Overview
Our project is a unique and comprehensive project that produces satisfactory results in the modeling system, synthetic biology system, and hardware system. This project aims to realize the programmable quantitative detection of a variety of miRNAs through the enzyme-free isothermal nucleic acid amplification technique HCR coupled with CRISPR/Cas12a technology. On this basis, a portable paper chip and incubator suitable for various application scenarios, combined with a smartphone and a classification model, are developed to realize CMP classification and assessment. Below are some engineering design cycles in each system.
miRNA Model
In order to be able to screen out two reliable miRNAs as our pain biomarkers, we need to optimize the screening methods and results obtained in the literature many times, and combine them with our concept to optimize the technology. Our screening model is solved under the local optimum (considering 119 miRNAs). Finally, hsa-miR-98-5p and hsa-miR-7d-5p were obtained as biomarkers.
Design
We chose to develop from a miRNA dataset obtained from an existing study by Dayer[1] et al.: screening from 119 miRNAs that can be detected in the blood.
Build&Test
According to the experimental data of miRNA directly related to pain, we first used the ROC curve to evaluate the importance of each miRNA. The results of the ROC curve showed that the two miRNAs with the highest importance were hsa-miR-98-5p and hsa-miR-7d-5p, respectively. However, according to the random forest results of Dayer et al., hsa-miR-98-5p and hsa-miR-320a were the best results, so we decided to reverify the screening results.
Learn
We found that the results obtained by the separate verification models suffered from high randomness. The results obtained by the random forest would fluctuate and differ in each screening, so we need to repeatedly verify the results through multiple experiments.
Design
Owing to the problems shown in the results, we adopted other models, including logistic regression, random forest, chi-square test, and ANOVA, on the basis of unchanged original ideas, and conducted multiple comparative experiments to study the reliability of the results.
Build&Test
Through repeated verification and screening, we finally concluded that hsa-miR-98-5p and hsa-miR-7d-5p were the top two miRNAs in importance.
Design
Since there is no clear correspondence between miRNA and pain, we need to learn and classify through some models that do not need explicit classification criteria. The decision tree model is used to automatically learn the features of the data, which is used as our final classification model.
Build&Test
We divide the existing clinical data into a training set and a test set. By building a decision tree, we finally obtained a model that could output the pain type after receiving the miRNA concentrations as input. By evaluating the importance of hsa-miR-98-5p and hsa-miR-7d-5p, the concentrations of the two miRNAs were put into the model, and then the final classification results were computed according to the importance obtained previously. The specific training parameters are detailed on the modeling page.
For more information, please click here to jump to the "Modeling" page
Learn
The final classification results showed that the importance of hsa-miR-98-5p far exceeded that of hsa-miR-7d-5p, so we chose to give priority to hsa-miR-98-5p for biological and hardware technology verification.
Biology
Design
To accurately distinguish between nociceptive and neuropathic chronic musculoskeletal pain, we conducted extensive research on pain-related markers in body fluids that are easy to obtain and discussed the feasibility of various pain-related markers for this study from multiple perspectives such as metabolism and genes.
After investigation, we found that miRNA has the function of determining the origin of chronic pain. Its expression is highly variable in patients with pain and its stability in plasma and serum is strong. Therefore, we decided to select miRNA as our biomarker.
Furthermore, we summarized and screened 14 miRNAs related to pain classification from literature [2], designed a variety of algorithms to establish models, and selected two highly correlated c-miRNAs, hsa-miR-7d-5p and hsa-miR-98-5p, as biomarkers for pain classification, so as to distinguish between the neuropathic and nociceptive origins of pain.
To detect the target miRNA, we identified HCR (Hybridization Chain Reaction) as our biological signal amplification method by searching the literature. The HCR has the following advantages: it does not require the assistance of enzymes and has mild and easy-to-control reaction conditions, it also requires simple instruments and is highly compatible with various types of detection.
Build&Test
We referred to the probe design in the paper of Jia and others [3] on microRNA detection using HCR and Crispr Cas technology and designed HCR probes for our targets miR-7d-5p and miR-98-5p. The results show that it can perform normal reactions, but there are some problems such as high false positive results and low reaction limit.
1: H1+H2+H3+ miRNA (300nM); 2: H1 +H2+H3+ miRNA (300nM);
3: H1+H2+H3+ miRNA (300nM); 4: H1+H2+H3+ miRNA (300nM);
5: H1-2+H2+H3; 6: D2000
Learn&Design
Based on the previous optimization, in order to further reduce the occurrence of false positive results. Referring to literature [4], we modified and optimized the H1 sequence, and optimized the design according to the stem length of H1 and the mismatch of the stem ends of H1, H2, and H3.
