Description

1. Background and Challenges

Respiratory infectious diseases have high incidence, rapid transmission, and a tendency to cause cluster outbreaks, making them a major global public health concern. Early clinical symptoms caused by different pathogens are highly similar and commonly include nonspecific symptoms such as fever, cough, sore throat, and fatigue, which makes accurate identification from symptoms alone difficult. Without timely diagnosis and intervention, some patients may progress to pneumonia, bronchitis, or even respiratory failure, while continued transmission may occur within households, schools, childcare facilities, eldercare institutions, and healthcare settings. The Global Burden of Disease 2023 analysis identifies lower respiratory infections as the world's leading infectious cause of death, with the highest burden concentrated in children younger than 5 years and adults aged 70 years and older ^[1]. Therefore, developing rapid and accurate methods for early pathogen detection that can be used in primary-level settings is important for improving diagnostic and treatment efficiency, helping stop transmission chains, and reducing the risk of severe disease.

This project focuses on five pathogens: respiratory syncytial virus (RSV), human metapneumovirus-B (hMPV-B), Streptococcus pneumoniae (SPN), Bordetella pertussis (BP), and Haemophilus influenzae (HI). Together, they include both viral and bacterial pathogens and are associated with a high disease burden, overlapping symptoms, hidden transmission, or the risk of invasive infection. They therefore represent different public health concerns that need early identification in childcare facilities, eldercare institutions, and primary healthcare settings.

Table 1 Clinicalepidemiological risks and inclusion rationale for the five targeted respiratory pathogens

These five pathogens represent several risk categories, including high-burden viral infection, clinically similar viral infection, severe bacterial infection, hidden transmission, and invasive infection. Developing a rapid combined detection system targeting their conserved sequences can help solve the diagnostic challenge in primary-level settings where similar symptoms may require different treatment and management paths, while also creating a reusable technical framework for future expansion to other respiratory pathogens.

2. Limitations of Existing Detection Technologies and POCT Products

The practical need in respiratory pathogen testing is not simply to achieve a lower limit of detection. Rather, rapid, specific, stable, and interpretable results must still be obtained under conditions involving low pathogen loads during early infection, variable sample quality, limited personnel and equipment, and overlapping symptoms caused by multiple pathogens.

2.1 Limitations of existing detection technologies

Current methods for respiratory pathogen detection can be divided into precise laboratory testing and rapid on-site testing. qPCR and RT-qPCR provide high sensitivity and specificity, but standard workflows often require nucleic-acid extraction, fluorescence amplification instruments, standardized laboratory conditions, and trained personnel, making direct use difficult in decentralized or resource-constrained settings ^[8,14].

Antigen or lateral-flow assays are simple to operate and provide rapid results, but clinical sensitivity can vary with viral load and the timing of testing; serological testing is also affected by the timing of the immune response ^[14]. Isothermal amplification reduces dependence on thermal cyclers, but assay performance still depends on careful primer and reaction design to control background and non-specific products. Recent studies of multiplex respiratory POCT also show the value of combined multi-pathogen detection, while successful on-site use also requires integrated solutions for sample processing, channel isolation, temperature control, signal readout, and user operation ^[8].

2.2 Limitations of current POCT products

There is still a gap between a technically workable method and a usable POCT product. Many existing products or prototypes cover only a single pathogen or a small number of pathogens, making them poorly suited to settings where multiple pathogens circulate at the same time and symptoms overlap. Expanding the detection panel can introduce primer competition, channel crosstalk, and more difficulty in interpreting results. Some approaches still require off-device nucleic acid extraction, open-tube transfer, or repeated pipetting, which increase training needs and operating errors as well as the risk of aerosol contamination by amplification products. Primary-level use also requires stable reagent storage, internal and external quality control, positive and negative controls, invalid-result prompts, device calibration, and data records that can be tracked. These product-development elements are often not fully considered during the proof-of-concept stage. Finally, a rapid result is not equivalent to a clinical diagnosis: the timing of sample collection, differentiation between colonization and infection, management following a false-negative result, and pathways for confirming positive results must all be clearly explained in the product instructions and operating procedures.

