Micro-Q Pro

Overview/Background

Traditional methods of diagnosing coronary artery disease (CAD) such as angiograms and echocardiograms are tedious and expensive, with costs ranging from $500 to $20,000. The length of time for diagnosis is prolonged by the fact it takes multiple days to receive test results for patients (Fogoros, 2023). To increase accessibility of CAD diagnostics, our team at Lambert iGEM developed a cost-effective biosensor that detects microRNA (miRNA) biomarkers correlated with CAD progression. Our biosensor utilizes rolling circle amplification (RCA) to correlate fluorescence intensity to specific miRNA concentrations, allowing doctors and clinicians to monitor CAD progression (see RCA). However, existing fluorometers are bulky, immobile, and have prices upwards of $10,000, making point-of-care testing (POCT) impractical (Hixson et al., 2022).

In 2022, our team developed Micro-Q: a ~$15 PCR-tube fluorometer capable of quantifying fluorescent samples in seconds with 97.01% accuracy (See Lambert iGEM Hardware, 2022). This affordable solution, paired with its user-friendly mobile app, increases the accessibility of POCT in underfunded labs. However, after testing Micro-Q in Thailand, a couple of limitations were revealed: 1) quantifying batches of samples is tedious; 2) tuning excitation spectra for diverse fluorophores is complicated. Ultimately, these issues make testing and experimentation inefficient.

This year, we addressed these issues by developing Micro-Q Pro—a camera-based fluorescence viewer under $10 capable of quantifying up to 8 PCR tubes simultaneously across the full visible wavelength spectrum. Since this detection uses camera imaging rather than photodiodes, users can record fluorescence videos and create real-time curves by snipping images at certain time points with 95.37% accuracy. This open-sourced fluorometer enables accurate and rapid sample quantification for diverse applications.

Fluorescence Background

Fluorescence is a type of luminescence that occurs when particular substances absorb light at one wavelength and then emit light at a longer wavelength. When a fluorescent sample is exposed to a specific excitation wavelength, it absorbs the light energy, raising its molecules to higher energy states. As these excited molecules return to lower energy ground states, they release excess energy as fluorescent light (see Fig. 1).

This fundamental process enables diverse applications for fluorescence, including bioimaging, medical diagnostics, forensics, and environmental monitoring. By tuning excitation light and detecting emission wavelengths, researchers can selectively visualize, quantify, and track fluorescent probes for many purposes.

Figure 1. The process of fluorescence excitation and emission from a sample.

Design

Parts List

All the codes, steps for assembly, and CAD files can be found in the open-source Github repository: Micro-Q Pro Files.

Light Source

Last year, Micro-Q used a singular 405 nm laser for high-precision green fluorescent sample excitation, minimizing needed software processing. However, when presented to Kiatichai Faksri of Khon Kaen University in Thailand, he commended its single-tube quantification accuracy but noted difficulties switching excitation lasers and filters to quantify different materials. Commercial spectrophotometers utilize white light lasers, which cover the full visible spectrum, in conjunction with diffraction gratings to choose the optimal excitation wavelength for a given biological sample (Dondelinger, 2011). However, these lasers often exceed $3,000, as they combine an array of laser wavelengths into a single compressed laser. Additionally, the many moving parts make these systems prone to breaking, while increasing size and bulkiness. A cost-effective alternative is a 5000K LED emitting the full 350-750 nm visible spectrum (Liu et al., 2018). To test our LED, we excited green fluorescein dyes at 30°, 45°, and 60°, finding 45° optimal for fluorescence visualization with minimal glare (Fig. 3). Additionally, we coated the inside of Micro-Q Pro with Musou Black Paint, which absorbs 99% of visible light, to eliminate excess ambient light for clearer imaging (Chu, 2019).

Figure 3. Fluorescein sample excited at A) 30° B) 45° C) 60°; 45° showed least glare and best fluorescence.

Fluorescence Quantification

The original Micro-Q design utilized a photoresistor (a light intensity sensor) behind an emission filter to quantify the fluorescence intensity of the sample. While this design worked relatively well for discerning between high fluorescence concentrations, it struggled to differentiate between lower ones. Its accuracy could significantly be improved by replacing the photoresistor with a photodiode (a more sensitive light intensity sensor), however this year, Lambert iGEM implemented a camera-based detection system to enable real-time visualizations and quantification of sample fluorescence. Additionally, with this quantification mechanism, more samples can be quantified in a fraction of the time. This camera-based detection enables users to automate fluorescence curves by quantifying fluorescence across video segments.

To quantify fluorescence, the Micro-Q Pro ESP32 camera captures two images: 1) a blank image with just light; 2) a sample image with fluorescent tubes (Fig. 4). Then, the algorithm subtracts the blank pixels (image 1) from the sample image (image 2) to remove non-fluorescent features, and isolate the sample’s fluorescence intensity. The image is then preprocessed to improve quality before calculating RFU values. Preprocessing involves median blurring to reduce noise, inverting the image, and applying gamma correction. This compensates for the non-linear relationship between true and measured sample brightness (see Fig. 5).