Build
We conducted simulation experiments on H1 stem length and tail mismatch.
In terms of H1 stem length, we used the simulation function of http://www.nupack.org/ to perform simulation tests of HCR responses under different designs. When designing the length of the H1 stem-loop, we conducted simulation tests with the length of the H1 stem-loop ranging from 7 to 22 when only H1 and miRNA were added and when H1, H2, H3, and miRNA were added. The test results show that the experiment effect is better when the H1 stem length is 11.
For more information, please click here to jump to the "Modeling" page
Further, in terms of probe tail mismatch, we adjusted the mismatch of H1, H2, and H3 stem tails when the stem length of H1 was 11. The simulation test showed that the effects of various mismatch adjustments were poor, false positives increased and even the reaction could not occur. In particular, different from the original literature we referred to, our mismatch attempt on H2 is still poor, which may indicate that the mismatch is uncertain and cannot bring deterministic effective results.
Test
We also conducted experiments in the laboratory using H1 of different stem lengths to compare their HCR effects. The results showed that for miR-98-5p and miR-7d-5p when their H1 stem lengths were 7, 9, 11, 13, and 15, HCR responses could be triggered, and their responses were best when the stem length of H1 was 11.
1:miRNA(300nM)+H1(15)+H2+H3;2:miRNA(300nM)+H1(13)+H2+H3;
3:miRNA(300nM)+H1(11)+H2+H3;4:miRNA(300nM)+H1(9)+H2+H3;
5:miRNA(300nM)+H1(7)+H2+H3; 6:H1(11)+H2+H3; 7:D2000
1:miRNA(300nM)+H1(15)+H2+H3; 2:H1(15)+H2+H3;
3:miRNA(300nM)+H1(13)+H2+H3; 4:H1(13)+H2+H3;
5:miRNA(300nM)+H1(11)+H2+H3; 6:H1(11)+H2+H3;
7:miRNA(300nM)+H1(9)+H2+H3; 8:H1(9)+H2+H3;
9:miRNA(300nM)+H1(7)+H2+H3; 10:H1(7)+H2+H3; 11:D2000
Learn&Design
After the aforementioned optimization of probes in HCR, we also optimized the reaction system of HCR. We carried out optimization experiments for probe concentration and Na+ concentration.
Build&Test
After we verified the feasibility of the HCR reaction for the two miRNAs, we optimized the relevant reaction conditions and tried to change the concentration of the H1 probe and Na+ concentration to select the most appropriate reaction conditions.
In the experiment of probe concentration optimization, we carried out HCR reactions with miRNA under the conditions of H1 concentration of 0.1μM, 0.3μM, 0.6μM and 0.9μM respectively, and found that the best reaction was achieved at 0.6μM.
3: miRNA(400nM)+H1(0.6μM)+H2+H3+Na+；4: miRNA(400nM) +H1(0.4μM)+H2+H3+Na+
5: miRNA(400nM) +H1(0.2μM)+H2+H3+Na+；6: miRNA(400nM) +H1(0.1μM)+H2+H3+Na+
7: H1(0.6μM)+H2+H3+Na+；8: D2000
1: D2000; 2: H1(0.6μM)+H2+H3; 3: miRNA(300nM)+H1(0.1μM)+H2+H3;
4: miRNA(300nM)+H1(0.3μM)+H2+H3; 5: miRNA(300nM)+H1(0.6μM)+H2+H3;
6: miRNA(300nM)+H1(0.9μM)+H2+H3
We conducted gradient experiments of 95nM, 140nM, 350nM, 550nM and 750nM for Na+ concentration, of which 350nM has the best Na+ reaction.
1: miRNA(400nM)+H1(0.6μM)+H2+H3+Na+ (750mM);
2: miRNA(400nM)+H1(0.6μM)+H2+H3+Na+(550mM);
3: miRNA(400nM)+H1(0.6μM)+H2+H3+Na+(350mM);
4: miRNA(400nM)+H1(0.6μM)+H2+H3+Na+(140mM);
5: miRNA(400nM)+H1(0.6μM)+H2+H3+Na+(95mM);
6: H1(0.6μM)+H2+H3+Na+(350mM); 7:D2000
Hardware
Design
According to the requirements of our biological route, we expected to build a platform for point-of-care testing in the hardware module, which can realize detection of multiple miRNAs targets in channels, and meet the goal of convenient operation and user interaction with software. So we should select the most suitable platform for our project ideal response platform through our research.
