Parts Creation and Success
Detection relies on identification, amplification, and reporting of a specific DNA region of the organism's genome. For our purposes, we searched for a highly conserved and nonpathogenic region of the Bd and Bsal genomes and found the ITS region. The ITS (internal transcribed spacer) region is highly conserved and widely used as the DNA barcode region in fungi such as Bd and Bsal. It is a spacer DNA that is nonpathogenic, adding an additional layer of safety; this will be the region amplified in downstream reactions such as RPA or LAMP.
We created a gblock part for the Bd ITS region and Bsal ITS region. Our experiment relies upon DNA extraction, amplification, and detection. For the latter two, we designed primer parts within our gblock region. These included those for PCR, qPCR, RPA, LAMP, and CRISPR. In the three weeks we had access to a biology lab space, we were able to confirm success of our gblocks and PCR and RPA primers, along with CRISPR guide RNA for Bd. See our results page for a more detailed description of how we confirmed success of these parts!
Diagnostic Device Engineering Design Process
The development of the diagnostic device has been following the engineering design cycle very closely. The team first started out with a very extensive research phase where we compiled all the existing literature regarding each component and process of the device and documented on the Notion platform The complete biological diagnostic pathway was also determined to help the engineering team divide the device into distinct components, as can be seen in the diagrams below:
Pipeline diagram of device pathway
Mapping diagram of device pathway
Meanwhile, we also reached out to stakeholders and experts to set constraints on the device Questions for Stakeholders from Engineering Side, Metrics and Requirement ). Using this knowledge and keeping the constraints in mind, for each process, both in biology and engineering, we have been researching which options could be implemented in the device. Based on these options, we have been also brainstorming alternate options, such as in case of reaction medium, and coming up with innovative ways to put the processes together as in case of sample preparation and reaction medium.
Using our selected options, which are still increasing as we brainstorm further, we are designing experiments using standard and proven methodologies, keeping factors such as errors, need of replicates, and experimental controls in mind. While our lab access has been very limited so far due to COVID-19, we have reached the initial prototyping stage for some processes such as fluorescence, have experimental results for RPA amplification, and we plan to prototype and test other options soon. Using the results from these tests, we will be improving or discarding the options to develop concepts which will be then combined under the constraints to provide a complete device design. Finally, this design will be prototyped, tested, and improved to become field deployable.
The complete engineering product development timeline can be seen here:
Engineering product development timeline
Given below are some brief examples of components of our diagnostic process that are meant to further explain the use of engineering design cycle and the engineering success in our methodology:
Engineering Success- LFA sensing Mechanism
Research and Imagine
This process involved us conducting research in order to implement an experimental analysis design that can then be tested and improved accordingly. This involved studying articles that used LFA as a sensing mechanism (especially those that used LFA with CRISPR based reactions).
After comparing articles and the methods they used, we implemented methods that seemed to be the most feasible in terms of Point of Care testing, and the methods that seemed to be most cost-effective and time-efficient
Why should we use Milenia HybriDetect?
When deciding on the brand of LFA strips to be used, we found that the Milenia Hybridetect 1 strips were highly popular among researchers and users and has been cited in a series of high quality publications. A list of all the publications can be found here . The products were suitable since we could detect products derived from amplification methods including RPA, PCR and LAMP. The strips are of reliable quality and are readily available.
For quantitative analysis, we tried to look for equipment such as those produced by Milenia Biotech, that would grant quantification of these results however, these were not proven to have reliable results. Therefore, we began looking at ways through which we could engineer a method to quantify these LFA strips.
After intensive research, we decided to follow the technique outlined in "Analyte Quantity Detection from Lateral Flow Assay Using a Smartphone," by the authors Foysal, Kamrul H et al. In this article, the writers carefully explain developing an image processing software which was used to quantify LFA results.
Image-processing software was then created and designed using Python, replicating the method explained in the article. Some adjustments were made depending on the requirements of out experiment and the expectation of the results. Therefore, the research that went into preparing the experimental analysis for quantitative LFA detection involved a lot of interdisciplinary collaboration between computer science studies and bio technology. This collaboration helped us derive an experimental design that can be used to successfully quantify the DNA. The development of the experimental design for LFA quantification involved research in both computer science and bio technology.
Design and Build
We designed the experiments using data from our research explained above. Experimental design of the qualitative and quantitative analysis is as follows:
Experimental Design for Qualitative Analysis
1) After completion of the pre-amplification step, there will be a 20 μL of reaction mix. This will be added to a 2-mL Eppendorf tube after which 80 μL of Hybridetect 1 Assay Buffer should be added into it. A nuclease-free 2-mL 96- well block can be used instead of Eppendorf tubes.
2) A lateral flow strip (Milenia HybriDetect 1, TwistDx) should then be added to the reaction tube and a result should be visualized after approximately 2 min. The lateral flow strip must be placed in an upright position and must be incubated at room temperature for at least 5 minutes. (unto 15 minutes). Results must then be analysed.
A single band, close to the sample application pad (the control band) will indicate a negative result, whereas two bands close to the top of the strip (test band) and the control band should indicate a positive result. The bands are clearly formed red bands. In case of very high concentrations of hybridisation product, control band’s intensity may be affected (during a positive test) Nevertheless, the control band should be still visible clearly.
3) After completion, remove the lateral flow strip and place it on a white background for visual inspection.
