As the amplification method for our diagnostic system, we selected Recombinase Polymerase Amplification (RPA) because it operates at low, constant temperatures close to body temperature, eliminating the need for thermal cycling and making it highly suitable for use in portable or at-home settings (1).
We evaluated five different RPA configurations:
To simplify development and troubleshoot effectively, we decided to focus on a single pathogen to establish a fully functional workflow before extending the method to others. Using the HPV16 L1 gene as a model sequence, we tested three primer pairs designed by our team (see the Experiments and Protocols Page) at two DNA concentrations: 1 ng/µL and 0.1 pg/µL.
Symmetrical and asymmetrical RPA yielded reliable amplification even at the lowest DNA input, as confirmed by agarose gel electrophoresis and Nanodrop analysis. For asymmetrical RPA, we screened different primer ratios (1:10, 1:20, and 1:50) based on literature reports (2). Due to technical challenges, we could not conclusively determine the optimal ratio, but the setup provides a clear framework for further optimization. With these two experiments completed, the first steps toward both detection pathways were successfully achieved.
To explore truly equipment-free conditions, we conducted a temperature and incubation time screening. We found that RPA can be reliably performed at room temperature if the reaction time is extended from 20 to 30 minutes. According to Nanodrop measurements, this setup achieved an amplification of approximately one million-fold compared to the input concentration.
| Temperature [°C] | Incubation time [min] | Order of gel | DNA amount [ng/μL] |
|---|---|---|---|
| 39 | 20 | 1 (positive control) | 106 |
| 39 | 20 | 2 (negative control) | 33.4 |
| RT / ~23.8°C | 20 | 3 | 12.9 |
| 25 | 20 | 4 | 38.7 |
| 32 | 20 | 5 | 34.1 |
| 37 | 20 | 6 | 83.5 |
| RT / ~23.8°C | 30 | 7 | 87.3 |
| 25 | 30 | 8 | 75.2 |
| 32 | 30 | 9 | 103.5 |
| 37 | 30 | 10 | 121.4 |
| RT / ~23.8°C | 60 | 11 | 108.3 |
| 25 | 60 | 12 | 96.1 |
| 32 | 60 | 13 | N/A |
| 37 | 60 | 14 | 51.1 |
We also tested RPA directly on paper strips following the 2017 iGEM Munich team (“CascAID”) protocol. While asymmetrical RPA did not yield detectable amplification, symmetrical RPA on strips produced a 3-fold increase, which is technically amplification, but with low efficiency compared to expectations. This highlights the need for further optimization of reaction conditions on paper substrates.
Legend: Set-up for RPA on the strip.
Although multiple input dilution series were initiated to characterize sensitivity more precisely, we were unable to complete them due to time and equipment limitations (see the Engineering and Cycles Page).
Legend left:
Legend left:
From these experiments, several key insights emerged: RPA successfully amplified the HPV16 L1 gene from as little as 0.1 pg/µL input, demonstrating high sensitivity. The reaction proved robust across a wide temperature range, including room temperature, as long as the incubation time was adjusted. Amplification efficiency reached approximately one million-fold compared to the input concentration, confirming strong DNA production even under low-resource conditions. On-paper RPA was shown to be feasible, though further optimization is needed to improve yield.
Looking ahead, we plan to:
Together, these findings help guide the refinement of our amplification system and support the goal of building a reliable, user-friendly diagnostic test.
A central aim of our project was to establish and evaluate DNA hybridization as a detection method. In contrast to CRISPR-Cas systems, which depend on proteins, guide RNAs, and carefully controlled reaction conditions, DNA hybridization offers a simpler, more cost-effective, and robust alternative, particularly well suited for paper-based diagnostic tools intended for home use (3)(4).