An ideal POCT for respiratory pathogens should therefore meet five related requirements: low dependence on specialized settings and trained operators, a short turnaround time, stable detection at low pathogen loads, highly specific discrimination among multiple targets, and full-process quality control from sample processing to result recording. The technical and device designs of this project are based on these requirements.

3. Integrated RT-RPA-CRISPR/Cas12a Solution

To address the technical challenges and product limitations described above, we first focused on recombinase polymerase amplification (RPA), an effective isothermal amplification method characterized by a moderate reaction temperature, a relatively short amplification time, and low equipment dependence, making it suitable for primary-level screening and on-site testing. However, the amplification efficiency and specificity of an RPA system depend strongly on primer sequences, the local structure of the target sequence, and reaction conditions. Poor primer design or complex target-sequence secondary structures may reduce amplification efficiency and lead to background amplification and the accumulation of non-specific products, making it difficult for RPA alone to achieve both high sensitivity and high specificity.

The CRISPR/Cas12a system provides a second sequence-recognition step to address these limitations. Guided by a crRNA, LbCas12a specifically recognizes the target double-stranded DNA and then activates trans-cleavage activity, cleaving a single-stranded DNA reporter to generate a fluorescent or visual signal. Placing this step after RPA amplification can reduce the risk of false positive results caused by nonspecific amplification products. Recent studies have shown that RPA coupled with CRISPR/Cas12 can support multi-target respiratory-pathogen detection in portable systems, while extraction-free one-pot RPA-Cas12a systems can simplify bacterial pathogen testing workflows ^[9,10].

On this basis, our project is designed for kindergartens and eldercare institutions. It integrates RT-RPA with LbCas12a trans-cleavage for multi-pathogen detection and includes the parallel development of a portable fluorescence-reading prototype. Instead of simply combining two technologies, the system uses RPA for rapid target increase and Cas12a for sequence-specific confirmation, while independent reaction channels and digital result recording make multi-target results easier to interpret.

3.1 Target pretreatment and RT-RPA amplification

Specific primers are designed against conserved gene fragments of the five pathogens. Viral RNA is reverse-transcribed and enters the RPA workflow together with bacterial DNA for isothermal amplification at about 37-43 °C, which reduces the need for precision thermal-cycling equipment.

3.2 Specific recognition by the LbCas12a RNP complex

A target-specific crRNA forms a ribonucleoprotein complex with LbCas12a. When an amplification product matches the matching crRNA, Cas12a is activated and cleaves a fluorophore-quencher ssDNA reporter, turning sequence recognition into a measurable signal.

3.3 Multi-target interpretation in independent channels

Amplification and Cas12a recognition for the five pathogens are performed in separate reaction channels to reduce primer competition and signal crosstalk. The portable reading module collects fluorescence from each channel and reports positive, negative, or invalid results based on predefined thresholds.

3.4 Digital result management

A mobile interface is used to display results, store testing times and sample identifiers, and support authorized data export, helping institutions keep records that can be tracked.

4.A Multidimensional Optimization Loop Integrating Dry-Lab and Wet-Lab Work

The development of RPA-CRISPR/Cas12a assays often relies on wet-lab experience and repeated trial and error. At the CRISPR layer, Cas12a activation kinetics are strongly affected by the target sequence context: bases in the PAM region, sequences within and outside the region complementary to the crRNA spacer, DNA topology, and target length can greatly affect activation and trans-cleavage performance ^[15]. At the integrated-assay level, combining nucleic-acid amplification with CRISPR cleavage in a one-pot format creates limits on reaction compatibility that must be addressed through system design and condition optimization ^[16]. Together with primer-placement and amplification constraints, these factors make testing all possible combinations slow and costly, while conditions optimized for one target may not work well for another. To address this challenge, the project establishes a two-stage optimization strategy: computational models prioritize candidate sequences and reaction parameters, while wet-lab experiments validate and refine these predictions.

