Dual digital nucleic acid detection method based on non-uniform liquid drops

Abstract

The invention discloses a dual digital nucleic acid detection method based on non-uniform liquid drops, which comprises the following steps: adding a nucleic acid sample containing at least two target nucleic acids and a primer sample containing at least two corresponding specific primers into an amplification reaction system to obtain an amplification reagent; generating a non-uniform liquid drop for the amplification reagent; carrying out amplification on the liquid drops; injecting the amplified liquid drops into a micro-channel, and imaging the liquid drops in the micro-channel by using imaging equipment to obtain a liquid drop image; extracting the size of the liquid drop and the optical density value of the precipitate in the liquid drop image through an image analysis algorithm, and determining the type of the liquid drop in combination with the liquid drop characteristic curve of each type of liquid drop, the size of the liquid drop and the optical density value of the precipitate; and counting the proportion of each type of liquid drops in the total number of liquid drops, and calculating the concentrations of at least two target nucleic acids in all liquid drops according to the proportion of each type of liquid drops in the total number of liquid drops. According to the invention, label-free and multi-target digital nucleic acid detection can be realized.

Description

Technical Field

[0001] The present invention relates to the field of biological detection technology, and in particular to a double digital nucleic acid detection method based on non-uniform droplets. Background Art

[0002] Multiple target detection has significant advantages over single target detection. For example, multiple target detection can not only obtain more comprehensive and rich biological information and improve detection efficiency, but also effectively reduce the cost of expensive enzymes and reagents. Especially when the number of samples is limited or precious, multiple detection can significantly save reagents and samples. In the prevention and control of infectious diseases, many diseases show similar symptoms. The use of multiple nucleic acid detection technology can efficiently and accurately identify pathogens, thereby providing patients with timely and targeted treatment options.

[0003] Existing digital nucleic acid detection technologies generally rely on fluorescent labeling strategies, such as TaqMan probes, nucleic acid binding dyes, and molecular beacons. These methods rely on the generation and reading of fluorescent signals and must be combined with complex light sources, optical paths, and integrated modules such as cameras, resulting in high costs and difficulty in achieving portability. Therefore, the development of label-free multi-target digital nucleic acid quantification technology has become a key issue that needs to be addressed urgently. This can not only further reduce detection costs and improve efficiency, but is also expected to promote the application and popularization of multiple digital nucleic acid detection technologies in a wider range of fields. Summary of the invention

[0004] The embodiment of the present invention provides a dual digital nucleic acid detection method based on non-uniform droplets, aiming to achieve label-free, multi-target digital nucleic acid detection.

[0005] The embodiment of the present invention provides a dual digital nucleic acid detection method based on non-uniform droplets, comprising:

[0006] Adding a nucleic acid sample containing at least two target nucleic acids and a primer sample containing at least two corresponding specific primers into an amplification reaction system to obtain an amplification reagent;

[0007] generating non-uniform droplets of the amplification reagent;

[0008] Amplifying the droplets; wherein at least a portion of the amplified droplets contain precipitates;

[0009] Injecting the amplified droplets into a microchannel and imaging the droplets in the microchannel using an imaging device to obtain a droplet image;

[0010] The size of the droplets and the optical density value of the precipitate in the droplet image are extracted by an image analysis algorithm, and the type of the corresponding droplet is determined by combining the droplet characteristic curve of each type of droplet, the size of the droplet and the optical density value of the precipitate; wherein the droplet characteristic curve is a pre-fitted mathematical model that describes the relationship between the optical density value of the precipitate and the droplet size;

[0011] The ratio of each type of droplets to the total number of droplets is counted, and the concentration of at least two target nucleic acids in all droplets is calculated according to the ratio of each type of droplets to the total number of droplets.

