For direct performance comparison between vertical and horizontal convective amplification systems, we first optimized capillary parameters (inner diameter and solution height) for the vertical configuration (fig. S9). To quantify amplification efficiency, we simulated nucleic acid concentration dynamics during convective cycling using the established kinetic model (text S2). As shown in Fig. 2H and fig. S10, the gray dashed lines indicate the time required for nucleic acid concentration to double. The results show that vertical convection achieves concentration doubling more rapidly than the horizontal case. However, after 40 min of amplification, the DNA concentration in the horizontal capillary substantially exceeds that in the vertical configuration. The enhanced amplification performance arises from fundamental differences in convective transport characteristics between the two systems. Detailed analysis of the spatiotemporal evolution of flow and temperature fields (fig. S11) revealed that the horizontal configuration exhibits two distinct advantages: (i) reduced convection velocities and (ii) expanded stagnant zones. These hydrodynamic properties markedly prolong dsDNA residence time in the 95°C denaturation region during the initial heating phase. The extended thermal exposure promotes more thorough DNA strand separation, thereby establishing optimal conditions for efficient primer annealing and extension in subsequent amplification cycles. Last, we determined the limit of detection (LOD) for the optimized capillary parameters (Fig. 2I). The horizontal cPCR platform reliably detected samples at concentrations as low as 10 copies/μl, demonstrating high stability and sensitivity.

Performance characterization of IHCS

We systematically characterized the fluid dynamics within the capillary, with three-dimensional flow field distributions presented in Fig. 3A. The flow field modulates the temperature distribution, expanding the effective reaction zones and enhancing amplification efficiency. Using the optical system illustrated in fig. S12, we experimentally tracked particle trajectories under horizontal convection (movie S1) and compared these with simulation results (Fig. 3B), demonstrating strong agreement. Furthermore, we calculated the temperature variation curves of individual particles during convection (Fig. 3C). Although particle motion within the capillary exhibits regular patterns, not every convection cycle passes through all three temperature zones, which is a primary factor affecting amplification efficiency. Nevertheless, simultaneous reactions across different zones compensate for this limitation.

Fig. 3. Flow field distribution in capillary and heating performance of IHCS.

(A) Heating method and fluid flow trajectories. (B) Particle trajectories in horizontal thermal convection. (C) Temperature changes of NA molecules during thermal convective motion. (D) The temperature rise curves in 10 holes on the heating device, showing a high degree of consistency. (E) Infrared top view image of the heating device. (F) Physical drawing of the integrated device. (G) Influence of tilt angle on the Tt value. (H) Vibration isolation performance at a construction site. (I) Amplification performance under different vibration conditions. Created in BioRender. Guo, W. (2025) https://BioRender.com/psbtaqo.

To achieve high-throughput detection, we developed a multicapillary heating system capable of simultaneously processing 10 samples. Temperature profiles during heating were measured (Fig. 3D), with intercapillary temperature variation coefficients calculated (fig. S13). All capillaries exhibited consistent heating, reaching 95°C from room temperature within 100 s, with excellent temperature uniformity. In addition, infrared imaging of the axial temperature gradients (Fig. 3E) confirmed a gradual cooling along the capillary, with the end point temperature around 50°C, consistent with simulation results. The fully integrated device (Fig. 3F) combines precision temperature control with real-time fluorescence readout on an embedded display, enabling seamless operation in diverse settings. Modular assembly is facilitated by plug-and-socket connectors installed at both ends of the heating device (fig. S14). For ultraportable applications, the keychain-sized convection unit features USB-powered operation (compatible with power banks, chargers, or computers) with light-emitting diode (LED)–based temperature indicators (fig. S15). System control and signal processing are managed by an STM32 microcontroller, which synchronizes thermal regulation, fluorescence acquisition, and data display (fig. S16).