4.1 DNABERT-6-Assisted Sequence Prioritization

A DNABERT-6 nucleic-acid language model is used to encode candidate sequences and capture contextual nucleotide patterns, including PAM-associated target context, primer regions, and longer-range sequence dependencies.. Sequence embeddings are combined with useful thermodynamic and design features, including GC content, melting temperature, minimum free energy, conservation, and predicted off-target risk. Conserved target regions retrieved from NCBI are used to rank suitable primer–crRNA combinations instead of optimizing each component separately. The resulting shortlist is then used for mechanism-based analysis and experimental testing.

4.2Mechanistic ODE modeling of the RT-RPA-Cas12a cascade

The reaction network is represented as a set of ordinary differential equations under the assumption that the reaction mixture is uniform. The model describes key molecular processes involved in the RT-RPA-Cas12a cascade, including primer-related interactions, recombinase-mediated strand invasion, polymerase-driven strand extension,Cas12a-crRNA complex formation, target recognition, and reporter cleavage. Component concentrations are initialized from the experimental recipe, rate constants taken from the literature are used as starting values, and the system is solved over 0-600 s with an adaptive solver such as LSODA. The simulated changes over time of amplified target, Cas12a-crRNA-target complex, and cleaved reporter link molecular events to signal rise time and endpoint fluorescence.

4.3 Wet-lab validation and DBTL iteration

Candidate crRNAs-primer pairs, and reaction formulations are tested by real-time fluorescence monitoring. Signal growth rate, endpoint intensity, analytical detection limit, and negative-control background are used to evaluate both sequence ranking and parameter predictions. Measurements are fed back into the sequence and kinetic models to update feature weights, rate assumptions, and parameter bounds, forming a Design-Build-Test-Learn cycle. This feedback structure is intended to reduce unnecessary trial and error while while keeping wet-lab evidence as the final basis for decisions.

5. Portable Detection Terminal for Institutional Settings

We developed ByeGerm, a portable fluorescence-reading platform, to bring the project’s respiratory detection assays into one instrument workflow. The platform provides the physical conditions needed to incubate reaction tubes, capture fluorescence and display the resulting records on a mobile device. Its intended settings include kindergartens and eldercare institutions, where staff need clearly identified reaction positions and results that can be traced to the corresponding test. The hardware is developed alongside the five-target assay panel described above.

Our development progressed through two generations. The first combined temperature control, a fixed optical observation area and embedded image processing. The second added a separate lysis zone and a rotating reaction chamber, with corresponding changes to control and acquisition software. These changes extend the tasks performed within the instrument and accommodate more reaction positions.

5.1 First-generation integrated platform

The first-generation instrument integrates thermal, optical, computing, communication and power modules in an enclosure approximately 7.5 × 14 × 11 cm in size. A heating assembly supports the reaction tubes, while the camera, excitation LEDs and filter are held in fixed positions. An Orange Pi 5 Max board based on the RK3588 processor handles image processing and communicates with the mobile interface. A rechargeable lithium battery supplies the instrument, and the enclosure and supports are produced by three-dimensional printing.

The temperature-control assembly combines ceramic heating elements, temperature feedback and fan-assisted heat dissipation. PID control and PWM output regulate heating power. The first-generation platform record reports a control range of 25–95 °C and temperature fluctuations of approximately ±0.7 °C. For the respiratory assays, the operating temperature is selected from the corresponding wet-lab protocol; the device’s control range describes its hardware capability.

Blue LED illumination excites the fluorescence signal, and a filtered camera records the reaction area. The enclosed housing limits ambient-light entry and keeps the relative positions of the tubes and optical components fixed. We compared images at five camera-to-tube distances and selected 50 mm for the first-generation mounting position. This turns the optical arrangement into a repeatable part of the assembled instrument.

After acquisition, YOLOv8n locates the reaction tubes and OpenCV processes the selected image regions. The mobile application provides device connection, program configuration, temperature display, image viewing and record storage. These functions establish the acquisition and display workflow that the second generation builds upon.