To begin, we tested single-probe hybridization using three different probe lengths: a 15-mer LNA probe, a 20-mer, and a 25-mer, each labeled with the fluorophore Dy681. As templates, we used DNA from three amplification sources: symmetrical RPA (dsDNA, 1 ng/μL input), asymmetrical RPA (ssDNA-enriched, with a 1:50 forward:reverse primer ratio), and PCR products (~200 ng/μL). Each reaction (10 μL total volume) included a probe at a 1:1000 dilution from a 100 μM stock, along with NaCl, nuclease-free water, and the respective DNA template.
After denaturation at 95 °C followed by annealing at 25 °C, we observed visible signals across all template types. Among them, symmetrical RPA produced the strongest hybridization bands, and the 15-mer LNA probe consistently yielded the most robust performance compared to the longer probe designs.
Legend:
Following the initial single-probe tests, we explored room-temperature hybridization without a denaturation step. To confirm specificity, we tested our probe against non-target pathogens including Chlamydia trachomatis, Neisseria gonorrhoeae, Treponema pallidum, and Trichomonas vaginalis, and observed no cross-reactivity. This demonstrated the probe's high specificity.
Initially, products from isothermal asymmetrical RPA did not generate any detectable signal. However, after redesigning the workflow, particularly with improved probe handling and preparation, fluorescence signals were successfully observed at room temperature. This confirmed that effective hybridization can occur without prior denaturation, an important step toward simplifying the workflow for at-home testing.
Next, we optimized the primer ratios for asymmetrical RPA to improve ssDNA enrichment. We compared 1:10 and 1:20 forward:reverse primer ratios and found that 1:10 provided reliable amplification across both standard (1 ng/μL) and diluted (0.1 pg/μL) DNA template inputs. In contrast, the 1:20 ratio failed to produce a signal at low template concentrations. Based on this, we standardized the 1:10 ratio for all downstream hybridization experiments.
To further improve detection flexibility and explore the potential for multiplexing, we introduced a second probe labeled with a distinct fluorophore, Cy7, allowing us to test double-probe hybridization. Initial tests only produced signal in one channel, which we traced back to photobleaching of the longer-wavelength dye. By switching to freshly prepared aliquots and protecting them from light exposure, we achieved successful double-probe hybridization.
This was a key milestone: we observed simultaneous binding of both probes on symmetrical RPA (dsDNA) and asymmetrical RPA products (ssDNA-enriched using a 1:10 forward:reverse primer ratio), at both 1 ng/μL and diluted (0.1 pg/μL) template inputs. These results confirmed that double-probe detection is feasible under our reaction conditions and further supports the robustness of our hybridization-based detection strategy.
Legend:
Legend:
Finally, we transferred our detection system onto a lateral flow strip using the Milenia HybriDetect kit. For this setup, we used two labeled oligonucleotides: a FAM-labeled forward primer and a biotin-labeled reverse primer. On the strip, anti-FAM antibodies conjugated with gold nanoparticles are embedded, which bind the amplicons and generate the visible signal. We tested two probe concentrations, 100 µM and 0.1 pM, in combination with two concentrations of symmetrical RPA products: 1 ng/µL (undiluted) and 0.1 pg/µL (a 10,000-fold dilution).
Clear and reliable signals appeared within 2-5 minutes when using the 100 µM labeled probes, both for the undiluted and the diluted RPA products. In contrast, no signal was observed at 0.1 pM probe concentration, likely due to insufficient input for detection on the strip. These results confirmed that our system can produce fast and visible results on lateral flow strips when operated under appropriate input conditions.
Legend:
Some of the early double-probe hybridization experiments presented challenges, primarily due to fluorophore bleaching and low amplification efficiency from isothermal RPA. These issues initially prevented the detection of clear signals. However, through repeated testing, we discovered that protecting fluorophores from light exposure, changing the order of readout steps, and increasing the template volume from 1 µL to 2-3 µL significantly improved both signal strength and consistency.
When transferring the system onto lateral flow strips, we found that probe concentration was a critical factor. Using probes at 0.1 pM failed to produce any visible signal, likely because this concentration falls below the detection limit of the strip format. In contrast, higher probe concentrations, such as 100 µM, resulted in clear, reliable signals across both undiluted and highly diluted RPA products. This highlighted the importance of adapting not only the reaction chemistry but also the input parameters to suit the strip-based context.