We further evaluated the robustness of the IHCS by testing its detection performance across varying inclination angles. Recognizing that precise horizontal alignment may be challenging in field settings, we combined both simulation and experimental results to demonstrate that the IHCS is capable of performing nucleic acid detection within an inclination range of −15° to 30° (Fig. 3G and fig. S17). The experimental results confirm that the amplification performance remains stable within this range, with the maximum variation in time threshold (Tt) values being less than 3%. The establishment of consistent convective flow requires stable thermal conditions, in particular in single heat source configurations. To ensure robust operation in environments with substantial temperature fluctuations, we used a thermostatic case to maintain the external temperature required for stable convection. In addition to stabilizing thermal conditions, the case also serves to protect the device and store consumables. This integrated design, combined with battery-powered operation (fig. S18), considerably enhances the system’s field adaptability while maintaining laboratory-grade performance standards.

To further minimize the impact of environmental vibrations on the IHCS, wave-shaped damping pads were integrated inside the case. Common sources of mechanical disturbance in real-world scenarios include vibrations from construction sites, manual handling, and vehicle transport. We evaluated the isolation performance of the damping system by measuring vibration acceleration before and after isolation (Fig. 3H and fig. S19). The results demonstrated that this passive isolation strategy effectively reduced both the acceleration magnitude and dominant vibration frequency. Subsequently, cPCR experiments were conducted under all three vibration conditions using a DNA concentration of 350 ng/ml. As shown in Fig. 3I, the implementation of vibration isolation enabled reliable amplification and detection even under mechanical disturbance (movie S4). Compared with a quiescent environment, the Tt value increased by only 1 min, indicating that this approach effectively preserves IHCS functionality during handling or transport and enables reliable diagnostics in mobile or field settings.

Capillary-driven self-loading and sealing

Given the limited solution volume (13 μl) in the capillary, rapid sample loading can be challenging. To address this, we developed a capillary action-based self-loading and bilateral sealing method (Fig. 4A). Similar to many portable amplification systems that use mineral oil to seal the reaction mixture and prevent evaporation (10, 26, 41), our approach involves vertically inserting the capillary into PCR buffer and mineral oil mixtures (step i). Under the influence of capillary forces, the solution spontaneously climbs along the capillary wall (step ii), ensuring rapid and bubble-free loading. Subsequent insertion into polydimethylsiloxane (PDMS) (iii) draws a small PDMS volume (iv), which solidifies within 2 min in a 95°C aluminum block, providing robust sealing and solution immobilization.

Fig. 4. Rapid sample loading and capillary performance evaluation.

(A) Schematic diagram of rapid sample addition using capillary action. (B) Height of the aqueous phase in the capillary at different oil contents. (C) Effect of hydrophobic agent concentration on liquid rise height. (D) The process of rapid sample loading using capillary action. (E) No significant decline in hydrophobic performance over 4 weeks. (F) Long-term sealing stability of capillaries after sample loading. (G) Real-time fluorescence curves of LOD measured using IHCS. (H) Agarose gel electrophoresis results for the LOD.

We precisely controlled the liquid column height within the capillary by tuning the contact angle (text S3). Since mineral oil was used to seal the system during heating, the resulting oil layer height inevitably exerted an influence on the aqueous phase (Fig. 4B). To achieve effective sealing without significantly affecting the capillary rise of the aqueous solution, we selected a mineral oil volume of 2 μl, corresponding to an oil layer height of approximately 1.5 mm. The contact angle θ was adjusted through hydrophobic treatment using varying concentrations of AQUAPEL (PPG, USA). As shown in Fig. 4C, the capillary rise height exhibited a linear dependence on hydrophobicity across the 0 to 7% concentration range (movie S2). On the basis of the fitted results, a concentration of 4.25% was selected, yielding a capillary rise height of 10 mm, corresponding to a sample volume of approximately 13 μl. To visualize the loading process, red ink was used as a model fluid (Fig. 4D), demonstrating complete sample loading within 4.5 s, highlighting the method’s simplicity and reliability (movie S3). Furthermore, we evaluated the long-term stability of the hydrophobic treatment (Fig. 4E). The results show that the commercial hydrophobic agent maintained effective water repellency for up to 4 weeks, ensuring consistent sample loading. The dual-end sealing was validated through thermal treatment at various temperatures (fig. S20 and table S3). In addition, a 4-week sealing test confirmed the long-term storage stability of the system (Fig. 4F).

