PDL

Goal

The overall goal of this cycle was to develop and validate a platform for CspZ protein detection by generating DNA products from single stranded DNA (ssDNA) aptamer constructs based on a novel application of the proximity-dependent ligation (PDL) process.

Iteration 1 – Initial PDL Experimentation

Design

To develop our diagnostic, our idea this year was to utilize protein-binding aptamers that would be brought in close proximity by a bridging sequence then ligated into a double stranded DNA (dsDNA) product (see Proximity-Dependent Ligation). We developed our ssDNA constructs using anti-CspZ aptamer sequences specific to Borrelia burgdorferi, the causative agent of Lyme. (see Aptamer Design). The goal of our first iteration was to test the feasibility of the PDL reaction in vitro with isolated protein samples. Building off of insights from previous literature, we also modified the protocol to include an extension step with T4 DNA polymerase following ligation to ensure better downstream compatibility in our overall assay (see Proximity-Dependent Ligation; see Experimental Modifications).

As a result, we decided to apply the redesigned PDL protocol: Aptamer protein binding DNA-DNA hybridization Ligation of aptamers using T4 DNA ligase Duplexing of ssDNA product via T4 DNA polymerase

Negative controls included samples with no aptamers and no ligase (see Validating PDL). Diagnostic efficacy of PDL was analyzed using 1.5% gel electrophoresis with TBE (Tris-Borate-EDTA) buffer, and a successful result was expected to show bands at 232 base pairs, corresponding to the full length of the PDL product (see Fig. 1).

Figure 1. Sequence map of the full 232 bp PDL product assembled in Benchling.

When developing our ssDNA aptamers constructs, we took advice from Dr. Mark Styczynski, an expert on metabolic dynamics from the Georgia Institute of Technology, to screen our constructs with NUPACK. By cross-referencing the minimum free energy (MFE) structure of the constructs with the MFE structure of the aptamers, we minimized the likelihood of secondary structure formation in our linker regions to ensure ease of bridge binding by making linkers (BBa_25SC2X09 and BBa_2537R5AB) as linear as possible (see Figs. 2-4).

Previous slideNext slide

Slide 0 of 0

Figure 2. NUPACK predictions for the MFE secondary structure of aptamer 11 (BBa-25xxui8b; 2a) compared to aptamer 11 with linker region (2b), showing linear structure of left linker (​​BBa-25sc2x09). Figure 3. NUPACK predictions for the MFE secondary structure of aptamer 4 (BBa_2543XHTV; 3a) compared to aptamer 4 with linker region (3b), showing linear structure of right linker (BBa_2537R5AB).

Figure 4. NUPACK predictions for the MFE secondary structure of combined PDL aptamers and linkers, showing near complete linearity of overall linker region.

Build

We purchased these aptamer and bridge ssDNA constructs from Integrated DNA Technologies (IDT), hydrated them in nuclease free water, and diluted to our desired stock concentrations (see PDL Lab Notebook). Additionally, we ordered T4 DNA ligase (M0202S) and T4 DNA polymerase (M0203S) from New England Biolabs and commercial CspZ protein from Rockland Immunochemicals, Inc. We carried out the PDL assay according to our initial protocol as below (see Table 1).

Test

The PDL assay showed no bands at the expected length for experimental and negative control samples, indicating a lack of successful product formation(see Fig. 5).

Figure 5. Gel electrophoresis results of initial PDL reaction showing no bands present for experimental and control samples.

Learn

From our initial experimentation, we concluded that PDL did not yield a sufficient number of products without amplification for visualization on gel electrophoresis. Results showed that post-PDL amplification is likely required for visualization and data was not sufficient to prove successful assay activity.

Iteration 2 – Validating PDL with PCR

Design

After realizing that PDL on its own did not generate a sufficient number of copies for visualization on gel electrophoresis, we decided to incorporate Polymerase Chain Reaction (PCR) to amplify the dsDNA PDL product. By doing so, we aimed to isolate the PDL step and confirm that our aptamers successfully bound to CspZ to produce a ligated DNA output. To begin our experimentation with post-PDL PCR, we manually designed PCR primer sequences by using Addgene’s PCR primer design recommendations . The PCR primer design restrictions set by Addgene are: Length: 18-24 nucleotides GC content: 40-60% Start or end with 1-2 G/C pairs Tm (Melting temperature): 50-60ºC Primer pairs should have a Tm within 5ºC of each other Primer pairs should not be complementary to one another

The sequences and properties for PCR primer pair 1.1 are as follows (see Figs. 1-2): PDL-PCR forward primer 1.1 (BBa_25HV7AXA): GGTCTGGTTGGCCCGTGTGTCATT 24 bp, 58% GC, Tm 60ºC

Previous slideNext slide

Slide 0 of 0

Figure 1. NUPACK predictions for the MFE secondary structure of PDL-PCR forward primer 1.1, showing predominant linearity with relatively low probability for hairpins. Figure 2. NUPACK predictions for the MFE secondary structure of PDL-PCR reverse primer 1.1, showing predominant linearity. ([BBa_25GA8G6C](https://registry.igem.org/parts/bba-25GA8G6C))

Our primer pair was screened for secondary structures and dimer formation using predictions from NUPACK, which shows that primer sequences are mostly linear (see Figs. 1-2).

Build

We purchased the designed primer sequences from IDT and used them with the Phusion High-Fidelity PCR Master Mix with GC Buffer (F532S) to perform PCR. By following the protocol below, our goal was to amplify the output of the PDL reaction to generate a detectable amount of dsDNA product (see Table 1).

Figure 3. Determination of annealing temperature for PDL-PCR primer set 1.1 using NEB Tm calculator (New England Biolabs, n.d.).

Test

Initial experimentation with PCR amplification of PDL once again showed no bands at the expected length for experimental and negative control samples, indicating a lack of successful product formation or amplification (see Fig. 4).

Figure 4. Gel electrophoresis results of initial post-PDL PCR reaction showing no bands present for experimental and control samples.

Learn

These results suggested that while the PDL reaction may have occurred, the yield of ligated product was either too low for successful amplification or the concentration of amplified DNA from PCR was not high enough for visualization on gel electrophoresis. This indicated the need to optimize reaction conditions, such as temperature, primer design, or initial concentration or reagents, to improve amplification sensitivity.

Iteration 3 – Troubleshooting Temperature Condition

Design

After unsuccessful amplification in the previous iteration, the first assay condition we decided to troubleshoot was the incubation temperature for the aptamer-protein binding step. We hypothesized that the lack of sufficient time for the anti-CspZ aptamers to properly bind to the target protein was limiting the production of dsDNA products. Therefore, we decided to use an incubation time of 1 hour at room temperature instead of the previous 15 minutes at 37°C.

Build

By using the updated protocol as shown below, the goal of this iteration was to investigate if a change in temperature conditions improved aptamer binding efficacy and enhanced ligation efficiency (see Table 1).

Test

Gel electrophoresis results continued to show no visible amplification for both temperature conditions, indicating that changes in temperature alone were insufficient or had an insignificant effect on increasing PDL yield (see Fig. 1).

Previous slideNext slide

Slide 0 of 0

Learn

Because there was no significant increase in detectable PDL product, we concluded that troubleshooting the temperature condition did not improve assay performance. To conserve time needed for experimentation, we decided to perform future iterations with the 37°C for 15 minutes condition.

Iteration 4 – Redesigning PCR Primers

Design

The next condition that we decided to troubleshoot was the functionality of our PCR primers. We hypothesized that our manually designed primer set 1.1 may have been faulty, which led us to use IDT’s built in PrimerQuest™ software tool to design additional PCR primer sets 1.2 and 1.3 (Integrated DNA Technologies, n.d.). Our goal in this iteration was to redesign primers that exhibit greater amplification efficiency within the PCR process. The sequences and properties for PCR primer pairs 1.2-1.3 are as follows (see Figs. 1-4): PDL-PCR forward primer 1.2 (BBa_25SMX70F): CCAGCTTATTCAATTGGTTGGTCT 24 bp, 42% GC, Tm 54ºC

Figure 1. NUPACK predictions for the MFE secondary structure of PDL-PCR forward primer 1.2, showing predominant linearity with relatively low probability for hairpins.

PDL-PCR reverse primer 1.2 (BBa_258FI332): TATCTACAAACCCAACCCATCC 22 bp, 45% GC, Tm 53ºC

Figure 2. NUPACK predictions for the MFE secondary structure of PDL-PCR reverse primer 1.2, showing predominant linearity.

PDL-PCR forward primer 1.3 (BBa_25BEQMUZ): GTTGGCCCGTGTGTCATTAC 20 bp, 55% GC, Tm 54ºC

Figure 3. NUPACK predictions for the MFE secondary structure of PDL-PCR forward primer 1.3, showing predominant linearity.

PDL-PCR reverse primer 1.3 (BBa_25DECTTF): TTGCACTTACTATCTACAAACCCAA 25 bp, 36% GC, Tm 53ºC

Figure 4. NUPACK predictions for the MFE secondary structure of PDL-PCR reverse primer 1.3, showing predominant linearity.

Both of our new primer pairs were screened for secondary structures and dimer formation using predictions from NUPACK, which shows that primer sequences are mostly linear (see Figs. 1-4).

Build

We purchased the designed primer sequences from IDT and performed post-PDL PCR according to the protocol in previous iterations, with annealing temperature of 62°C and 63°C for primer sets 1.2 & 1.3, respectively (see Figs. 5-6; see Iteration 2 – Validating PDL with PCR.

Figure 5. Determination of annealing temperature for PDL-PCR primer set 1.2 using NEB Tm calculator (New England Biolabs, n.d.).

Figure 6. Determination of annealing temperature for PDL-PCR primer set 1.3 using NEB Tm calculator (New England Biolabs, n.d.).

Test

PCR amplification of PDL was repeated with both new primer pairs while keeping other reaction conditions constant. However, no visible bands were observed on gel electrophoresis, suggesting that either PDL was unable to properly produce the dsDNA output or that PCR continued to fail in detecting the ligated product (see Fig. 7).