These findings have shaped our next steps in improving the hybridization workflow. To simplify the readout and make the system more suitable for at-home use, we plan to replace fluorescent detection with alternatives such as gold nanoparticles or enzymatic colorimetric systems, which would allow for easy visual interpretation without special equipment. In parallel, we plan to refine probe designs, testing shorter sequences and introducing additional LNA modifications, to enhance hybridization efficiency and specificity. Optimizing buffer conditions is another important goal, particularly to ensure that hybridization on paper strips is both rapid and reliable under minimal-infrastructure settings.
Overall, while specificity and successful double-probe binding have already been clearly demonstrated in gel-based assays, transferring the method to paper-based strips remains a work in progress. The current strip results are promising but not yet fully consistent, and we consider them preliminary. In line with our commitment to scientific transparency, all such data are clearly marked as ongoing and subject to further optimization.
Although CRISPR-Cas12a was initially considered as the main detection method due to its high specificity, the project later shifted toward DNA hybridization as a simpler and potentially more robust alternative for at-home use. Despite this change in focus, we continued to evaluate both methods in parallel to compare their performance under the same conditions.
For the Cas12a-based detection, we used self-purified LbCas12a protein together with a target-specific crRNA and a fluorophore-quencher ssDNA reporter (ROX/BHQ2). In early tests, the reporter concentration was too low to produce a visible signal. After increasing the reporter levels, a clear fluorescence change became visible to the naked eye, especially at 32 °C and 37 °C within 30 minutes, and also at room temperature after longer incubation. These results confirmed that LbCas12a functioned reliably in our setup and provided a basis for direct comparison with the DNA hybridization method.
Tube 2: Incubation at 37 °C
Tube 3: Incubation at 32 °C
Tube 4: Incubation at 25 °C
Tube 5: Incubation at RT (23°C)
Legend:
To evaluate the compatibility of our LbCas12a-based assay with paper-based formats, we transferred the reaction onto a lateral flow strip using the HybriDetect kit. We tested two reporter concentrations (50 nM and 100 µM) alongside two RPA product inputs (1 ng/µL and 0.1 pg/µL) to assess performance under varying conditions. In all four combinations, the control line appeared, confirming that the tests ran correctly. A visible test line developed within 2-5 minutes, indicating successful target detection. As expected, the strongest signals were observed with the higher RPA input (1 ng/µL), suggesting that increased template concentration enhances detection clarity. Overall, these results confirmed that LbCas12a-based detection works not only in tube-based reactions but also reliably on lateral flow strips.
Legend:
In the early stages of our Cas12a-based detection experiments, some assays failed to produce a visible signal due to insufficient reporter concentrations. Through systematic troubleshooting, we learned that several factors, particularly the concentration of the reporter, the amount of template DNA, and the incubation temperature, are critical for achieving consistent and reliable detection.
To further improve performance, we plan to pre-assemble the Cas12a-crRNA complexes prior to each reaction and explore alternative reporter designs, including variations in length and fluorophore/quencher chemistry (e.g., FAM-Biotin or FITC-Biotin). If CRISPR-based detection proves too complex for home use in the long term, one alternative strategy could be to position it as a clinic-based solution, while continuing to develop DNA hybridization as the preferred option for at-home testing.
Despite initial setbacks, CRISPR-Cas detection demonstrated reliable performance in tube-based assays, consistently producing a clear and visible color change. The system also worked effectively on lateral flow strips, delivering rapid results under various conditions. Stronger signals were observed with higher input DNA concentrations, but detection was still successful with diluted templates, highlighting the potential of Cas12a as a flexible and sensitive detection method.
Successfully developed a predictive kinetic model to simulate the entire RPA-CRISPR/Cas diagnostic system.