[0012] Furthermore, extracting the size of the droplets and the optical density value of the precipitate in the droplet image by using an image analysis algorithm includes:

[0013] Using a first deep learning model to detect and segment the droplets in the droplet image, and obtain the position and attribute of each droplet in the droplet image; wherein the attribute of the droplet includes positive and negative, a positive one represents that the droplet contains sediment, and a negative one represents that the droplet does not contain sediment;

[0014] Screening out positive droplets according to the attributes, and cutting out the screened positive droplets according to the positions to obtain the areas and / or volumes of the corresponding droplets;

[0015] Using a second deep learning model to detect and segment the sediment in the cut-out droplet to extract sediment area information; wherein the sediment area information includes the location and size of the sediment;

[0016] The precipitation feature is extracted according to the precipitation area information; wherein the precipitation feature at least includes the optical density value of the precipitation in the droplet.

[0017] Further, the generating non-uniform droplets of the amplification reagent comprises:

[0018] Vortexing is used to generate non-uniform droplets of the amplification reagent.

[0019] Furthermore, the step of generating non-uniform droplets of the amplification reagent by vortex oscillation comprises:

[0020] Inject droplet generation oil as the oil phase into the loading tube;

[0021] Adding the prepared amplification reagent into the loading tube as an aqueous phase;

[0022] The loading tube was placed on a vortex shaker for shaking and standing to generate droplets of different sizes.

[0023] Furthermore, the precipitate contained in the droplet after amplification is an amplification product of a single target nucleic acid or an amplification product of at least two target nucleic acids;

[0024] The counting of the proportion of each type of droplets to the total number of droplets, and calculating the concentration of at least two target nucleic acids in all droplets according to the proportion of each type of droplets to the total number of droplets, includes:

[0025] Counting the types of each droplet, and obtaining the ratio of each type of droplet to the total number of droplets and the size of each droplet; wherein the types of droplets include droplets without target nucleic acid, droplets containing only one target nucleic acid, and droplets containing at least two target nucleic acids;

[0026] Based on the proportion of each type of droplets in the total number of droplets and the size of each droplet, the concentration of at least two target nucleic acids is calculated using the maximum likelihood estimation method.

[0027] Furthermore, the concentrations of at least two target nucleic acids are calculated using the maximum likelihood estimation method based on the proportion of each type of droplets to the total number of droplets and the size of each droplet, including:

[0028] The concentrations of the first target nucleic acid and the second target nucleic acid are calculated according to the following formula:

[0029]

[0030] Wherein, N0, N1, N2, and N3 are the number of droplets without target nucleic acid, droplets containing only the first target nucleic acid, droplets containing only the second target nucleic acid, and droplets containing both the first target nucleic acid and the second target nucleic acid, respectively; V 0,i 、V 1,i 、V 2,i 、V 3,i are the volumes of droplets without target nucleic acid, droplets containing only the first target nucleic acid, droplets containing only the second target nucleic acid, and droplets containing both the first target nucleic acid and the second target nucleic acid, respectively; i is a number; C1 and C2 are the concentrations of the first target nucleic acid and the second target nucleic acid, respectively.

[0031] Furthermore, the method of determining the type of corresponding droplets by combining the droplet characteristic curves of each type of droplets, the size of the droplets, and the optical density value of the precipitate includes:

[0032] A second-order polynomial is used in advance to fit the relationship between the optical density value of the precipitate in each type of droplet and the size of the droplet, so as to obtain a droplet characteristic curve of each type of droplet;

[0033] The size of the droplets extracted from the droplet image and the optical density value of the precipitate are substituted into the droplet characteristic curves of each type of droplets for matching, so as to determine the type of each droplet in the droplet image.