Previous slideNext slide

Slide 0 of 0

Figure 7. Gel electrophoresis results of PDL-PCR reaction showing no bands present for primer sets 1.2 (a) & 1.3 (b).

Figure 7. Gel electrophoresis results of PDL-PCR reaction showing no bands present for primer sets 1.2 (a) & 1.3 (b).

Learn

The results in this iteration demonstrated that the limited success we observed was likely not due to faulty primer design alone, but other factors that made the ligation process inefficient. However, because we were unsure of which primer set was able to successfully amplify the PDL product, the next iterations continued to also test all three different primer pairs.

Iteration 5 – Adjusting Protein Concentration

Design

After repeatedly failing to properly generate and amplify the dsDNA PDL product, we hypothesized that the concentration of CspZ protein was not optimal for aptamer binding and ligation. Initially, we used a high protein concentration of 10 ng/µL, which is equivalent to 8.88 × 10¹⁰ molecules/µL for the 67.8 kDa commercial protein (see Adjusting Protein Concentration). However, after finding that this value exceeded the 2.00 × 10⁶ CspZ molecules/µL predicted by our computational model, we speculated that the excess of protein may have saturated the reaction and interfered with aptamer reactivity (see Fig. 1; see CspZ Protein Concentration). To address this issue and ensure our assay better reflected physiological conditions, we repeated the PDL reaction using a protein concentration that is consistent with in vivo levels reported in literature.

Figure 1. Estimated CspZ concentration in the bloodstream up to 300 days post-infection using Mass Action reaction kinetic laws.

Build

By using the updated protocol as shown below, the goal of this iteration was to investigate if a change in protein concentration improved assay efficacy and amplification efficiency (see Table 1).

Test

Gel electrophoresis of the post-PDL PCR products continued to show no visible bands at 232 bp, suggesting that lowering protein concentration alone did not significantly improve amplification (see Fig. 2).

Figure 2. Gel electrophoresis results of post-PDL PCR reaction at more physiologically relevant protein concentration, lacking bands for experimental samples.

Learn

These findings suggested that the lack of successful amplification was not simply from the high protein concentration, but may be linked to other confounding variables. To ensure that our assay mimicked physiological conditions as best as possible, we continued future testing at the lowered protein concentration.

Iteration 6 – Upscaling PCR Reaction

Design

We hypothesized that amplifying only a portion of the 50 µL PDL reaction may have limited dsDNA yield, as the low product concentration may not be homogenous in the solution. To increase the likelihood of detecting the PDL product, we upscaled the PCR reaction to use the full 50 µL of the PDL reaction while adjusting all reagents proportionally.

Build

By using the updated protocol as shown below, our goal was to determine if increasing input volume of PDL product would generate detectable dsDNA for downstream applications (see Table 1).

Test

Despite adjusting the reaction volume, gel electrophoresis revealed that amplification of PDL still failed, once again suggesting that additional factors were likely limiting product formation (see Fig. 1).

Figure 1. Gel electrophoresis results of post-PDL PCR reaction at upscaled PCR volume, lacking bands for experimental samples.

Learn

These findings suggested that the unsuccessful results we observed were not simply from a lack of available product. However, because we were unsure if upscaling the volumes for PCR did not impact the reaction, we continued future testing at the increased reaction conditions (see Table 1).

Iteration 7 – Optimizing Aptamer Concentration

Design

After observing consistently unsuccessful results with our PDL experimentation, we decided to develop a deterministic ODE (ordinary differential equation) model for PDL. From our ODE model, we identified our aptamer concentration as another potential limiting factor. At the low concentration of 20 pM being used up to this point, our PDL ODE model predicted that only ~0.6 dsDNA copies of PDL product would be produced per reaction (see Fig. 1; see PDL Modeling). This value is insufficient for reliable amplification, informing us that the aptamer concentration characterized in literature is not optimal for our assay.

Figure 1. ODE model of generated dsDNA PDL product over time at low aptamer concentration created by correlating the initial concentration of reagents to PDL product over time.

To address this, we decided to increase working aptamer concentrations to 20 nM, which our model predicted would generate a much greater copy number of ~800 (see Fig. 2). We also included an additional negative control lacking CspZ protein (no protein) to further verify the specificity of our assay.

Figure 2. ODE model of generated dsDNA PDL product over time at high aptamer concentration created by correlating the initial concentration of reagents to PDL product over time.

Build

By using the updated protocol as shown below, our goal in this iteration was to evaluate if increasing aptamer concentration could produce a detectable amount of dsDNA with PDL-PCR (see Table 1).

Test

In this iteration, gel electrophoresis results of PCR products revealed a clear band at ~200 bp in the experimental sample with primer pair 1.2, while negative controls showed no bands (see Fig. 3).

Figure 3. Gel electrophoresis results of post-PDL PCR reaction with increased aptamer concentration, showing band for experimental sample at the expected length of ~200 bp for primer set 1.2.

Learn

The length of fluorescent DNA bands produced was consistent with the expected 216 base pairs of the amplified PDL product with primer set 1.2, confirming a successful PDL assay reaction. Increasing the aptamer concentration produced a visible PCR product, confirming that previous failures were due to insufficient dsDNA copies from low aptamer input. These model-informed wetlab results validated our computational predictions and established the basis for a functional PDL reaction for downstream amplification. Moving forward, the final protocols followed for our PDL assay and PDL-PCR reaction are taken from our most optimized experimental workflows (see Table 1; see Iteration 6 –Upscaling PCR Reaction.

Iteration 8 – Testing Protein Concentration Over Time

Design

Following the optimized protocol for PDL finalized in iteration 7, we also sought to evaluate how expected protein concentrations over multiple time points of simulated infection affected PDL efficacy. The goal was to determine the range of time for which our assay could reliably generate detectable dsDNA products and be compatible with downstream amplification.

Results were qualitatively visualized using 1.5% gel electrophoresis where a successful result was expected to show bands at 216 base pairs, corresponding to the length of the PDL amplicon (see Iteration 7 –Optimizing Aptamer Concentration. The success of PDL was also quantified using the ImageJ software, using the process below :

Build

We prepared PDL reactions using expected CspZ concentrations corresponding to 2, 25, 50, 75, 100, and 150 days post-infection (pi) as shown below, while keeping other reagent concentrations consistent (see Table 2; see CspZ Protein Concentration).

All final reactions, including controls, were repeated in triplicates to ensure that results were reproducible and statistically significant.

Test

Gel electrophoresis of post-PDL PCR products revealed fluorescent bands at the expected ~200 bp for protein concentrations corresponding to 2, 25, and 50 days pi, qualitatively demonstrating successful dsDNA generation (see Fig. 1).

ImageJ analysis of standardized integrated density of fluorescent DNA bands was also performed to quantify assay performance across tested time points (see Fig. 2).

Figure 2. ImageJ analysis of post-PDL PCR on agarose gel electrophoresis with protein concentrations from 2-150 days pi, quantitatively validating PDL viability until at least 100 days pi.

Learn

These results confirm that our PDL assay is able to produce sufficient dsDNA product for amplification and remain functional for up to 100 days pi. Band intensity was greatest at earlier time points and gradually decreased over time, resulting in a loss of assay activity by 150 days pi. Negative controls exhibited significantly lower signal intensities, validating the specificity of our assay and minimizing the risk of false positives. Overall, LANCET demonstrates reliable and specific detection of protein concentrations in blood across multiple stages of Lyme disease, supporting its use in downstream diagnostic workflows. With these iterations of development, our findings demonstrate that LANCET is able to reliably detect protein concentrations in the blood, even in low volumes, supporting PDL’s potential for conjugation with downstream processes within our proposed diagnostic pipeline.

RPA

Goal

Following successful detection of CspZ by proximity-dependent ligation (PDL), the next cycle of the diagnostic process sought to reliably amplify double stranded DNA (dsDNA) products by using Recombinase Polymerase Amplification (RPA), enabling high sensitivity.

Iteration 1 - RPA Positive Control Experimentation

Design

To enhance the sensitivity of our diagnostic, we utilized RPA, an isothermal amplification method capable of rapidly generating dsDNA copies that is well-suited for point-of-care diagnostics; (see Recombinase Polymerase Amplification). Because the dsDNA products initially formed by PDL are produced at a very low concentration, RPA amplifies the sequence to create a sufficient number for visualization on gel electrophoresis or for downstream quantification in our overall assay. To validate the feasibility of using RPA within our diagnostic, we first tested the commercial TwistAmp® Exo and Basic kits using a positive control provided by TwistDx. Therefore, the goal of this first iteration was to ensure that the RPA kits could successfully amplify a known target before applying RPA to experimental constructs. The functionality of the kits was assessed with 1.5% agarose gel electrophoresis with TBE (Tris-Borate-EDTA) buffer and fluorescent output on a plate reader, and controls included negative reactions with no template DNA to reduce the possibility of non-specific amplification.

Build

We purchased the TwistAmp® Basic and Exo kits and the positive control oligo mix and template DNA from TwistDx. Positive control reactions were performed according to the instructions by TwistDx as shown below, and we ran triplicates to confirm reproducibility (see Table 1).

Test

Gel electrophoresis results of RPA positive control experimentation revealed a clear band for positive control samples at the expected ~150 base pairs, while negative controls showed no bands.

Figure 1. Gel electrophoresis results of RPA positive control reaction, showing band for positive control samples at the expected length of ~150 bp.

Additionally, the positive control showed a significant increase in fluorescence under the plate reader, confirming that the kit functioned as expected (see Fig. 2).

Figure 2. Fluorescent output of TwistAmp® Exo kit quantified using plate reader; results indicate a significant difference in RFU between positive and negative controls.

Learn

The RPA positive control produced clear bands at the expected 150 bp and exhibited a significant fluorescence increase, confirming that our reagents were functioning as intended. These results validated the amplification efficacy of the commercial RPA system and established a foundation for applying RPA on dsDNA templates generated from the PDL reaction.