To understand and optimize our diagnostic's performance, we first developed a mechanistic model based on ordinary differential equations (ODEs). This model simulates the two core processes of our system: the amplification of target DNA via Recombinase Polymerase Amplification (RPA) and its subsequent detection by the CRISPR-Cas12a system. By describing the dynamic interactions between every component, the model allowed us to perform virtual experiments, test hypotheses, and analyze system behavior under a wide range of conditions.
Experimentally validated the model's core RPA amplification module against wet lab data.
A model's predictions are only as valuable as their connection to reality. The most critical step in our work was to validate our model against experimental data. Due to the proprietary composition of commercial RPA kits, we performed a sensitivity analysis to predict a plausible range of DNA yields. Our model predicted a final DNA output between ~45 nM and 2364 nM. Our wet lab team's benchmark experiment yielded a quantified result of 405 nM.
Implications:
The successful alignment of our experimental data within the model's predicted range is a crucial achievement. It validates the foundational RPA module of our model, confirming that it accurately captures the core amplification dynamics. This validation provides a strong anchor of confidence for all subsequent predictions and insights generated by the integrated model.
Identified a critical influence of cis-cleavage that would cause system failure.
With the RPA module validated, we used the CRISPR/Cas12a & RPA integrated model to investigate potential interactions between the amplification and detection steps. The simulation revealed a critical bottleneck: the cis-cleavage activity of the Cas12a enzyme. The model predicted that at low concentrations of amplified DNA, typical during the early phases of the reaction, the Cas12a enzyme would bind and destroy the target DNA faster than RPA could produce it. This creates a "deadlock" effect that prevents the DNA concentration from ever reaching the threshold required for a detectable fluorescent signal.
Implications:
This in silico discovery was one of our project's most significant findings. It identified a fundamental design flaw that would have led to persistent experimental failure, saving invaluable time and resources. This highlights the power of modeling to proactively identify and diagnose problems that are difficult to observe at the bench.
Proposed a model-driven “two-chamber” hardware design to resolve the predicted system failure.
The insight from our model pointed directly to a rational solution. Since the cis-cleavage issue is most severe at the early time phase of RPA, we could resolve it by separating the two reactions in time. Our model simulations confirmed that allowing the RPA reaction to proceed for several minutes before introducing the CRISPR-Cas12a components would allow a sufficient concentration of amplicons to accumulate, rendering the effects of cis-cleavage negligible.
Implications:
This finding translates directly into a tangible engineering solution: a "two-chamber" hardware design for a future device. This design would physically separate the amplification and detection steps, ensuring the sample incubates in an RPA chamber before flowing into a second chamber containing the CRISPR reagents. This model-driven design provides a de-risked and scientifically grounded blueprint for a robust and functional diagnostic device.
Designed a theoretical framework for predicting enzyme stability.
A functional diagnostic must also be stable over time. To address this, we developed a theoretical framework to model enzyme degradation and predict diagnostic shelf-life. The model is based on principles of protein denaturation kinetics.
The results reveal how storage conditions dramatically affect shelf life. At refrigeration temperatures (e.g., 4°C), enzyme activity declines slowly over many months. Once brought to room temperature (e.g., 21°C), degradation accelerates substantially, but the assay remains viable for weeks of typical use.
Integrated all modules to create a comprehensive, end-to-end simulation.
To create a truly holistic tool, we integrated the Enzyme Stability Framework with our validated RPA-CRISPR Kinetic Model. This multi-scale simulation can now forecast not only the immediate signal generation of a fresh test but also how its performance alters after months of storage under varying temperatures. The results were critical, particularly insightful for our entrepreneurial plans.
The heatmap shows that with proper refrigerated storage (e.g., 100 days at 4°C), followed by room-temperature use, the assay reliably crosses the threshold within 30-40 minutes. However, prolonged room-temperature storage (e.g., 200 days at 21°C) leads to substantial activity loss, extending detection times or preventing threshold crossing altogether.
Implications:
This integrated analysis provides a powerful, quantitative validation for our initial entrepreneurial strategy of incorporating freeze-drying. The model clearly shows that without a method to preserve enzyme stability, the test's performance would be unreliable for real-world use where a cold chain is not guaranteed. This result is a cornerstone of our business case, proving that lyophilization is not just a desirable feature but an essential requirement for our diagnostic to be effective and accessible globally.