Experimental Design for Quantitative Analysis
The analyte quantity in the sample is expected to be proportional to the intensity of the red colored test line region. Hence, a weighted summation of red color in the test line region can be considered as the parameter for the detection of analyte quantity, where the weight is considered to be the color intensity of each pixel. The intensity of the red could then be quantified by image acquisition followed by an image processing software to measure the band intensity and to determine a positive detection against a positive-control dilution series.
To develop the software to be ready for quantification, data points have to be obtained for a calibration curve to be formulated. Once this has been fed into the system, images to be tested will be studied against the calibration curve to determine a quantified result for the test.
The method below refers to the creation of the calibration curve. Firstly, data points have to be obtained to formulate the curve.
1) Firstly, we need to get known quantities of the target analyte. We hope to consider 5 different concentrations, such as: 10 ng/mL, 1 ng/mL, 100 pg/mL, 10 pg/mL, and 1 pg/mL.
2) Three sets of samples must be obtained for this, with each set containing 5 LFA strips of different concentrations. 10 μL of the sample should be prepared. The quantities we could study are therefore : 100pg, 10pg, 1pg, 100fg, 10fg. 15 LFA strips will be used. Each reading should be taken 5 times to reduce error. 75 readings should be obtained overall. The sets of readings will be used to train the machine learning model.
3) Using the grid view of the camera, the region of interest (ROI) should be positioned inside the center box. The image will then be captured, and the ROI obtained. The data acquisition procedure will be performed with a smartphone. Which brand has not been anticipated yet, but it should be kept constant. The flashlight would be turned off during the data acquisition procedure. Images of the LFA strips will be captured under different ambient lighting conditions in the laboratory. The illumination power of the room must be noted. The quantification method will be evaluated based on acquired data using MATLAB.
Next, we analyse the results using the method below, to obtain the regression and classification of our points.
Feature extraction- The number of red pixels accumulated on the test line is considered the feature. Assuming that the cumulative sum of red pixels’ intensities in the test line varied proportionally with analyte quantity under a specific lighting environment, we carried out the test. In order to do this:
The testing procedure is as follows:
(1) A test strip is cut according to consistent measurements with only the test line and control line regions visible. It is placed under the smartphone on a white background. Using the grid view of the camera, the ROI was positioned inside the center box.
(2) The image was captured. The application then created a mask using the preprocessing technique mentioned in the proposed algorithm (Otsu's method).
(3) In the masked region, the weighted sums of red pixels’ intensities of the test and the control line regions were calculated.
(4) T/C ratio values that corresponded to the quantity of the analyte were calculated.
(5) A regression curve is plotted which will be used as the calibration curve.
Citation: Foysal, Kamrul H et al. “Analyte Quantity Detection from Lateral Flow Assay Using a Smartphone.” Sensors (Basel, Switzerland) vol. 19,21 4812. 5 Nov. 2019, doi:10.3390/s19214812
Test, Learn and Improve
We plan on testing our experimental designs explained above, and repeat with any alterations that may be required after detailed observation of the results.
Engineering Success - Fluorescence
To start our process, we started researching different possible methods for fluorescence detection in Point of Care devices. This detection mechanism was one of our options as we plan on taking advantage of the fluorescence quenching-based reporting that can be easily combined with CRISPR-Cas reactions. We mapped all possible methods that included different types of excitation sources, sensors and setups using different case studies. A full list of this can be found on the Fluorescence — page. We then classified these methods based on their sensitivity and cost as the time taken in this method is negligible for all setups.
Mapping of fluorescence options
After the research was completed, we concluded on two possible methods, 1 of which involved filters and 1 that did not. The dichroic mirrors that are used as filters are expensive but increase the overall sensitivity of the device. Therefore instead of just picking one between cost effectivity and sensitivity, we decided to test both setups once we got lab access. The final components we decided can be found on the Hardware page.
Design and Build
We designed a shell to block the light out of the detection medium to increase the signal to noise ratio. Finally, we developed a code to run initial tests and check the sensor that we plan on using.
Fluorescence detection schematic
We plan on testing this initial setup with amplified samples and control samples using the SYBR green fluorescence dye provided by the bio team to calibrate the device. The calibratrion would be done as follows:
The device should then be able to detect whether an unknown sample is positive or not using the above data
1. We will first take 20 samples that we know are positive for the target DNA. We will place these in the prototype and check the sensor readings for both the Green and Blue component.
2. We will repeat the above steps for 20 control (negative) samples as well.
3. Finally, we will incorporate the values we obtained in steps 1 and 2 into the code and use this to predict if an unknown sample placed in the prototype is positive or not.
Learn and Improve
We will then make the necessary changes and recalibrate it to sense quantitatively for a positive sample using CRISPR. The recalibration process will involve the following steps:
1. Preparation of multiple samples with different concentrations of the target DNA.
2. Find the average sensor value for 20 repeats of each concentration.
3. Plot a graph of the average sensor value (x axis) against concentration (y axis)
4. Use this in the code to give a quantitative result
Once the above processes are completed successfully for both our methods, we will compare them based on biology and engineering requirements and make a decision on which method to move forward with. We hypothesize that the dichroic mirror method will be more sensitive but significantly more expensive while other parameters such as size and time taken would be similar.
We will then repeat the above steps and try and make the device more sensitive and durable.