Iteration 2 – RPA Target Construct Experimentation

Design

Once we confirmed that our commercial RPA kits functioned as expected, we tested RPA on a synthetic target DNA construct that simulated the PDL product (see Fig. 1). To do this, we purchased a dsDNA RPA target construct (BBa_25SSHRPA) which contained binding sites for all current and future primer sets. This allowed us to isolate the RPA process and first evaluate whether we could reliably amplify the target construct without confounding variability from PDL and actual protein samples.

Figure 1. Sequence map of the RPA target construct assembled in Benchling.

For our experimentation with LANCET-specific RPA, we began with manually designed RPA primer sequences developed using the TwistDx® Design Manual. The guidelines for RPA primer design are:

• Length: 30-36 nucleotides
• GC content: 20-70%
• Tm (Melting temperature): 50-100 ºC
• Maximum length of a mononucleotide repeat: 5 nucleotides

The sequences and properties for RPA primer pair 1.1 are as follows (see Figs. 2-3):

• PDL-RPA forward primer 1.1 (BBa_250TF2EO): GGTTGGTCTGGTTGGCCCGTGTGTCATTAC 30 bp, 57% GC, Tm 66ºC

Figure 2. NUPACK predictions for the minimum free energy (MFE) secondary structure of PDL-RPA forward primer 1.1, showing predominant linearity with low probability for hairpins.

• PDL-RPA reverse primer 1.1 (BBa_25X1JFV4): ACAAACCCAACCCATCCCATCCCGCCACCA
• 30 bp, 60% GC, Tm 67ºC

Figure 3. NUPACK predictions for the MFE secondary structure of PDL-RPA reverse primer 1.1, showing predominant linearity.

By using the NUPACK software, we screened the primer pair for secondary structures and dimer formation. We found that the MFE structures for the primers were mostly linear with few interfering secondary structures, which is critical for RPA, as the process is isothermal and does not allow for DNA denaturation (see Figs. 2-3). We also intentionally positioned our primers on both aptamer constructs to facilitate amplification only in the presence of fully ligated products, rather than single aptamers (see Fig. 4).

Figure 4. Sequence map of the RPA amplicon resulting from primer set 1.1 assembled in Benchling.

To ensure maximum efficacy of our RPA assay, we designed our target construct to contain all RPA amplicons. Amplicons were developed with key parameters from TwistDx® DNA Amplification Kit Assay Design Manual in mind:

• Length: 100-200 nucleotides
• GC content: 40-60%
• Minimize number of repetitive, palindromic sequences

The properties for RPA amplicon 1.1 are as follows (see Fig. 5):

• 196 bp, 47% GC

Figure 5. Repetitive sequences were minimized in the amplicon, specifically the linker regions, to improve amplification speed by reducing the product-to-noise ratio.

Build

We purchased the designed primer and target DNA sequences from IDT to perform RPA. By following the protocol below, our goal was to amplify the output of the PDL reaction to generate a detectable amount of dsDNA product (see Table 1).

Test

Figure 6. Gel electrophoresis results of RPA on target construct with primer set 1.1, showing no bands present for experimental and control samples.

Learn

No visible bands were observed for either the experimental or control samples, indicating that RPA primer set 1.1 failed to produce a detectable level amplification. This outcome suggested that the primer design was likely suboptimal for use in our assay. Based on these findings, we concluded that further troubleshooting was necessary to improve RPA performance and ensure reliable amplification of PDL products.

Iteration 3 – Redesigning RPA Primers

Design

Following unsuccessful amplification using primer set 1.1, our objective for this iteration was to redesign RPA primers to improve amplification efficiency. To achieve this, we used the EZassay software to design multiple new primer candidates. From the software’s set of possible primer combinations, we filtered the results to meet specific requirements of our assay which EZassay’s algorithm did not account for. We excluded amplicons that were too short for our assay (<100 base pairs) or those that did not encompass both aptamer constructs, which could lead to a false positive result (see Iteration 2 – RPA Target Construct Experimentation). Our goal in this iteration was to redesign primers that exhibit greater amplification efficiency within the RPA process. The sequences and properties for RPA primer pairs 1.2-1.4 are as follows (see Figs. 1-5):

• PDL-RPA forward primer 1.2 & 1.3 (BBa_25VUQNNR): CGTGTGTCATTACGGGTTGGATAAGATAGTA
• 31 bp, 42% GC, Tm 60ºC

Figure 1. NUPACK predictions for the MFE secondary structure of PDL-RPA forward primer 1.2, showing predominant linearity with relatively low probability for hairpins.

• PDL-RPA reverse primer 1.2 (BBa_25RMXLR4): ACCACCACCAAGTGAATTGAATAAGCTGGTA
• 31 bp, 42% GC, Tm 60ºC

Figure 2. NUPACK predictions for the MFE secondary structure of PDL-RPA reverse primer 1.2, showing predominant linearity with relatively low probability for hairpins.

• PDL-RPA reverse primer 1.3 (BBa_25NVR9HK): ATAAGCTGGTATAAACACACAACCAAACAAC
• 31 bp, 35% GC, Tm 58ºC

Figure 3. NUPACK predictions for the MFE secondary structure of PDL-RPA reverse primer 1.3, showing predominant linearity with moderate probability for hairpins.

• PDL-RPA forward primer 1.4 (BBa_25PUGKNO): GGTTGGCCCGTGTGTCATTACGGGTTGGATA
• 31 bp, 42% GC, Tm 60ºC

Figure 4. NUPACK predictions for the MFE secondary structure of PDL-RPA forward primer 1.4, showing predominant linearity with relatively low probability for hairpins.

• PDL-RPA reverse primer 1.4 (BBa_257J0EPT): CCAAGTGAATTGAATAAGCTGGTATAAACAC
• 31 bp, 35% GC, Tm 58ºC

Figure 5. NUPACK predictions for the MFE secondary structure of PDL-RPA reverse primer 1.4, showing predominant linearity with moderate probability for hairpins.

All of our new primer pairs were screened for secondary structures and dimer formation using predictions from NUPACK, which shows that primer sequences are mostly linear (see Figs. 1-5). The properties for RPA amplicons 1.2-1.4 are as follows (see Fig. 6):

• 153 bp, 42% GC
• 133 bp, 42% GC
• 154 bp, 44% GC

Build

The redesigned primers were synthesized by IDT and used in RPA under the same conditions as the previous iteration. Each primer set was tested separately with the synthetic dsDNA template, and negative controls (no template) were included to validate specificity.

Test

Experimentally, RPA primer sets 1.2 and 1.4 demonstrated greater efficacy, producing gel bands within the expected base pair length around ~100-150 base pairs for each amplicon and showing significantly greater expression than the negative control (see Fig. 6).

Learn

The presence of clear bands in experimental samples and the absence of amplification in negative controls confirmed that RPA can effectively amplify dsDNA products synthesized through PDL. These results validated the compatibility and functional integration of RPA with PDL, demonstrating a key step in establishing LANCET as a cohesive diagnostic workflow capable of detecting CspZ protein and amplifying the resulting DNA signal for downstream analysis.

CRISPR-Cas12a

Goal

Designing a CRISPR-Cas12a system to turn a DNA product into measurable fluorescence.

Iteration 1 – Initial Cas12a Positive Control Experimentation

The purpose of this workflow is to optimize the Cas12a protocol and test for fluorescence to ensure the functionality of the DETECTR system with a verified target sequence and crRNA V-crRNA: Verified construct CRISPR RNA V-Target: Verified target construct L-crRNA: Lyme construct CRISPR RNA (complementary to RPA amplicon) L-target: Lyme target construct (Based on RPA amplicon)

Design

In order to visualize the results of the LANCET experimental design for our verified DNA constructs, we utilized a protocol provided by Signalchem regarding their Cas12a(LbCpf1, His) Kit. Using the protocol outline, we created an initial adaptation of the adjustable protocol to our experiment. Literature on the successful use of Cas12a to diagnose Mycoplasma Pneumoniae was referenced for selecting a verified crRNA that can successfully bind Cas12a to the verified target construct. Thus, we developed our first experimental protocol using our verified DNA constructs. Add enzyme, buffer, and crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FQ reporter, V-Target, and nuclease free water(as needed to achieve final volume) Incubate at 37°C for 15 minutes before pipetting the final solution into a 96-well plate Run the plate reader for 45 minutes at pre-heated 37°C

Fluorescence expression would result from the separation of 6-FAM and BHQ1. The result is read using a microplate reader at the optimal excitation and emission settings to maximize accuracy of detection for 6-FAM. Alongside the positive control, a negative control and “no enzyme”(blank) were run alongside the positive. This was done to ensure the lack of false positives or high fluorescence in reactions without target DNA or enzyme.

Build

Our enzyme, Cas12a (LbCpf1) was provided by New England Biolabs, alongside the following; an ssDNA FQ reporter composed of 6-FAM and BHQ1, Cas12a Reaction Buffer, and our V-crRNA. The target construct was ordered from IDT via custom oligonucleotide synthesis as dsDNA.

Test

An experimental value showed a significantly high level of background fluorescence that remained static for the read time in the plate reader. This level of fluorescence persisted throughout the duration of the read (see Fig. 1).

Figure 1. Failed Cas12a reaction with verified constructs showing inconsistent and abnormally high background fluorescence.

Learn

The lack of a single positive with significantly higher fluorescence values in comparison to the negative and “no enzyme”(blank) controls reflects a failed reaction where the fluorescence in controls gave inaccurate results. An ideal reaction would show similarly low results in negative and no enzyme controls, while the positive control would show a much greater increase in fluorescence. The no enzyme control showing high fluorescence as well as a high background fluorescence in the other two controls was identified as a potential reagent issue.