For real-time fluorescence detection, a corresponding detection module for the 10 capillary positions was designed, as shown in fig. S21. The excitation light from LEDs, filtered appropriately, is directed axially into the capillary, while a photodiode positioned directly beneath the capillary detects the radial fluorescence intensity. On this basis, we conducted a comprehensive LOD evaluation on the IHCS. The real-time fluorescence curves (Fig. 4G) show that the system can reliably detect samples with concentrations as low as 10 copies/μl within 40 min. Agarose gel electrophoresis further validated the accuracy of the horizontal convective amplification (Fig. 4H).

Clinical serum assay for HBV

To evaluate the clinical applicability of IHCS, we conducted tests on hepatitis B, COVID-19, and influenza A (Fig. 5A; see table S4 for patient information). For hepatitis B, venous blood was collected from 70 volunteers, and the infection status was independently confirmed by the First Affiliated Hospital of Harbin Medical University (Harbin, Heilongjiang). The research protocol was approved by the ethics committee of the First Affiliated Hospital of Harbin Medical University.

Fig. 5. Clinical serum testing for hepatitis B.

(A) Clinical trial protocol. A total of 130 tests were analyzed by IHCS and conventional RT-PCR. (B) Normalized fluorescence signal levels in 70 tests, including 35 patients with hepatitis B and 35 healthy individuals. The positive threshold was set to five times the negative standard deviation. (C) Tt value of all positive samples (+) is significantly lower than that of negative (−) (****P < 0.0001). (D) Evaluation of the consistency between IHCS and RT-PCR analyses showed positive correlations. (E) Real-time fluorescence curves for clinical testing of HBV. sID represents the serial number of the sample in table S5.

Each sample was analyzed using both the IHCS (following the protocols described in text S4 and movie S5) and a benchtop reverse transcriptase PCR (RT-PCR) system under identical primer conditions, and the normalized fluorescence detection results are shown in Fig. 5B. The results demonstrated that 34 of 35 HBV-positive samples were successfully detected, while all negative samples were correctly identified, yielding a sensitivity of 97.14%. The Tt values obtained from the real-time fluorescence detection of the positive samples were all below 40 (Fig. 5C and table S5). Subsequently, we compared the Tt values (from IHCS) with the cycle threshold (Ct) values of a benchtop RT-PCR system. As shown in Fig. 5D, a strong correlation was observed between the two methods (Pearson’s r = 0.85). Figure 5E presents the real-time fluorescence curves for a subset of the detection results. For samples that were not detected, the corresponding viral load was 2 copies/μl according to the clinical data, which is below the LOD of the IHCS.

Nasopharyngeal swabs clinical testing

To demonstrate the IHCS’s suitability as a home-use POC device, we evaluated its performance with nasopharyngeal swab samples, which offer easier collection compared to blood samples. We collected nasopharyngeal swab samples from 60 volunteers, and those with influenza A infection was independently confirmed by First Affiliated Hospital of Harbin Medical University (Harbin, China), and COVID-19 status was verified by the Center for Disease Prevention and Control (Harbin, China). The results obtained using the IHCS were compared with those from the hospital’s clinical PCR system (table S6).

Analogous to the HBV assay, the measured fluorescence values were standardized (Fig. 6A), with four vertical color blocks representing the results from the same volunteer and RPP30 served as the positive control for RNA extraction. For severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the IHCS successfully detected the virus in 28 of 30 COVID-19 samples by identifying both the N1 and N2 genes. Two samples yielded indeterminate results, as only one of the two target genes was detected (Fig. 6B), with the detection sensitivity was 96.67% (58 of 60). The initial viral loads of the two undetected samples were both below 10 copies/μl, which is lower than the LOD for IHCS. A comparison between RT-PCR Ct values and IHCS detection outcomes revealed a strong positive correlation for both genes (Fig. 6C). Moreover, the receiver operating characteristic (ROC) curve analysis indicated that the IHCS exhibited high specificity of 100% (60 of 60; Fig. 6D). On the other hand, regarding influenza A virus detection, all 30 positive samples were correctly identified, and their Tt values were consistently below 40 (Fig. 6E). The detection results also showed a strong positive correlation with RT-PCR Ct values (Fig. 6F). No false-positive results were observed among the negative samples, further confirming the high specificity of the IHCS (Fig. 6G).