Build
We planned to launch a series of research, on the basis of previous research which focused on hydrogel, microfluidic centrifugal chip and paper-based chip as our form of platform construction.
Test
During our research, we found that the shape coding function of hydrogels is in line with our needs. The specific reaction in hydrogels can produce corresponding signal output, and the advantage of shape coding hydrogels is that they can detect multiple targets in the same chamber. However, in our biological route, the specific reaction is the HCR reaction, and the reaction that produces fluorescence is not a specific reaction. The products of both miRNAs after HCR can activate Cas protein cleavage to produce fluorescence. That is to say, the specificity of the reaction is lost in the last step of fluorescence generation, and then shape-encoded hydrogels cannot be used in our project.
Microfluidic centrifugal chips and paper-based chips can meet our needs well in comparison. However, considering the cost and portability issues, we want to build a relatively simple device. Centrifugal chips require additional motors as a driving force to achieve multi-channel detection. In conclusion, paper-based chips are simpler to achieve our goal.
Learn
Finally, we chose paper-based chips as our platform for biological reactions.
Design
We decided to build a paper-based chip with 10 mm × 10 mm squares as our reaction platform, and with temperature control device, in order to achieve the ultimate goal of portable detection. At the same time, we needed to ensure that the two layers of PES filter in the chip are well hydrophobic to ensure that the liquid can reach the reaction pad in the way we want.
In order to achieve this goal, we need to (1) Determine the processing methods of paper-based chip, including the number of holes on the paper chip and the hydrophobic processing methods of PES filter ; (2) Build a circuit board that can control the temperature.
Build
(1) Processing methods of paper-based chip
Considering that we wanted to perform a repeated experiment of a sample on a chip and set negative quality control on the chip, we constructed a paper-based chip with a six-hole structure :
For hydrophobic treatment, we intended to use photoresist to form a hydrophobic layer on the PES filter to achieve our goal.
(2) The setting of temperature control circuit
We set up the temperature control circuit to control the temperature in the form of a heating sheet as follows:
Test
(1) Processing methods of paper-based chip
Setting of the number of holes
We found that it is difficult to process the six-hole paper-based chip during the actual processing. It is hard to ensure the symmetrical distribution of the holes. Another serious problem is that the reaction pad we designed is too small to load enough amount of liquid.
Hydrophobic treatment
Through our research, we found that although the photoresist method can achieve the purpose of hydrophobic, the isoprenol used in the last step of etching in the operation is a polar solution, while the PES filter will dissolve in the polar solution, so we can not hydroponically treat the PES film in this way.
(2) The setting of temperature control circuit
After testing, our temperature control circuit can meet the needs of temperature control.
For more information, please click here to jump to the "Hardware" page
Learn
We initially designed a paper-based chip with a six-hole structure and considered using the photoresist, but the actual situation is not in line with our expectations. Obviously , it needed further improvement.
What excited us is that we had built a temperature control circuit that met the requirements. However, it also needed further improvement.
Design
In Cycle 2, we found that six-hole paper-based chips are difficult for us to produce, and the hydrophobic mode needs further consideration. At the same time, we wanted to set up a shell for the heating plate we built.
Build
(1) Processing methods of paper-based chip
Setting of the number of holes
Based on the six-hole structure, we reduced the number of holes and designed the following four-hole structure :
Hydrophobic treatment
We planned to test whether our purpose can be achieved by using crayon smear, paraffin smear or the method of soaking pes filter in molten paraffin to form wax printing layer.
(2) Construction of incubator
As follows, we built an incubator for our temperature control circuit as the circuit shell :
Test
(1) Processing methods of paper-based chip
Setting of the number of holes
This time, we successfully processed a four-hole structure of the PES filter, and with the subsequent hydrophobic treatment, we also successfully processed a complete paper-based chip, and its four holes have good uniformity.
For more information, please click here to jump to the "Hardware
Hydrophobic treatment
We tried to use the method of soaking wax, but found that the wax printing layer was not uniform.
For the choice of hydrophobic method, we conducted a simple experiment. After dropping a drop of water droplets on the four treated PES filters and placing them for a while, except for the water droplets on the film coated with crayons, the water droplets on the PES filters treated by other methods did not condense.
Learn
In Cycle 3, we finally confirmed that the paper-based chip with four-hole structure as the final processing methods, and the hydrophobic treatment of the PES filter was carried out by crayon smearing. At the same time, we built an incubator as the shell of our temperature control circuit.