Iteration 2 - FQ Reporter Troubleshooting

Design

Initial results provided insight that an abnormal and inconsistent amount of background fluorescence was being produced across the controls in our reaction. To establish and validate the root cause of this issue, we reviewed previous experiments where it was most likely a leaky reporter was present in the reaction. In addition, a custom FQ reporter was purchased from IDT as a replacement for the “leaky” reporter from Signalchem. The leaky reporter was quantified from a failed reaction alongside a blank control of nuclease free water in order to differentiate the fluorescent values to prove that the FQ reporter from signalchem had a higher background fluorescence compared to a solution with no reporter. The IDT ordered FQ reporter was then run in the plate reader to measure background fluorescence in comparison to nuclease free water and FQ reporter.

Build

The Signalchem FQ reporter as part of the Cas12a kit, the custom IDT FQ reporter, and a blank of nuclease free water were used to design reactions that could determine the main reporter issue with background fluorescence.

Test

High levels of background fluorescence in the faulty reporter quantified in comparison with a blank (see Fig. 1).

Figure 1. The graph reveals the abnormally high fluorescence values of the fault FQ reporter.

The custom FQ reporter from IDT showed very insignificant background fluorescence that would not interfere with reaction data (see Fig. 2).

Figure 2. Graph showing little to no cleaved 6-FAM concentration produced from the FQ reporter on its own, demonstrating proper reporter probe stability.

Learn

The Signalchem reporter caused issues with background fluorescence which was seen in previous V-target reactions. The reporter was deemed at fault for leaking high amounts of background fluorescence. The IDT FQ reporter was shown to have very minimal and statistically insignificant background fluorescence, having the same level of fluorescence as nuclease free water solutions with no reporters, representing a high level of stability and low levels of background fluorescence.

Iteration 3 - Verified Construct Experimentation

Design

After extensive communication with Signalchem regarding potential issues with our reactions, we learned that their available protocol could be further optimized to suit our reagents and experimentation. With their feedback we were able to streamline the protocol and suit it better to our needs. Volumes, concentrations, and the protocol were all adjusted to maximize compatibility with our verified reaction after addressing and ensuring a lack of any other present issues. Add enzyme, buffer, and crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FQ reporter, V-Target, and nuclease free water(as needed to achieve final volume). Directly pipette the final solution into a 384-well plate. Run the plate reader for an hour at 37°C. Fluorescence expression would function as mentioned in Iteration 2

Build

The enzyme, Cas12a(LbCpf1) was bought from New England Biolabs, alongside their NE Buffer r2.1. The V-crRNA was bought from Signalchem while the V-target DNA construct was ordered from IDT via custom oligo synthesis. Sequence length of the ssDNA reporter was determined by literature on the effectiveness of FQ reporters. The FQ reporter was then ordered from IDT.

Test

Fluorescent reactions were triplicated to ensure significant results across positive control for verified constructs. A stable increase of fluorescence was found in the positive control over the span of an hour (see Fig. 1).

Figure 1. Graph of final Cas12a reaction with verified sequences showing greater cleaved 6-FAM reporter concentration in experimental samples as compared to control samples, demonstrating successful Cas12a activity.

Learn

The quantification of significantly increasing fluorescence in the positive reaction well and the lack of any significant fluorescence in our negative controls verifies that our optimized DETECTR system works in producing a fluorescent signal when the correct target DNA is present. This allows us to begin testing results with an LFA in order to create a simple and visualized result that does not require the use of a plate reader.

Iteration 4 – Experimentation with Lateral Flow Assay and Biotin Reporter

Goal

Be able to visualize the results of Cas12a on a Lateral Flow Assay to verify the functionality of the DETECTR system when integrated with another method of reaction quantification.

Design

As DETECTR serves as a means to accurately detect the results of the prior PDL and RPA, we used fluorescence as a reliable quantitative output. However, LANCET needs to be visualized in an accessible way; hence, the use of the lateral flow assay. Adapting our protocol to this visual output required us to replace the FQ reporter with an ssDNA reporter composed of 6-FAM and biotin. A successful reaction would result in the cleavage of the FB reporter. The biotin would bind to streptavidin found on the control line. The 6-FAM labeled fragments of the reporter would bind to gold nanoparticle-conjugated anti-FAM antibodies, which travel downstream of the LFA and bind to the goat anti-mouse antibody painted test line. Add enzyme, buffer, and V-crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FB reporter, V-Target, and nuclease free water(as needed to achieve final volume). Incubate at 37C for 15 minutes Place the sample pad of the Lateral Flow Assay into the tube, ensuring contact between the solution and sample pad, then wait 5-10 minutes for results of reaction to fully travel up the LFA and show the control or test line.

Build

The build is replicated from the initial positive control except for the FQ reporter which is replaced by an FB reporter. In addition, a Cas Nucleic Acid Test Strip from Signalchem was bought due to its complementarity with the FB reporter.

Test

A successful reaction visualized the separation of the FB reporter on the LFA. The test line was faintly visible, but still present. There was also a strong presence of the control line (Figure 1).

Figure 1. LFA visualization of Cas12a reaction with verified sequences showing a visible test line alongside the control line in the experimental sample (top) and only a control line for the negative sample (bottom), confirming successful cleavage of the FB reporter.

Learn

The results verified the presence of a separated FB reporter, a result of the successful cleavage of the target sequence which led to the indiscriminate cleaving of the FB reporter. The faint yet present test line for the positive control reaction on the LFA can be explained by the timing of inserting the LFA into the test tube. We came to the conclusion that the V-target positive reaction was not given enough time to run at 37C, thus a majority of FB reporters were left uncleaved. Those reporters were well visualized on the control line while the few FB reporters that managed to get collaterally cleaved by the successful cleaving of the target DNA were able visualize a positive reaction on the test line. The verified negative control proves that only the control line will show an accurate negative result of our reaction. Regardless, a positive result is given, validating our DETECTR system with verified constructs in an LFA.

Iteration 5 – Cas12a Target Construct Experimentation

Goal

Proceeding the verification of the functionality of the DETECTR system, target construct experimentation required the designing of our own crRNAs and target sequence. This experimentation validates our construct designs that are complementary to RPA.

Design

Using the verified constructs experimentation as a foundation for experimentation with our own sequences, the previous protocols were optimized to the L-target DNA and L-crRNA in both a plate reader and LFA reaction. Add enzyme, buffer, and L-crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FQ reporter, L-Target, and nuclease free water(as needed to achieve final volume). Directly pipette the final solution into a 384-well plate. Run the plate reader for an hour at 37°C. (Run for fluorescent expression over time read in a plate reader.) Add enzyme, buffer, and L-crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FB reporter, L-Target, and nuclease free water(as needed to achieve final volume). Incubate at 37C for 15 minutes Place the sample pad of the Lateral Flow Assay into the tube, ensuring contact between the solution and sample pad, then wait 5-10 minutes for results of reaction to fully travel up the LFA and show the control and/or test line. (Visualized output on LFA)

The crRNAs used for experimentation consisted of a common scaffold sequence used for Cas12a crRNAs as well as a ‘TTTV’ PAM site. The spacer region of the crRNA was then designed complementary to our L-Target construct adjacent to the upstream PAM site. Primarily, the crRNAs were designed complementary to regions of the RPA amplicon that included the primers. They were also designed using Benchling’s guide RNA software in order to minimize off-target effects (Benchling 2025) (see Fig. 1).

Figure 1. Predictions from Benchling software showing high on-target effects.

Build

Similarly to verified construct experimentation, the enzyme, Cas12a(LbCpf1) was bought from New England Biolabs, alongside their NE Buffer r2.1. The L-crRNA and the L-target DNA construct were ordered from IDT via custom oligo synthesis. Sequence length of the ssDNA reporter was determined by literature on the effectiveness of FQ reporters. The customFQ reporter was then ordered from IDT. The FB reporter is included in the experimentation in LFAs.

Test

Run in triplicates to ensure the significance of data and validity of our sequences, the experimental controls all displayed a significant increase in fluorescence over time. The negative controls displayed a low level of fluorescence consistently throughout the read. This describes a successful reaction, justifying the accuracy of our amplicon based target construct as well as the complementary crRNA (see Fig. 2).

Figure 2. Graph of Cas12a reaction using synthetic Lyme target DNA construct, demonstrating successful Cas12a activity indicated by greater cleaved 6-FAM reporter concentration in experimental samples than in control samples.

The lateral flow assay displayed a present and visible test line, indicating that the FB reporter was cleaved successfully while the negative control LFA displayed only the control line, meaning that there was no cleavage of the reporter. This reverifies the accuracy of our sequences and our protocol in both the LFA and plate reader reactions (see Fig. 3).

Figure 3. LFA visualization of Cas12a reaction with synthetic Lyme target DNA construct showing a distinctly visible test line in the experimental sample (top) compared to the control line demonstrated in the negative sample (bottom), confirming successful cleavage of the FB reporter.

Learn

These successful results indicate that the protocol in both plate reader and LFA testing have now been optimized to a high efficacy and flexibility (as seen by exchange in reporters). Our designed target and guide constructs are validated through successful results translated into different outputs in multiple successful reactions. Because the constructs were centered around the RPA amplicon, this indicates that when experimentation with an RPA amplicon resulting from PDL is used as the target, our DETECTR system will be able to quantify positive results, leading to a successful conjugation of the diagnostic components.

Iteration 6 – Conjugating Cas12a with PDL and RPA

Design

Conjugating every workflow of the diagnostic results in the accuracy of our diagnostics and validates the functionality of LANCET. Following successful conjugation of PDL and RPA, the amplified construct served as the target for our DETECTR system. Thus, with further compiled knowledge of our protocol and functionality, the protocol was further optimized to incorporate the PDL+RPA target. Add enzyme, buffer, and L-crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FQ reporter, PDL+RPA construct, and nuclease free water(as needed to achieve final volume). Directly pipette the final solution into a 384-well plate. Run the plate reader for an hour at 37°C. (Ran for fluorescent expression over time read in a plate reader.) Add enzyme, buffer, and L-crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FB reporter, PDL+RPA construct, and nuclease free water(as needed to achieve final volume). Incubate at 37C for 15 minutes Place the sample pad of the Lateral Flow Assay into the tube, ensuring contact between the solution and sample pad, then wait 5-10 minutes for results of reaction to fully travel up the LFA and show the control or test line. (Visualized output on LFA).