Fig. 6. Detecting nasopharyngeal swab samples with IHCS.

(A) Standardized fluorescence levels of 30 patients with COVID-19 and 30 patients with influenza A. (B) Fluorescence detection results of COVID-19 showing notable differences between positive and negative samples (****P < 0.0001). (C) Evaluation of analytical concordance between IHCS and RT-PCR. The results showed a positive correlation. (D) ROC curves for COVID-19 testing. Since one sample of both the N1 and N2 genes was not successfully detected, their ROC curves are identical. (E) Fluorescence detection results of influenza A (****P < 0.0001). (F) Comparison of IHCS results with RT-PCR Ct values. (G) ROC curves for influenza A testing. (H) COVID-19 testing with key-sized devices. (I) Fluorescence detection on IHCS. (J) Evaluation of analytical concordance between key-sized devices and RT-PCR.

Last, we used the keychain-compatible convective amplification device for SARS-CoV-2 detection, which enables simultaneous amplification and detection in two capillary tubes (Fig. 6H). Although real-time fluorescence monitoring is not implemented on this miniaturized platform due to spatial constraints, such monitoring is not strictly necessary for nucleic acid testing, especially in qualitative diagnostic applications (3, 11). End point detection, which evaluates fluorescence after the reaction is complete, enables clear visual interpretation of results without the need for external benchtop instruments (fig. S22). Although the results on the key-sized device can be directly interpreted by the naked eye, we used the IHCS to assist in acquiring quantitative fluorescence data to compare the consistency between the key-sized device and commercial PCR assays. The fluorescence intensities of the N1 and N2 genes in clinical samples are shown in Fig. 6I. To better illustrate the linear relationship between fluorescence intensity and the benchtop RT-PCR system, we applied a logarithmic transformation [expressed as −log₂(Ct)], which revealed strong consistency between the two methods (Fig. 6J and fig. S23). Compared to existing rapid antigen tests (42), the key-sized device demonstrated strong concordance with commercial PCR results even at low viral loads (Ct > 25), further underscoring the advantages of this horizontal convection amplification approach.

DISCUSSION

Our results demonstrate that IHCS enables the detection of multiple types of clinical samples, including blood and respiratory specimens, facilitating rapid and reliable diagnosis in nonclinical settings. Compared to respiratory infections, blood-borne diseases such as HIV, syphilis, and hepatitis B generally exhibit longer incubation periods (43). Although serum-based nucleic acid testing requires additional processing steps, including blood collection, serum separation, and nucleic acid extraction, blood biomarkers remain of high clinical relevance. While many laboratory-based approaches have been proposed for molecular detection of blood-borne pathogens, their reliance on complex instrumentation, multistep procedures, and high operational costs limits POCT suitability (44–46). IHCS overcomes these limitations through a horizontally driven thermal convection strategy powered by a single heat source. This design offers several key advantages: (i) By leveraging horizontal thermal convection, IHCS minimizes environmental perturbations while maintaining amplification efficiency. (ii) Capillary action enables rapid sample loading and efficient dual-end sealing, enhancing thermal stability, minimizing contamination, and preventing high–viral load samples from interfering with adjacent tests. (iii) The single-end heating configuration simplifies the device architecture, making it suitable for household infectious disease screening, with a testing cost as low as $1.40 and a detection limit comparable to standard hospital-based assays. (iv) In a clinical evaluation involving 130 patient samples, IHCS achieved accurate detection of HBV, SARS-CoV-2, and influenza A virus, yielding a diagnostic sensitivity of 96.84% (92 of 95) and a specificity of 100%.