Design
The volume of the incubator we constructed was relatively large. We hoped that after further modification, the circuit and its shell would be smaller, which meets our demand for portability of hardware equipment.
Build
We built a heating platform as our improved temperature control shell as follows :
Test
For the selection of equipment materials, at first, we chose resin as the material of the whole structure. During the process of practice, we found that its heating performance was not as good as expected, so we changed its heating chamber to aluminum, which improved the overall performance.
In the subsequent test, in order to control the stability of the temperature, we adopted the form of screw cover plate. By adjusting the height of the screw, the cover plate part of the aluminum heating chamber can be pressed so that the paper-based chip and the silicone rubber heating film can be more fully contacted, while avoiding damage to the paper chip.
In order to adapt to our paper-based chip and mobile phone fluorescence detection equipment, we also designed a strip. The paper-based chip is placed in a groove in the middle of the strip, which allows it to be heated and fluorescent photographed in a fixed area while conveniently placing the chip.
Learn
By continuously improving the temperature control equipment and the overall structure in the actual test, we finally formed a system which could fit our chip well.
CRISPR Model
Design
In the CRISPR system adopted in this project, crRNA can specifically recognize the amplified sequence of target miRNA, and then activate the side-chain cleavage activity of Cas12a protein. Cas12a protein can specifically recognize and cut fluorescent probes, thus realizing specific signal detection.
We hope to fit the kinetic indexes KM and kcat of the enzyme based on the fluorescence curve of the experimental output, so as to seek the direction of experimental improvement in theory.
Build&Test
In the beginning, we found that the initial reaction velocity is inversely proportional to the Michaelis-Menten constant KM by using the classical Michaelis-Menten equation.
Michaelis-Menten constant KM is a characteristic constant of enzyme, which can be used to express the affinity between enzyme and substrate. The smaller the enzyme is, the higher the affinity between the enzyme and substrate. Therefore, we can measure the KM of the enzyme at different temperatures and PH, and the experimental conditions with lower KM are better.
But this is obviously not enough if we want to make predictions about experimental outcomes, including reaction times. Furthermore, if we have a solution for the amount of product as a function of time, we can predict the results of the experiment, including the approximate detection time; The parameters can also be analyzed to find the direction of optimization experiment; The correctness and self-consistency of the experimental results can also be evaluated, that is, if the experimental results deviate greatly from the analytical solution, the correctness of the experimental results is worthy of discussion.
Therefore, we improve the classical Mie equation on the basis of an additional condition of CRISPR experiment and obtain the analytical solution, so as to meet our needs.
Learn&Design
As mentioned in literature [5], the reporter gene concentration in CRISPR reaction is generally much lower than KM. Therefore, the Michaelis-Menten equation is improved by combining equation (1), and the ordinary differential equation is solved to obtain:
Where
Clearly, τ decreases with increasing
Build&Test
Therefore, based on this analytical solution and a series of CRISPR experiments, we proposed three indicators to evaluate the self-consistency and correctness of the experiment. By analyzing the influence of the parameters in equation（2）on the solution, the improvement direction of the experiment is put forward. The experiment was predicted by plotting the normalized variables with t/τ as the horizontal axis and P/S0 as the vertical axis.
Reference:
[1] Dayer, Camille Florine, et al. "Differences in the miRNA signatures of chronic musculoskeletal pain patients from neuropathic or nociceptive origins." PloS one 14.7 (2019): e0219311.
[2] Staff P O . Correction: Differences in the miRNA signatures of chronic musculoskeletal pain patients from neuropathic or nociceptive origins[J]. PLoS ONE, 2019, 14(7):e0220486-.
[3] Haiyan Jia, Hongli Zhao, Ting Wang, et al.A programmable and sensitive CRISPR/Cas12a-based MicroRNA detection platform combined with hybridization chain reaction[J].Biosensors and Bioelectronics,2022,114382.
[4] Kexin Zhao, Zhao Peng, Hao Jiang, et al.Shape-Coded Hydrogel Microparticles Integrated with Hybridization Chain Reaction and a Microfluidic Chip for Sensitive Detection of Multi-Target Mirnas[J]. Sensors and Actuators B: Chemical.Volume 361,2022,131741,ISSN 0925-4005. https://doi.org/10.1016/j.snb.2022.131741.
[5] Ramachandran, Ashwin, and Juan G. Santiago. "CRISPR enzyme kinetics for molecular diagnostics." Analytical Chemistry 93.20 (2021): 7456-7464.