Build

The L-crRNA used in target construct experimentation was reused for this workflow as it is complementary and functional to the PDL+RPA construct which was used as the target. Reagents were used as per the successful previous reactions. (see Cas12a Target Construct Testing.

Test

Like the L-target reactions, the PDL+RPA construct reactions were run in triplicates to ensure statistical significance in our reaction data. The results show a much higher level of fluorescence found in the experimental control wells as opposed to the negative control (see Fig. 1).

Figure 1. Graph of Cas12a reaction on RPA-amplified PDL product, demonstrating successful Cas12a activity indicated by greater cleaved 6-FAM reporter concentration in experimental samples than in negative controls.

As previously mentioned about LFA reactions, the test line becomes present when the FB reporter is successfully cleaved. Our experimental reaction when visualized on an LFA, the test line was highly visible and present. The negative control showed no test line and only the control line (see Fig. 2).

Figure 2. LFA visualization of Cas12a reaction with RPA-amplified PDL product showing a distinctly visible test line in the experimental sample (top) compared to the control line demonstrated in the negative sample (bottom), confirming successful cleavage of the FB reporter.

Learn

The results of testing the PDL+RPA construct with both a plate reader and LFA produced positive results. This justifies the functionality and successful conjugation of all the major diagnostic components of LANCET. As reactions succeeded from PDL, and the direct results of the reactions used in RPA, then given as amplicons that were “targets” for Cas12a with the L-crRNA, every reaction in our wetlab diagnostics conjugated one after another, led to an accessible and distinctive visualization of the results of LANCET.

Iteration 7 – Experimentally Validating crRNA Sequence from CASPER

Design

Due to the lack of crRNA designing tools, software developed CASPER, a design tool that generates high-scoring crRNAs and RPA primers based on target sequences. In order to validate the generated crRNAs, we chose the highest scoring crRNA design based off of our RPA amplified target sequence (L-Target). Our experimental protocol for reactions in the plate reader was then optimized for the CASPER crRNA. Add enzyme, buffer, and CASPER crRNA to microPCR tube and incubate at room temperature for 15 minutes Following incubation, add FQ reporter, L-Target, and nuclease free water(as needed to achieve final volume). Directly pipette the final solution into a 384-well plate. Run the plate reader for an hour at 37°C. (Ran for fluorescent expression over time read in a plate reader.)

Build

The reagents used in our successful L-Target experimentation were also used for this validation (see Cas12a Target Construct Experimentation). The L-crRNA was replaced with the CASPER designed crRNA.

Test

After an hour in the plate reader, the experimental control that utilized the CASPER crRNA displayed a significantly higher level of fluorescence that increased overtime compared to the static negative controls with an ambient level of fluorescence (see Fig. 1).

Figure 1. Graph of Cas12a reaction comparing the results of CASPER crRNA enzyme activity with positive and experimental reactions, demonstrating successful Cas12a activity indicated by greater cleaved 6-FAM reporter concentration than in negative controls.

Learn

This details a working reaction where the experimental controls show a significant level of fluorescence while negative controls show a very low and unchanging level of fluorescence. Then the CASPER crRNA can be validated by its successful usage in the DETECTR system as a crRNA. The graph shows that the CASPER crRNA is on par with our L-crRNA. The design tool is then proven to be able to design high scoring and functional crRNAs.

Iteration 7 – Addressing Efficacy of DETECTR in Low Target Concentrations

Design

In LANCET’s real world application, the PDL+RPA construct may result in varying concentrations. However, because the conjugation of the diagnostic cannot use a nanodrop to determine the concentration of the target, we ran a concentration curve to verify the ability of Cas12a to cleave at different concentrations.

Build

Utilizing the procedure for verified construct experimentation, the protocol was optimized to run different concentrations of target DNA including a negative and no enzyme control. The experimental build consists of reagents used in our successful reactions.

Test

The plate reader was run for an hour with identical settings to our standardized reaction, providing the results in the graph shown below (see Fig. 1).

Figure 1. Graph of Cas12a reaction at varying target DNA concentrations (250 nM, 100 nM, 40 nM, 16 nM), showing increased Cas12a activity with higher target DNA as indicated by greater cleaved 6-FAM reporter concentration than in negative controls.

Learn

The significant level of fluorescence shown in the lowest target concentration out of the experimental controls ran indicates the ability of our DETECTR system to accurately identify the presence of smaller traces of target DNA. In addition, the other concentrations of the target construct reflected successful reactions as predicted with the highest concentration of target fluorescing the most and decreasing with the concentration. This then is able to prove that LANCET will function optimally and show a significant quantifiable positive result even at varying concentrations of target DNA, low and high.

CRISPRi

Goal

The goal of this cycle was to validate our CRISPRi system in a cell-free TXTL environment by testing dCas9–sgRNA complexes targeting a synthetic Borrelia burgdorferi target sequence, ensuring functional transcriptional repression before in vivo application.

Iteration 1 - Validating TXTL

Design

We first verified that GFP could be efficiently transcribed and translated in the TXTL reaction, establishing a quantitative baseline for repression analysis. The dCas9 plasmid was included to confirm that its expression alone would not reduce fluorescence, ensuring any repression would be caused by our sgRNAs.

Build

After we prepared our deGFP plasmid, we then ran a positive control at the proper experimental concentrations, which optimize the efficiency of the CRISPRi reaction, as detailed in our protocol below:

Test

Measuring the fluorescence of our positive and negative controls indicated that our prepared plasmid was indeed producing GFP, as there is a noteworthy difference in relative fluorescence units between the sample with and without GFP (see Fig. 1).

Figure 1. Graphs of deGFP expression from our positive control and negative control assays; data derived from 2024 project, SHIELD (Lambert iGEM 2024).

Learn

This iteration verified that all of our reagents were competent and were ready to be expressed alongside targeting sgRNAs in a CRISPRi reaction. We confirmed that any subsequent decrease in fluorescence in later tests would directly correspond to sgRNA directed repression, not background effects from the target sequence or system variability.

Iteration 2 - Single-Gene Target Construct Experimentation

Design

We obtained the complete sequence of the genes Bb0250 and Bb0841 and utilized Benchling’s software to select sgRNA binding sites that would optimize on-target effects. This analysis led us to identify ideal sgRNA sequences expected to maximize on-target effects on the non-coding strand of each gene for the CRISPRi system. We designed our target constructs by assembling linear DNA constructs tagged with deGFP following the sgRNA binding regions (see Figs. 1-2). Successful binding of our sgRNAs to our genes of interest will lead to a quantifiable decrease in fluorescence, as dCas9 inhibits downstream transcription of GFP.

Figure 1. Linear DNA Construct of the Bb0250 target construct.

Figure 2. Linear DNA Construct of the Bb0841 target construct.

Build

We purchased the sgRNA constructs and abridged & spliced constructs as linear gBlocks from IDT after designing the constructs. We performed PCR on all of the linear constructs to amplify their concentrations. We followed the volumes/concentrations for CRISPRi found in the table of reagents below (see Table 1):

Test

Measuring the fluorescence of our target constructs as a positive control indicated that our construct was producing GFP, as there was a significant difference in fluorescence between the positive and negative controls. This indicated that our target constructs were competent and could be expressed alongside targeting sgRNAs in a CRISPRi reaction (see Fig. 3).

Figure 3a. Target construct functionality of Bb0250sequence

Figure 3b. Target construct functionality of Bb0841 sequence

We ran reactions with sgRNAs with highest on-target scores, and determined which had the greatest repressive capability. The constructs were all used at their amplified concentrations. The results show sgRNA60.5 for Bb0250 and sgRNA54.3 for Bb0841 to exhibit the highest repression between candidates (see Fig. 4).

Figure 4a. Characterization of repressive capability of sgRNA60.5 for Bb0250 respectively.

Figure 4b. Characterization of repressive capability of sgRNA54.3 for Bb0841 respectively.

Learn

This iteration verified that sgRNAs for Bb0250 and Bb0841 were functional and capable of directing transcriptional repression in vitro. With successful testing individually with both genes, we moved to running multiplexed CRISPRi reactions, where the reactions for each gene were combined.

Iteration 3 - Multiplexed In Vitro Reaction

Design

Following the successful validation of individual sgRNAs for Bb0250 and Bb0841, we then tested whether simultaneous targeting of multiple sites could enhance repression efficiency. Our overall assay provides one fluorescent signal output but converges the results for multiple genes in hopes to prove that we are able to induce repression of different genes in one reaction tube.

Build

We followed the volumes/concentrations for CRISPRi derived from the protocol before, using reagents for both Bb0250 and Bb0841 simultaneously (see Table 1):

Test

With sgRNA60.5 for Bb0250, and sgRNA54.3 for Bb0841, we ran triplicate multiplexed reactions of the CRISPRi system, in addition to our positive and negative controls. Similarly to the results of our model with 74.63% repression, we achieved 76.01% repression on average between the reactions. This reveals how repressive capability was not lost in the combining of the systems, and that multiplexed CRISPRi does allow for strong repression across the genome of a bacteria (see Fig. 1).

Figure 1. Triplicate multiplex reaction shows an average of 76.01% repression overall, showing successful combined downregulation

Learn

This iteration demonstrated that multiple sgRNAs could be effectively co-expressed and simultaneously guide dCas9 to distinct genomic regions within the same target construct. The results confirmed that multiplexing can amplify transcriptional repression strength in a predictable, additive manner without compromising system stability. These findings established the flexibility of our CRISPRi platform for multi-site gene targeting and provided valuable insights for our future in vivo application, which required coordinated silencing of the rpsL gene in Escherichia coli (see In Vivo Experimentation)..

One-Pot

Goal

Create uniform layers of wax at two separate melting points to control the hydration of reagents.