Although thermally driven cPCR has been explored for disease detection, such as for classical swine fever (25) and methicillin-resistant Staphylococcus aureus (26), these systems typically rely on vertical thermal convection. Similarly, while various studies have proposed POC devices based on this method, most lack validation with clinical samples (21, 24, 47). A key innovation of IHCS lies in its implementation of horizontal thermal convection, ensuring more stable flow dynamics and efficient amplification with smaller sample volumes. This configuration enhances robustness under variable environmental conditions. In contrast, photon-based PCR platforms often sacrifice multiplexing capability for sensitivity (9, 11), limiting their scalability and practical utility in large-scale outbreak scenarios.

An economically efficient diagnostic platform like IHCS has the potential to reshape the molecular diagnostics market. Commercial PCR systems, with costs exceeding $19,000 per unit and approximately $8 per test (48), are typically restricted to well-resourced facilities. In contrast, IHCS can be manufactured for just $98, with an estimated per-test cost of $1.40 (table S7). Notably, a notable portion of the cost arises from the optical filter system ($42), which could be greatly reduced through large-scale production using cost-effective alternatives, further improving affordability and scalability.

As pathogens evolve and global health threats intensify, there is a pressing need for adaptable diagnostic systems (49). Simultaneously, changing environmental and socioeconomic conditions demand ongoing adaptation of diagnostic technologies. Unlike traditional POC systems that depend on sealed, proprietary microfluidic cartridges, IHCS features a modular design that can be readily reconfigured to address emerging challenges. Its streamlined architecture supports rapid switching of reagents using standardized, commercially available packaging formats, expanding its applicability to diverse specimen types such as stool or tissue. This flexibility is critical for responding to emerging diagnostic demands driven by changing environmental and socioeconomic conditions.

The COVID-19 pandemic underscored the importance of rapid and decentralized diagnostics for infection surveillance and public health response. Despite the use of capillary action to simplify sample loading, RNA virus detection still requires a reverse transcription step, necessitating two pipetting steps per assay. Further integration and automation of these processes will be essential to enhance user-friendliness. Future development could include integration with digital health platforms (fig. S24) to enable real-time result sharing and remote monitoring. Such connectivity positions IHCS as a foundational tool within a broader, cloud-based diagnostic ecosystem, facilitating the early, accurate, and scalable detection of infectious diseases worldwide.

MATERIALS AND METHODS

Simulation of thermal convection and PCR amplification

We performed simulations of thermal convection and the PCR process using COMSOL 6.3 (data S1). The capillary contains mineral oil, water, and PDMS. By applying single-end heating to capillaries of various diameters and lengths, we calculated the flow velocity and temperature distribution within the capillary. The PCR amplification process within the capillary was simulated under both vertical and horizontal orientations to compare their differences (text S2). The material properties and boundary conditions are detailed in table S8.

Characterization of flow field

The flow field distribution within the sample solution was characterized using 10-μm polystyrene fluorescent microspheres. A mixture of deionized water and fluorescent microspheres was loaded into the capillary, and the flow field was visualized using a lateral microscope under the optical setup illustrated in fig. S12. Fluorescent particle tracing images were captured to analyze flow patterns.

Reaction conditions

HBV quality control samples were purchased from Kangchesitan Biotechnology (Beijing, China), and the DNA Maker was obtained from Takara (Beijing, China). The primers used in this study are listed in table S9, and all primers were synthesized by Sangon Biotech (Shanghai, China).

For the evaluation of the LOD, gradient-diluted HBV quality control samples were added to an Eppendorf tube containing 25 μl of a reaction mixture, and then the mixture was drawn into a capillary for amplification analysis. The reaction mixture consisted of 0.25 μl of Z-Taq (Takara, China), 2.5 μl of 10× Z-Taq buffer, 2 μl of dNTP mixture (each at 2.5 mM), 1.5 μl of primer mixture (forward, reverse, and probe, each at 0.2 μM), and 18.75 μl of nuclease-free water. The probe is a dual-labeled TaqMan probe with FAM at the 5′ end and BHQ1 at the 3′ end, and the detailed sequence is provided in table S9. The thermal cycling protocol was set at 95°C for 40 min, with fluorescence detection performed every minute, and the results displayed on a screen.