Figure 4. Software pipeline of Micro-Q Pro ESP-32 Camera.

Figure 5. Tubes of fluorescein from 200µm 1.5µm a) before preprocessing b) after preprocessing.

After preprocessing, Otsu’s thresholding masks out the remaining background noise, leaving only the foreground fluorescence signal. Individual binary masks are created for each PCR tube, and the average pixel brightness within each mask is calculated to determine the RFU values (Fig. 6). Since the masking algorithm will isolate the sample of fluorescence from excess light, users can quantify multiple samples without emission filters, unless a tube contains multiple fluorophores.

Figure 6. Tubes of fluorescein from 200µm 1.5µm a) original b) after masking fluorescence.

3D-Printed Shell:

The original Micro-Q device was initially designed to easily quantify individual PCR tubes. To enable multisample and multispectral quantification, Micro-Q Pro was redesigned as a 10.8 cm x 8.7cm x 8.3 cm 3D printed box with three key compartments. A sample chamber, an exchangeable emission filter, and a camera holster (see Fig. 7). The sample chamber was modified to hold a strip of 8 PCR tubes separated by dividers to minimize crosstalk between fluorescent signals, which could skew the data (Arrpe et al. 2017).

Moreover, Micro-Q Pro uses a 45-degree angled white LED strip paired with the emission filter to detect fluorescence across all relevant wavelengths. The ESP-32 camera module is placed 7.8 cm away from the samples to ensure it can view all 8 tubes at once. These optimized elements work seamlessly together to enable robust, high-quality fluorescence imaging of multiple samples.

Results and Analysis

Preliminary Testing

To validate Micro-Q Pro’s accuracy, we quantified a range of fluorescein concentrations (0-200 μM) similar to GFP working ranges (de Jonge et al.). Triplicate PCR tubes of 8 were quantified for each of the concentrations below.

Initial comparisons of raw relative fluorescence values (RFU) from Micro-Q Pro, and a plate reader revealed a logarithmic relationship, rather than the expected linear correlation between fluorescence and concentration (Fig. 8)(Itagaki, 2000). After further investigation, we discovered this was due to the camera’s reduced sensitivity at higher concentrations, causing oversaturation. This means small perceived camera differences equate to larger fluorescence intensity differences, as proven by the plate reader. To account for this non-linearity, the data requires processing.

Figure 8. Raw RFU values calculated from the original quantification algorithm).

To linearize our data, we applied a power series regression on all the trials from the 3 triplicates. The curve below achieved the highest R2 value when linearized, so we chose to apply this formula to linearize any further results (See Fig. 9).

Figure 9. Power series regression to linearize Micro-Q values to plate reader values.

Applying the power series regression formula to the previous triplicate test, we get a linear relationship (See Fig. 10).

Figure 10. Micro-Q Pro triplicate results after power series linearization.

Final Results

After confirming the linearization of the RFU values from Micro-Q pro, we reran triplicates of the same experiment, and compared the results to the plate reader. Since relative fluorescence units (RFU) are relative, we scaled the output of Micro-Q to match the scale of the output from the plate reader so that they can be compared (see Fig. 11).

Figure 11. Measurements from plate reader and Micro-Q Pro from 0-200 μM.

In order to have an accurate comparison of the data from the plate reader and Micro-Q, we added points at the origin to our data set, and calculated a slope from a linear regression for each measurement device, assuming a y-intercept of 0. The slopes of the data are ~72.1 and ~75.6 from Micro-Q Pro and the Plate reader, respectively, achieving a percent error of -4.629%

Future Directions & Conclusion

Micro-Q Pro demonstrates promising accuracy in quantifying green fluorescent samples, with no significant differences between Micro-Q Pro and plate reader outputs. This validates the fluorescence measurement capabilities of our hardware device. Although we primarily characterized performance with green fluorophores, preliminary data with red fluorescent materials is also encouraging.

Future iterations could incorporate internal heating pads, enabling fully autonomous fluorescence curves creation during incubations. We also aim to develop an intuitive mobile app for live-streaming fluorescence data and make sample quantification as simple as a click of a button. Additionally, we will resume full characterization across additional spectra once we return to the lab.

Micro-Q Pro retains the core functionalities of the original Micro-Q, while dramatically enhancing the ability to efficiently quantify multiple samples across diverse fluorescent spectra—all within a simple $10 device.

As characterization of Micro-Q Pro continues on different fluorophores, Micro-Q Pro makes robust fluorescence quantification widely accessible regardless of financial limitations. Ultimately, this technology enables portable, efficient point-of-care diagnostic testing our miRNA biosensors and other fluorescence-based assays.

References