Developed a predictive model for asymmetrical RPA
We developed a mechanistic model of asymmetrical RPA that accurately reproduces experimental kinetics reported in five independent papers. The model captures all three biological phases observed during the reaction and maintains perfect mass balance while providing quantitative predictions for assay optimization. This work represents a key step toward our ultimate goal of integrating a DNA hybridization model to generate predictions for our second approach.
Developed an Epidemiological model estimating public health impact.
To connect our technical work to its real-world context, we developed an epidemiological model to simulate the spread of an STI in a population. The model compares scenarios with and without the availability of an accessible, rapid diagnostic like ours. The model showed that introducing our test could significantly lower disease prevalence over time by enabling earlier detection and treatment. This finding offers strong public health justification for our project.
To summarize our work, our main achievement was the development of a kinetic model that made a critical prediction. By first checking the RPA amplification part of our model against a real lab experiment, we gained the confidence we needed to trust its insights. This trust was essential when the full model predicted that our combined system would fail due to the "cis-cleavage deadlock." This finding was invaluable, as it allowed us to propose a tangible hardware solution—the two-chamber device-before wasting time and resources on experiments that were destined to fail. Furthermore, by integrating the Enzyme stability model with diagnostic system performance, we gained invaluable insights regarding the real-world lifecycle of our test and validated our plan to use lyophilization. This step made us more clear in our entrepreneurial strategies.
However, our Dry Lab's effort to tackle all the edges of our project extended far beyond that. Our Epidemiological model was an invaluable global health justification for our project and served for us as a validation of our primary goal to combat STI epidemics with fully user oriented diagnostics.
However, in the spirit of scientific honesty, it's important to be open about the significant challenges we faced. Our journey was not a straight line, and these difficulties taught us a great deal. A major hurdle was that the commercial kits we used were essentially "black boxes," with unknown concentrations of key ingredients. We learned to navigate this uncertainty by using our model to test a wide range of possibilities, a technique called sensitivity analysis. This experience showed us how computational modeling can be a powerful tool for working with the kind of incomplete information that is common in real-world science. We also learned a hard lesson in interdisciplinary communication when an experiment to get numbers for our Cas12a model didn't work out as planned. We had to adapt quickly, relying on data from previously published papers, which taught us to be flexible and resourceful. Finally, we learned about the importance of project management. Our initial ambition to build two complex models at once was greater than the time we had. Making the tough decision to pause work on the second model to focus all our efforts on the main one was a critical lesson in prioritizing depth and validation over breadth.
Throughout our project, we engaged with over 20 stakeholders from diverse fields, whose feedback helped us refine our approaches and ensure our project addresses real-world needs.
To understand public testing behavior and the demand for STI diagnostic tools, we conducted a 25-question survey, collecting 417 responses over one month. The results confirmed our design approach, provided insights into current testing rates, allowed us to predict the potential impact of our product, and helped analyze our target market for entrepreneurial opportunities.
During our interviews, we soon realized that the test itself is only half of the work done. People need proper guidance as soon as they open the package containing the test until they have received appropriate treatment. We therefore decided to build a product page, displaying the content of our future webpage, guiding people throughout the whole process. We made sure the page is easy to read, displaying only the most important information, while still providing enough knowledge on STIs, our test and what to do with a test result. A lot of this work is based on an interview with a health psychologist from the university of Zurich, Walter Bierbauer. Make sure to read his interview to see how his input was integrated into the product page.
We developed a comprehensive business case to translate our diagnostic innovation into a viable product. Our startup concept addresses an unmet need in accessible, affordable STI testing. We defined our minimum viable product (MVP), created a business model canvas, performed market and competition analyses, and established a financial and scaling strategy. We identified key partners and regulatory pathways to ensure market readiness. Our approach demonstrates both feasibility and sustainability.