Iteration 1 - Determining Oil:Wax Ratios

Design

We began designing a one-pot system with the reagents for each reaction separated by layers of solidified paraffin at a set oil-to-wax ratio. After a drop of blood is added to the tube, the dehydrated reagents of the reaction are activated. When the system is heated to a specific temperature, the corresponding wax barrier is melted, rehydrating the next reagents to continue the sequential reaction. Following standardized melting points for paraffin wax found in previous research, we first tested with oil-to-wax ratios of 10:6 and 10:7. We selected waxes with melting points directly above reaction temperatures,ensuring that the layers stayed intact during the reactions while avoiding unnecessarily high temperatures that could potentially degrade the reagents.

Build

We determined both the type of wax to use and the method for adjusting the melting points from previous literature. Following the measurements provided by the paper, we weighed out the amounts of wax and oil, melting them together to create the ratios. We initially tested two different ratios of wax: 10:6 and 10:7. These ratios had measurements of 0.625 grams liquid + 0.375 grams solid and 0.588 grams liquid + 0.412 grams solid respectively. After creating these ratios of wax, we tried pipetting them into a microcentrifuge tube, however, without any protocol for pipetting wax, the layers quickly collapsed in on each other.

Figure 1. Solidified wax in pipette tip.

Test

During testing, we found that the wax solidified in the tip before it could be pipetted, creating an obstacle in creating usable layers (see Fig. 1). We also noticed that the 10:6 ratio exhibited a consistency that was too soft for setting a solid layer of wax.

Learn

Since a 10:6 oil-to-wax ratio was unsuitable for our application, we decided to instead experiment with a 10:8 ratio. To address the solidification of the wax in the pipette tip, we looked into ways of preheating the tip to keep the wax molten.

Iteration 2 - Addressing Solidifying Wax

Design

To prevent the wax from solidifying in the tip we began placing the tips in a heat bath prior to pipetting. This allowed for more time to carefully pipette the layers. We also changed the wax preparation methods to keep the ratios between both the oil and solid constant. Instead of melting the oil and wax separately and then combining them, which caused minor variations in the ratio, we weigh and mix the components first.

Build

We prewarmed the tips by placing them in a heat bath at the melting point of the wax for >1 minutes. We tested a 10:8 system adding 0.444 grams wax and 0.556 grams oil. The weighed wax and oil were heated together at >56˚C (paraffin wax melting point) until completely melted to create a consistent ratio.

Test

The 10:7 and 10:8 wax melted at their intended temperatures and we are able to pipette them with the pre-warmed pipette. However, when attempting to pipette the layers, we struggled to create two distinct and even disks.

Learn

We finalized our design using 10:7 and 10:8 for our wax ratios, as they exhibited low melting points with a sufficient temperature difference between them to control when each layer melted. Pre-warming the tips successfully prevented the solidification of the wax in the tip, however the shape of the microcentrifuge tube continued to cause problems in forming stable layers.

Iteration 3 - Creating Microcentrifuge Tube

Design

To combat the challenges the shape of the microcentrifuge created, we decided to design a custom tube in CAD. We divided the tube into 3 sections, so the layers could be made separately before the full tube was assembled (see Fig. 1).

Figure 1. CAD model of custom split microcentrifuge tube.

Build

The model was printed on default slicer settings, using a carbon-fiber 3D printer and PLA filament.

Test

The print failed due to the small size we were working at and a lack of optimization of the 3D printer setting (see Fig. 1).

Figure 1. Initial print failing to print ridges.

Learn

We learned that the default setting on this printer would not be able to achieve the precision we looked for in this print. We looked into the main changes that would be possible such as nozzle temperature, bed temperature, and print speed to help stabilize our print.

Iteration 4 - Fixing Tube Design

Design

We manually adjusted the printer’s setting to improve the effectiveness of the print. We also began thinking about how we would make the layers once the print was successful, and decided upon polytetrafluoroethylene (PTFE) tape. PTFE tape was selected for its hydrophobicity and thermal stability, ensuring negligible water uptake and durability at elevated temperatures.

Build

The settings were changed as follows (all other settings remained as default):

• Wall loops: 3
• Detect thin wall: Yes
• Initial layer speed: 20 mm/s
• Initial layer infill: 40 mm/s
• Outer wall speed: 10 mm/s
• Inner wall speed: 30 mm/s
• Small perimeters: 45% The sections of the tube were lined with PTFE tape, as its low surface energy means that molten wax does not adhere to it.

Test

The print was successful, providing us with a friction fit system to easily assemble and disassemble the tube (see Fig. 1). The PTFE tape, while easily peeling off and leaving the layer in the tube, resulted in very uneven disks (see Fig. 2).

Figure 1. Two parts of the successful printed custom tube.

Figure 2. Uneven wax layer achieved on PTFE tape.

Learn

The manual adjustments to the printer’s setting were effective and allowed us to replicate our print. However, the PTFE still did not provide us with the uniform layers necessary for our system.

Iteration 5 - Improving Layer Uniformity

Design

In efforts to improve the wax layers, we added a layer of water over the tape before pipetting the wax. The water ensured that the wax spread evenly over the surface. We also decided to switch from PLA filament to Polycarbonate (PC) to increase the heat resistance of our tube.

Build

The same printer settings were used to print the new tube out of the PC. To create the layers, we first lined the sections of the custom tube with PTFE tape, and pipetted 20 microliters of water onto the surface of the tape. The section was gently warmed, while the wax components were being measured, combined, and melted at 56˚C. The pipette tips were warmed at both 45˚C and 50˚C to prevent the solidification. The molten wax was then pipetted over the water, dispersing to create uniform disks.

Figure 1. Even layers of wax separating colored water.

Test

The printing of the tube was successful as was the use of water to create uniform layers. The layers were thin and consistent, successfully dividing sections of colored water (see Fig. 1).

Learn

The water coupled with the PTFE tape allowed us to create uniform barriers. Our current one-pot experimentation shows strong potential for LANCET to be a user-friendly point-of-care diagnostic, requiring only a prick of blood from the user.

References

Heating Pad

Goal

Create a user-friendly heating pad to control the progression of the temperatures for each step of the one-pot reaction.

Iteration 1 - Initial Pad Creation

Design

We began developing a heating element using an N-channel power MOSFET, an external power supply, a heating pad, and a TMP102 temperature sensor breakout board. Our program logic involved heating the pad when its temperature was below the target temperature as detected by the sensor and switching the heating pad to an inactive state when its temperature was above the target temperature.

Figure 1. Initial prototyping of our heating pad.

Build

Our initial prototype for the heating element involved the TMP102 temperature sensor being adhered directly to the surface of the heating pad using electrical tape. The heating pad was powered by the external power supply while the temperature sensor obtained power from the Arduino Uno board (see Fig. 1). The Arduino Uno board controlled the current flow to the heating pad by obtaining data from the temperature sensor.

Test

We found that the TMP102 temperature sensor was more effective at determining ambient temperature rather than surface temperature. Therefore, the temperature readings for the heating pad were often much lower than the actual temperature, causing excessively high heating.

Learn

Based on our testing, we decided to purchase an alternative temperature sensor that is more effective in determining surface temperature rather than ambient temperature.

Iteration 2 - Increasing Accuracy of Temperature Sensing

Design

We purchased a waterproof DS18B20 temperature sensor probe. The DS18B20 module is more effective in determining surface temperature than the TMP102 sensor since the probe’s metal surface conducts heat directly to the temperature sensing module within the probe.

Build

For more accurate temperature readings and even heating, we wrapped both the probe and the one-pot system within the heating pad and held the system upright by taping it to the breadboard.

Test

The new temperature sensor outputted more accurate surface temperature readings. However, after reviewing our program code, we determined that when the Arduino Uno sent current through the heating pad via the external power supply, the heating pad’s temperature spiked to its maximum temperature. When the Arduino Uno stopped sending current through the heating pad, the temperature rapidly decreased.

Learn

Based on our experimentation, we determined that another insulative material or substance was needed to maintain the temperature of the system without major fluctuations.

Iteration 3 - Maintaining Constant Temperature

Design

Due to its high specific heat capacity, water can maintain its temperature for an extended period of time. Therefore, we designed the heating pad to heat two tubes of water: one containing the waterproof DS18B20 probe for temperature monitoring and the other working as a water bath holding the on-pot. This design circumvents the issue of the heating pad rapidly spiking in temperature.

Build

We used the bottom portion of two 15 mL conical tubes cut at the 9 mL line. Each of the tubes is filled with 6 mL of water, and both tubes are wrapped within the heating pad. The conductive metal portion of the DS18B20 probe is fully submerged in one tube of water while the one-pot tube is submerged in the other tube of water (see Fig. 1).

Figure 1. Final water bath setup of heating pad.

Test

The temperature readings from the DS18B20 probe proved to be extremely accurate. Since the one-pot system no longer faced major temperature fluctuations, the waxes melted at their respective target temperatures within one minute each. The heating device required only about 6 minutes to heat to 37° Celsius.

Learn

Use of a water bath provided consistent, accurate temperature monitoring. This demonstrates that the heating pad is reliable in maintaining the conditions needed for the LANCET diagnostic to run effectively.

References

Modeling

Model — PDL

Goal

The main goal of Cycle 1 was to determine whether varying aptamer concentrations and binding affinities could optimize ligation yield and assay sensitivity. By testing a range of these values, we aimed to identify the combination that produced the most efficient and reliable ligation reaction for our assay.

Iteration 1 - Initial PDL Modeling

Design

Our initial model assumed that PDL yield scaled directly with the concentration of target protein (CspZ) while holding aptamer concentration constant. This simplification allowed us to isolate the effect of protein availability on overall ligation efficiency and evaluate how changes in CspZ levels alone could influence the formation of double-stranded DNA product.

Build

By using deterministic ODEs in MATLAB, we set aptamer concentrations to 20 pM, expecting this to reflect physiological serum conditions. This baseline allowed us to evaluate whether the modeled PDL system could generate measurable product under the same conditions we used experimentally.