For clinical RNA virus detection, the lysed samples were first subjected to reverse transcription (RT). The RT reaction mixture included 5 μl of 4× all-in-one qRT SuperMix (Vazyme R433, China), 4 μl of RNA template, and 11 μl of nuclease-free water. Approximately 6 μl of the solution was drawn into a capillary and sealed at both ends. Since the RT is typically carried out under isothermal conditions, the small volume ensures that the entire solution is within the heated block, thereby preventing the formation of thermal convection. Following the manufacturer’s instructions, the solution was heated at 50°C for 5 min and then at 85°C for 30 s before being removed for subsequent cPCR operations.

Design and assembly of integrated device

We designed the 3D structure of the device using SolidWorks and fabricated the device housing using 3D printing. The housing was printed in nylon (PA12, JLCPCB) to ensure that it could withstand the high temperatures during operation. The device is controlled by an STM32F103C8T6 microcontroller, and the printed circuit board was produced by JLCPCB. For fluorescence detection, a blue LED was used for excitation, and a photodiode (VEMD8080, VISHAY, USA) was used to collect the fluorescence. Both the excitation filter (transmitting light between 400 and 500 nm) and the emission filter (transmitting light above 510 nm) were purchased from PHTODE (Beijing, China).

Rapid sample loading via capillary action

The sealing mineral oil was purchased from Sigma-Aldrich, and the PDMS base and curing agent (Sylgard 184) were supplied by Dow Corning (USA). To facilitate field use, we adopted a simplified PDMS mixing and degassing procedure (fig. S25).

Glass capillaries were cut to a length of 20 mm using a glass cutter. Each capillary was then immersed in a 4.25% solution of AQUAPEL (PPG, USA) for 10 s, followed by rapid drying with nitrogen gas. The treated capillaries were subsequently baked in a dry oven at 80°C for 2 hours. After drying, the capillaries were stored in a cushioned storage box until further use.

Virus detection in clinical samples

All clinical procedures were carried out in accordance with the guidelines of the ethics committee of the First Affiliated Hospital of Harbin Medical University (approval no. 2024IIT453). All researchers and clinical staff wore personal protective equipment during the experiments. Clinical blood samples were collected on site from adults undergoing infectious disease screening at the First Affiliated Hospital of Harbin Medical University. Before testing, viral lysis and nucleic acid extraction were performed using the FastPure Viral DNA/RNA Kit (Vazyme, China). The extracted nucleic acid samples and PCR buffer were then added to the capillary for amplification and real-time fluorescence detection. For comparison with RT-PCR results, nucleic acid extraction was performed using Pre-NAT II system (PerkinElmer, USA), and amplification and quantitative fluorescence detection were carried out using the ABI7500 system (Thermo Fisher Scientific, USA). The viral load in the samples was determined using an HBV nucleic acid quantification kit (Sansure, China) (50–58).

Nasopharyngeal swab samples were collected from patients undergoing respiratory virus testing at the First Affiliated Hospital of Harbin Medical University and stored in inactivated virus sampling tubes. Before testing, viral lysis and nucleic acid extraction were performed using the FastPure Viral DNA/RNA Kit (Vazyme, China). Nucleic acid samples were mixed with PCR buffer and loaded into capillaries for amplification and real-time fluorescence detection. For comparison with RT-PCR results, nucleic acid extraction was performed using the SSNP-9600A system (BioPerfectus, China), and amplification and quantitative fluorescence detection were conducted using the SLAN-96P system (Hongshitech, China).

Statistical analysis

Data were analyzed and plotted using Origin software and reported as means ± SD. One-way analysis of variance (ANOVA) was used to analyze the data with IBM SPSS Statistics software, followed by Tukey’s post hoc test for multiple group comparisons. The P values were used to indicate the level of statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001).

Supplementary Materials

The PDF file includes:

Text S1 to S4

Figs. S1 to S25

Tables S1 to S9

Legends for movies S1 to S5

Legend for data S1

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S5

Data S1