5.2 Second-generation thermal and mechanical integration

The second generation is approximately 15 × 30 × 20 cm and brings sample pretreatment and the main reaction chamber into separate regions of one enclosure. Its architecture combines device control, image processing, temperature control, optical acquisition, power management and mechanical positioning. Dedicated boards organize power distribution, sensing and actuator connections, while the RK3588 platform processes the camera images.

A separate lysis zone provides two heated positions at the front of the device. Its metal heating block is physically separated from the main reaction area, with supporting and insulating structures around the thermal zones. A front lid gives access to the lysis zone, while the upper lid opens the main reaction chamber. This layout makes the two loading areas distinguishable and gives each stage its own thermal assembly.

We developed describes a rotating reaction assembly with three observation faces and five tube positions per face. A fixed camera acquires images as the faces are brought into view. Mechanical positioning and feedback associate each acquired image with the relevant observation face. The architecture therefore expands the number of reaction positions without requiring a separate camera for each tube.

Reaction positions and diagnostic targets are different quantities. For the five-target respiratory panel, individual tubes are assigned to the corresponding assays, controls and any required replicates. Multiple targets can be accommodated in one instrument workflow, while optical acquisition proceeds by observation face. The two-position lysis zone and the larger reaction chamber also have different capacities, so pretreatment and readout must be considered together when evaluating whole-device throughput.

5.3 Image analysis and mobile operation

The second-generation software retains the image-analysis and mobile-communication functions established in the first generation and adds controls for the rotating assembly. The application supports device connection, run settings, temperature display, fluorescence images and curves, and record storage. The operator can switch observation faces through the relevant controls, allowing images from several tube positions to enter the same analysis workflow.

The readout proceeds from image acquisition to tube localization, signal extraction and association with the run record. Tube-detection confidence indicates how confidently the image model has located a tube; the fluorescence measurement is a separate output. Positive, negative and invalid classifications for the respiratory panel must follow the rules and controls validated for the corresponding assay. The interface and saved records should preserve the relationship between the sample, target and reaction position.

During operation, the user connects the mobile terminal, selects the assay program and confirms the reaction-position assignment before starting acquisition. Temperature and fluorescence records allow the run to be followed over time. Images and processed records can then be reviewed together. Battery power, wireless communication and accessible loading lids support the intended use outside a conventional benchtop instrument setup.

5.4 Connecting the instrument to assay evaluation

The hardware provides the incubation and fluorescence-reading environment for the assays developed through the dry-lab and wet-lab workflow. Assay optimization and analytical evaluation are presented in Measurement; hardware evaluation addresses the quality of the conditions under which those signals are acquired. The first-generation records include temperature-control characterization, camera-distance comparison and integrated operation. Second-generation records add acquisition from multiple tube positions, image-region identification and tests of the mobile application on several phone systems.

For the respiratory panel, whole-device evaluation links these engineering functions to the target-specific reactions. The main questions are whether temperatures remain consistent across occupied positions, whether illumination and repositioning alter the readout, and whether the saved records preserve the correct tube-to-target assignment. These measurements connect the biological assay’s performance to the assembled device and guide the next changes to the instrument.

Table 2 Hardware functions supporting the respiratory assay workflow

Together, these modules connect the project’s independent respiratory assays to a portable instrument. The first generation establishes integrated heating, imaging and mobile readout; the second adds pretreatment heating and the mechanical organization needed for more reaction positions. This provides a practical platform for evaluating the complete detection workflow in the institutional settings that motivated the project.