Test

Our model simulation output predicted less than one product molecule was formed per reaction. Consistent with these results, the experimental PDL trials also showed no detectable PCR bands, confirming that the reaction produced levels of DNA product too low for visualization which aligned closely with the model’s prediction.

Learn

Product formation was limited not by protein concentration but by insufficient aptamer binding, which restricted the overall efficiency of the ligation step in the PDL pathway. This finding indicated that even when CspZ was present, the low number of bound aptamer complexes prevented effective product generation. To better understand this limitation, we incorporated a sensitivity analysis step to quantify how variations in individual parameters affected total product yield. This addition provided a clearer picture of which variables showed the greatest influence on assay performance and helped guide adjustments for subsequent modeling cycles (see Fig. 1).

Figure 1. Bar graph showing the sensitivity of PDL product based on initial concentrations of aptamer 1, aptamer 2, and cspZ protein.

Iteration 2 - Increasing Aptamer Concentrations

Design

Guided by the sensitivity analysis, we hypothesized that increasing aptamer concentrations would enhance the likelihood of successful ligation events while avoiding excessive binding that could saturate the protein. By slightly elevating aptamer levels, we aimed to create conditions that favored more frequent binding to the CspZ protein, improving the overall efficiency of product formation without disrupting the natural balance of the reaction.

Build

We simulated aptamer titrations ranging from 20 pM to 20 nM to observe how increasing concentrations influenced overall ligation efficiency. This range allowed us to compare low and high binding conditions within realistic experimental limits and evaluate how increases in aptamer availability affected the formation of the PDL product.

Test

The model predicted an approximately $10^6$-fold increase in dsDNA product yield at the optimal aptamer concentrations, showing a dramatic improvement in ligation efficiency compared to the initial conditions. Consistent with this outcome, wet-lab validation at 20 nM aptamer concentration produced strong, clearly visible PCR bands, confirming successful product formation and demonstrating agreement between simulated and experimental results (See Fig. 1).

Figure 1. Bar graph showing the sensitivity of PDL product based on initial concentrations of aptamer 1, aptamer 2, and cspZ protein.

Learn

The model accurately reflected the nonlinear relationship between aptamer concentration and product formation, showing that increases in output eventually leveled off at higher concentrations. This behavior revealed that the main constraint in the reaction came from binding affinity between the aptamers and the target protein rather than from limitations in the enzymatic ligation process.

Iteration 3 - Integrating Decay Calculations

Design

During later stages of the ligation simulation, we observed signs of product and protein degradation that affected overall reaction stability. To address this, we introduced decay constants for both CspZ and the dsDNA product, allowing the model to account for natural breakdown over time and produce more accurate long-term behavior.

Build

The revised ODE framework incorporated new terms accounting for protein and product degradation, as well as ligase inhibition kinetics observed in later reaction stages. These updates enabled the model to more realistically represent the balance between synthesis and decay throughout the assay.

Test

When run under the updated parameters, the model’s predictions aligned more closely with experimental endpoint yields measured after 75 minutes, showing improved consistency between simulated and observed reaction behavior.

Learn

By refining the model to include degradation and secondary inhibition, our PDL Model achieved greater accuracy in capturing real reaction behavior over time. These improvements allowed the model to reflect both the final product yield and the timing of key reaction changes, making it predictive of overall yield and sensitivity.

Model 2 — RPA

Goal

Our goal was to model amplification efficiency across different template concentrations and reaction durations to understand how varying input DNA levels influence total yield. By simulating these conditions, we aimed to capture the exponential nature of RPA amplification and determine the optimal range of template concentrations for reliable and consistent DNA replication within the system.

Iteration 1 - Initial RPA Modeling

Design

Our first model treated RPA as a continuous exponential amplification process, following principles similar to those used in PCR modeling (see Fig. 1). This approach assumed constant reaction efficiency and target amplification.

Figure 1. Initial ODE Graph of RPA amplification over time

Build

Using a simple doubling function embedded within an ODE framework, we modeled DNA amplification as a continuous exponential process to capture how product concentration increases over time; this structure allowed us to simulate rapid growth in DNA yield under ideal conditions.

Test

Experimental RPA reactions plateaued much earlier than predicted, while the model continued to exhibit unchecked exponential growth. As a result, the simulation overestimated DNA yield by more than 200%, highlighting a major gap between idealized modeling assumptions and the reaction’s actual self-limiting behavior.

Learn

We discovered that polymerase availability and recombinase binding limits introduced nonlinear phases in the amplification process. As the reaction progressed, enzyme saturation and reduced primer accessibility slowed amplification, explaining the deviation from the model’s predicted continuous exponential growth.

Iteration 2 - Adding more Accurate Biochemical Limits

Design

We modified the system of ordinary differential equations to more accurately represent the biochemical limits of the RPA reaction. Specifically, the updated model incorporated enzyme saturation and resource exhaustion to capture the self-limiting behavior observed experimentally. This adjustment allowed us to reflect how amplification efficiency naturally slows over time as essential reactants—such as polymerase and nucleotide pools—are consumed, rather than continuing indefinitely at a constant rate.

Build

To implement this refinement, we introduced Michaelis–Menten kinetics to govern polymerase velocity, defining a maximum reaction rate ($V_m$) and substrate affinity constant ($K_m$) that control enzyme behavior at different DNA template concentrations. Additionally, we added a depletion term for single-stranded binding proteins (SSBs), since these are progressively used up as new strands form and stabilize. Together, these additions allowed the model to simulate the gradual decline in amplification efficiency caused by enzymatic and substrate limitations.

Test

The updated model produced sigmoidal amplification curves that closely followed experimental profiles, showing clear lag, growth, and plateau phases. However, it continued to predict slightly lower yields at equilibrium than wet-lab measurements.

Learn

Discussion with Dr. Styczynski revealed that the polymerase velocity constant ($V_{m,poly}$) had been underestimated. Increasing this constant by a factor of 2 corrected the amplitude mismatch and aligned the simulated and experimental curves (see Fig. 1).

Figure 1. Final ODE Graph of RPA amplification over time

Iteration 3 - Validation of RPA Model

Design

To validate the accuracy of our refined model, we ran simulations using different initial template concentrations and compared their resulting amplification curves. This step tested whether the model could correctly scale amplification dynamics with changing input levels, ensuring that predicted reaction behavior remained consistent across realistic experimental conditions.

Build

We simulated a series of 10-fold serial dilutions of the PDL product to observe how varying template input affected overall amplification behavior. This setup enabled direct comparison of model performance across a wide concentration range, testing whether predicted yields scaled proportionally with decreasing initial template amounts.

Test

The model accurately reproduced a twofold amplification trend for reactions lasting up to 25 minutes, aligning with early experimental kinetics. However, beyond 40 minutes, the model began to diverge from observed data, as it did not account for recombinase dissociation, which limits continued strand invasion and reduces amplification efficiency over extended durations.

Learn

After adding dissociation and reformation terms for recombinase–primer complexes, the model achieved an exact fit with experimentally observed amplification kinetics. Statistical validation using a two-way ANOVA confirmed that both amplification method and reaction time had significant effects on signal intensity, verifying the model’s accuracy in representing RPA behavior under experimental conditions.

Model 3 — Cas12a

Goal

Our goal was to predict fluorescence output as a function of target double-stranded DNA (dsDNA) concentration and Cas12a enzyme activation kinetics. By modeling the relationship between enzyme activation and collateral cleavage, we aimed to quantitatively link target abundance to fluorescent signal intensity over time.

Iteration 1 - Initial Cas12a Model

Design

We began by modeling Cas12a activation as an instantaneous event that occurred immediately upon binding to its target double-stranded DNA (dsDNA). This simplified assumption treated enzyme activation as a direct one-step process, eliminating any delay between target recognition and trans-cleavage activity. The purpose of this design was to establish a foundational kinetic model that captured the essential reactions of the Cas12a system before introducing additional regulatory or rate-limiting steps.

Build

Using this framework, we implemented three core reactions representing enzyme binding, target cleavage, and fluorescence release. The first reaction simulated Cas12a’s binding to the target dsDNA through its guide RNA; the second described target cleavage and enzyme activation; and the third modeled the subsequent collateral cleavage of reporter probes, resulting in fluorophore–quencher separation. Together, these reactions allowed us to simulate the complete activation and signal generation process in a simplified, linear form.

Test

When simulated under these conditions, fluorescence increased linearly with time, reflecting a constant and unrestricted cleavage rate. However, experimental data collected from plate-reader assays displayed a distinct early plateau, indicating that fluorescence accumulation slowed significantly after initial activation. This difference revealed that the real Cas12a reaction did not maintain a constant rate of cleavage as the simplified model predicted.

Learn

We concluded that modeling Cas12a activation as instantaneous failed to capture key aspects of the enzyme’s turnover behavior. In practice, Cas12a activation is followed by a finite period of catalytic activity before enzyme efficiency declines. By over-simplifying this step, our model ignored enzyme reuse limitations and the gradual slowing of reaction kinetics over time. To correct this, we introduced catalytic saturation kinetics in the next iteration, enabling the model to represent a more realistic transition from rapid activation to plateaued fluorescence output.

Iteration 2 - Introducing Kinetic Modeling

Design

We rebuilt the activation reactions using Michaelis–Menten parameters derived from previous literature to more accurately represent Cas12a’s enzyme kinetics. This approach replaced the earlier assumption of instantaneous activation with a rate-dependent model that considers substrate binding, catalytic turnover, and enzyme saturation. By doing so, we aimed to capture the gradual buildup of activated Cas12a molecules that occurs as target DNA is progressively cleaved.

Build

To implement this refinement, we introduced activated Cas12a as an explicit intermediate species within the ODE system. This allowed the model to differentiate between inactive complexes that have just bound the target and active enzymes capable of trans-cleaving reporter probes. We then simulated the interaction between activated Cas12a and the fluorophore–quencher reporters to observe real-time fluorescence output under dynamic activation conditions (See Fig. 1).

Figure 1. Representation of the Cas12a reaction pathway using SimBiology

Test

The updated simulation produced fluorescence curves that followed a sigmoidal trajectory, characterized by an early lag phase, mid-reaction acceleration, and a plateau as substrate availability decreased. This pattern closely mirrored the experimental fluorescence profiles recorded in plate-reader assays, validating that enzyme activation and turnover were now properly represented in the model.

Learn

Incorporating an activation delay successfully captured the true kinetic behavior of Cas12a, bridging the gap between theoretical and experimental trends. This refinement demonstrated that accurate modeling of enzyme activation dynamics is essential for predicting the timing and amplitude of fluorescence signals in CRISPR-based detection systems.

Iteration 3 - Refinemint through Feedback Inhibition

Design

We observed inconsistencies in fluorescence saturation between replicate experiments, suggesting that additional factors beyond enzyme kinetics were influencing the reaction’s long-term behavior. To address this, we refined the model to include effects that could cause signal loss or variability at extended time points.

Build

We introduced a feedback inhibition term that simulated the depletion of quencher molecules and the gradual photobleaching of fluorophores during prolonged illumination. This addition allowed the model to capture decreases in apparent fluorescence intensity that occur independently of enzyme activity, reflecting the photochemical limitations of fluorescence-based detection.

Test

With these refinements, the simulated fluorescence curves displayed a distinct flattening after approximately 30 minutes, aligning almost perfectly with experimental data across replicates. The new model reproduced both the rate of fluorescence increase and the eventual plateau, resolving the earlier discrepancies between simulations and observed results (See Table 1).

Figure 1. Experimental (left) vs. Modeled (right) Cas12a fluorophore concentration over time

Learn

This iteration revealed that accounting for photochemical effects such as quencher depletion and fluorophore photobleaching is critical for achieving long-term kinetic accuracy. Incorporating these parameters made the model more representative of real experimental conditions, improving its reliability for predicting fluorescence behavior in extended Cas12a reactions.

Model 4 — CRISPRi

Goal

Our goal was to model how varying sgRNA concentrations and dCas9 binding dynamics influence GFP repression in both single and multiplexed CRISPRi systems. By establishing this relationship, we aimed to quantify how effectively dCas9–sgRNA complexes block transcription of target genes (Bb0250 and Bb0841) over time and to predict the combined effects of simultaneous repression events.

Iteration 1 - Initial CRISPRi Model

Design

In the first iteration, we constructed an initial ODE network capturing transcription, translation, and repression for a single target, Bb0250. This version focused on representing the sequential flow from sgRNA transcription to dCas9 binding and GFP output, forming the foundation of the CRISPRi simulation.

Build

During implementation, we mistakenly represented repression using a positive Hill function, which modeled activation rather than inhibition. As a result, the mathematical framework treated dCas9–sgRNA complex binding as a promoter-enhancing event instead of a transcriptional block.

Test

When simulated in MATLAB, GFP expression continuously increased without reaching a steady state, producing unrealistic fluorescence trajectories that contradicted the expected repression behavior observed experimentally.

Learn

We resolved this issue by replacing the positive Hill term with a negative Hill function to properly represent inhibition. This adjustment corrected the feedback mechanism and successfully replicated GFP repression dynamics, establishing a realistic baseline for future CRISPRi model iterations.

Iteration 2 - Adding Second Target Gene

Design

In the second iteration, we expanded the ODE framework to include a second gene target, Bb0841, enabling the simulation of multiplexed CRISPRi repression. This design allowed us to examine how two separate guide–target interactions behave within the same system and whether simultaneous repression could be sustained without interference.

Build

We implemented independent repression pathways for each gene (See Fig. 1), assigning distinct sgRNAs and dCas9–sgRNA complexes to their respective targets. Both repression mechanisms were modeled in parallel within the same reaction network, allowing the system to account for individual transcription and translation processes while sharing global resources such as polymerase and dCas9.

Representation of the CRISPRi reaction pathway using SimBiology

Test

When compared to wet-lab multiplexed reactions, the simulated results predicted stronger repression than was experimentally observed. While the model suggested near-complete inhibition for both targets, fluorescence data showed that GFP expression remained higher than expected, particularly during later time points.

Learn

The discrepancy revealed that our assumption of unlimited dCas9 availability was incorrect. In reality, both sgRNAs compete for the same pool of dCas9 proteins, limiting complex formation and reducing repression efficiency. Recognizing this competition was a key insight that guided the next iteration of model refinement.

Iteration 3 - Editing Enzyme Pool and Binding Constants

Design

In this iteration, we refined the multiplexed model by introducing shared dCas9 resource competition terms to represent the finite enzyme pool available for sgRNA complex formation. We also adjusted the forward and reverse binding constants to better reflect binding affinity differences between sgRNAs targeting Bb0250 and Bb0841.

Build

Using experimental fluorescence decay data, we re-estimated the kinetic constants (k_f and k_r) for both target systems. These updated values were integrated into the ODE framework to improve alignment with the measured repression kinetics. The new equations accounted for reduced complex formation rates under limited enzyme conditions, reflecting realistic competition within the cell-free system.

Test

The revised model successfully reproduced the plateauing behavior seen in both experimental fluorescence curves, accurately capturing the slower repression onset and steady-state GFP levels for each target. Predicted trajectories now matched experimental outcomes across single and multiplexed configurations (See Table 1).

*Table 1. Experimental vs. Modeled Multiplexed CRISPRi GFP Concentration graphs*

Learn

This iteration demonstrated that multiplexed repression operates under nonlinear resource-sharing dynamics rather than simple additive inhibition. Accurately modeling competition for shared dCas9 resources was critical for explaining differential repression strengths and achieving quantitative agreement with experimental results.

Model 5 — HP

Goal

Our goal was to forecast future Lyme disease prevalence using an integrated framework that combines spatial, demographic, and climatic data. By linking historical incidence records with environmental and socioeconomic variables, we aimed to predict geographic risk distribution and temporal disease trends at the county and regional levels. This model provides a data-driven foundation for anticipating outbreaks and guiding proactive public health interventions.

Iteration 1 - Initial HP Model

Design

In the initial iteration, we developed a simple linear regression model that used per-capita Lyme disease case counts as the dependent variable and key environmental predictors—such as forest cover, precipitation, and temperature—as independent variables. This approach aimed to establish a baseline relationship between ecological conditions and disease incidence.

Build

The linear regression was trained on county-level data spanning 2010–2023, standardizing all variables using z-scores to enable cross-feature comparison. While the model produced interpretable coefficients, it performed poorly overall, with an R² value of 0.41, indicating limited explanatory power. In particular, it failed to capture nonlinear relationships between forest cover and case density—variables known to interact in complex ecological ways.

Learn

This iteration revealed that simple linear assumptions cannot adequately represent the nonlinear and multifactorial nature of Lyme disease transmission. The weak fit emphasized the need for more flexible modeling approaches capable of handling complex feature interactions, leading us to explore advanced machine learning techniques in subsequent iterations.

Iteration 2 - Transitioning to XGBoost

Design

In the second iteration, we transitioned from simple linear regression to an XGBoost regression framework to capture nonlinear patterns and feature interactions influencing Lyme disease incidence. This shift allowed the model to account for complex dependencies among environmental, demographic, and temporal variables that could not be represented by a purely linear approach.

Build

We engineered temporal lag features such as cases_lag_1 and cases_lag_2 to help the model recognize year-to-year correlations in case trends. All predictors—including climate, land cover, and socioeconomic factors—were normalized using z-scores to standardize scale and reduce feature dominance. The XGBoost model was tuned through iterative hyperparameter optimization to balance bias and variance.

Test

Model accuracy improved substantially, with root mean square error (RMSE) dropping from 32.5 to 21.8 across cross-validation runs. However, predictions remained unstable in sparsely populated counties, where limited data led to exaggerated case estimates and inconsistent trend extrapolation.

Learn

The introduction of temporal correlation improved predictive accuracy, but regional bias persisted. This revealed that population density and data sparsity significantly affected model reliability, motivating the inclusion of spatial smoothing and demographic weighting in later iterations.

Iteration 3 - Enhanced Spatial Realism

Design

In the third iteration, we enhanced spatial realism by grouping counties into clusters based on geographic proximity and inferred human mobility patterns. This approach aimed to reflect real-world transmission dynamics, where movement between adjacent or socially connected regions influences disease spread more than distant interactions.

Build

Latitude and longitude coordinates were converted into spatial embeddings, allowing the model to recognize neighboring relationships numerically. Additionally, income strata were introduced as categorical socioeconomic indicators, enabling the model to account for disparities in healthcare access and reporting accuracy. These adjustments improved both the model’s spatial sensitivity and its contextual awareness.

Test

The updated framework generated smoother predictions, reducing abrupt case fluctuations across neighboring counties. Hotspot formation became more consistent and stable over time, with regional clusters aligning closely to historical surveillance data.

Learn

Integrating geographic and socioeconomic context reduced noise from underreported or sparsely populated regions, leading to more coherent spatial risk mapping. This iteration underscored the importance of embedding human and environmental connectivity into predictive epidemiological modeling.

Iteration 4 - Validation of Model

Design

We validated the model through rolling cross-validation across temporal subsets to ensure consistency in multi-year forecasting. To extend the framework’s global relevance, we integrated subnational European datasets, adapting the model for different reporting standards, environmental scales, and healthcare systems.

Build

The model was refined to handle variations in population scale and data density between U.S. counties and European NUTS2 regions. Normalization layers were updated to adjust for regional reporting discrepancies, while feature distributions—such as temperature, forest cover, and income—were rescaled to maintain comparability across datasets.

Test

Performance metrics demonstrated strong predictive stability, with mean RMSE converging around 20.5 across all validation folds. The model consistently localized hotspots corresponding to known endemic regions in both North America and Europe, confirming its adaptability to distinct epidemiological contexts.

Learn

After four iterative refinements, the HP model evolved into a robust and globally generalizable predictive framework. Its integration of demographic, environmental, and spatial-temporal features enables reliable forecasting of Lyme disease trends across diverse geographies, supporting early intervention and resource allocation worldwide.