6. Policy Support and Human Practices

6.1 Policy context

The intended application of this project is in line with the broader direction of Chinese public-health policy, which focuses on prevention, stronger primary-level capacity, and improved surveillance and early warning. The 14th Five-Year Plan for National Health called for enhanced monitoring, intelligent early warning, rapid-response capacity, stronger infectious-disease testing capacity in primary healthcare institutions, and improved services for priority populations ^[11]. The Healthy Children Action Enhancement Plan (2021-2025) provided policy context for strengthening child-health services and disease prevention during that planning period ^[12]. The 2024 Guiding Opinions on Establishing and Improving an Intelligent Multi-Point Trigger Surveillance and Early-Warning System for Infectious Diseases further call for acute respiratory infection syndromic surveillance, multi-pathogen testing from a single sample, monitoring in key institutions, and stronger rapid screening and testing capacity at the primary level ^[13]. These policies do not mean approval or purchasing commitments for any specific product; rather, they show the policy focus on earlier detection, multi-pathogen surveillance, and stronger primary-level public-health capacity.

6.2 Human Practices

Within the Human Practices framework, policy documents indicate only the broader direction and cannot replace direct input from real-world users. To bridge this gap, we used a closed-loop feedback strategy. Based on the intended settings and chain of responsibility, we engaged directly with managers of childcare and eldercare institutions, primary healthcare and disease-control professionals, laboratory and regulatory professionals, and caregivers of children and older users. Their specific field feedback was used to guide important design changes and evaluation measures for our device.

6.2.1 Institutional managers and frontline care staff

These stakeholders are particularly concerned with whether the workflow contains enough few steps, whether training costs are manageable, and whether records can be retained rapidly when a cluster anomaly occurs. Initially, our prototype assumed basic operator familiarity with sampling. However, field discussions revealed that any complex operation in a non-laboratory setting created training barriers that were too high and contamination risks. As a result, the device design was changed to reduce open-tube transfer and repeated pipetting, and the software interface was updated to provide clear positive, negative, and invalid states, ensuring traceable testing records without making the process too difficult to understand.

6.2.2 Primary healthcare and disease-control professionals

These professionals require interpretable results that can be used together with confirmation, referral, and outbreak-reporting procedures. Based on their input, the system was refined to include strict process controls and invalid-result prompts. Furthermore, we collaborated with them to clearly define a pathway for confirming positive results, while avoiding presenting app-based risk notifications as final clinical diagnoses.

6.2.3 Laboratory, regulatory, and biosafety professionals

These professionals are particularly concerned with analytical performance, lot-to-lot consistency, cross-contamination, sample compatibility, and quality control. In response to their practical concerns about waste management in grassroots facilities lacking standardized biohazard disposal, we are developing built-in physical inactivation protocols for the closed reactions and waste-disposal procedures. Subsequent validation will carefully cover the limit of detection, inclusivity and exclusivity, interference resistance, cross-reactivity, repeatability, stability, and agreement with an RT-qPCR reference method.

6.2.4 Caregivers, children, and older users

These stakeholders are concerned with sampling comfort, result privacy, affordability, and anxiety caused by abnormal results. To address data privacy, we created a data management framework where the mobile application collects only the minimum necessary data, using local anonymization on the device before any authorized export. Additionally, we included clear communication and plain-language explanations of results into the product design, while strictly avoiding diagnostic claims to non-professional users in the absence of clinical validation.

7. Project Value and Future Outlook

Using five respiratory pathogens that are important to public health as development targets, this project integrates rapid RT-RPA amplification, sequence-specific LbCas12a recognition, DNABERT-6 screening, mechanistic ODE modeling, differential-evolution parameter search, and wet-lab DBTL iteration into one continuous development workflow.The core value of the project is therefore both a multi-pathogen POCT architecture and a data-driven method that can be used for other targets for RPA-CRISPR system design.

Our portable detection platform supports fluorescencesignal acquisition, ondevice result interpretation and traceable reporting. In this study, fullset performance evaluation of the assay system has been completed, and the well-developed detection scheme has been implemented on our selfdeveloped portable hardware with preliminary wholedevice adaptation accomplished. On this basis, further efforts will be devoted to indepth instrument calibration and scenario adaptation, with continuous refinement targeting thermal stability, optical consistency and practical application performance. Integration with primary-level surveillance workflows will be gradually improved to promote technical implementation and fieldscale deployment.