[0034] Furthermore, the concentration of each specific primer in the primer sample is determined as follows:

[0035] Presetting multiple sets of combinations of specific primers with different concentrations and adding them to the amplification reaction system respectively to obtain multiple sets of amplification reaction solutions that do not contain target nucleic acid;

[0036] Each group of amplification reaction solutions not containing target nucleic acid is divided into multiple portions, and then an excess of a single target nucleic acid and multiple target nucleic acids are added to each amplification reaction solution respectively;

[0037] Each group of multiple amplification reaction solutions is used as an aqueous phase to generate uniform droplets with known contents, and the droplets are collected, amplified, and imaged, and finally droplet images of multiple groups of uniform droplets under different primer concentration conditions are obtained;

[0038] Extracting the optical density value of the precipitate in the droplet image as a characteristic value, and drawing a characteristic value distribution histogram of various types of precipitates;

[0039] A threshold is set according to the characteristic value distribution histogram, and the droplets are classified using the threshold, and the accuracy of the droplets corresponding to each type is calculated in combination with the gold standard;

[0040] According to the accuracy of each type of droplet and the size of the precipitate, a primer concentration combination with high accuracy and large precipitate is selected.

[0041] Furthermore, the microchannel is a microchannel with a height of 46.5 μm and a predetermined floating value, and the microchannel has an inlet and an outlet.

[0042] Furthermore, the top cover temperature during amplification is higher than the amplification temperature.

[0043] The embodiment of the present invention provides a dual digital nucleic acid detection method based on non-uniform droplets, which realizes label-free, multi-target digital nucleic acid detection by utilizing the precipitation effect of LAMP amplification products and the generated non-uniform droplets. It not only maintains high sensitivity but also significantly reduces the dependence on complex equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0045] Figure 1 A schematic flow chart of a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0046] Figure 2 A first experimental schematic diagram of a double digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0047] Figure 3 A second experimental schematic diagram of a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0048] Figure 4 A third experimental schematic diagram of a double digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0049] Figure 5 A fourth experimental schematic diagram of a double digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0050] Figure 6 A schematic diagram of establishing a droplet characteristic curve in a double digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0051] Figure 7 A schematic diagram of precipitation yield in a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0052] Figure 8 A frequency histogram of precipitation amount in a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0053] Fig. 9 A schematic diagram of model classification in a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0054] Fig.10 A schematic diagram of image analysis in a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0055] Fig.11 A schematic diagram of image annotation in a double digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0056] Fig.12 A network framework diagram of a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0057] Fig.13 A schematic diagram of data set construction in a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0058] Fig.14Another schematic diagram of a process of a double digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention;

[0059] Fig.15 A schematic diagram of microfluidic chip processing in a dual digital nucleic acid detection method based on non-uniform droplets provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0062] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0063] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0064] See below Figure 1 An embodiment of the present invention provides a double digital nucleic acid detection method based on non-uniform droplets, specifically comprising: steps S101 to S106.

[0065] Step S101, adding a nucleic acid sample containing at least two target nucleic acids and a primer sample containing at least two corresponding specific primers to an amplification reaction system to obtain an amplification reagent;

[0066] Step S102, generating non-uniform droplets of the amplification reagent;

[0067] Step S103, amplifying the droplets; wherein at least part of the amplified droplets contain precipitates;

[0068] Step S104, injecting the amplified droplets into the microchannel and imaging the droplets in the microchannel using an imaging device to obtain a droplet image;

[0069] Step S105, extracting the size of the droplets in the droplet image and the optical density value of the precipitate by an image analysis algorithm, and determining the type of the corresponding droplet by combining the droplet characteristic curve of each type of droplet, the size of the droplet and the optical density value of the precipitate; wherein the droplet characteristic curve is a pre-fitted mathematical model describing the relationship between the optical density value of the precipitate and the droplet size;

[0070] Step S106: Count the proportion of each type of droplets to the total number of droplets, and calculate the concentration of at least two target nucleic acids in all droplets according to the proportion of each type of droplets to the total number of droplets.

[0071] In this embodiment, a nucleic acid sample containing at least two target nucleic acids and a primer sample containing at least two specific primers are firstly combined to configure an amplification reagent, and then non-uniform droplets are generated by the amplification reagent. The droplets are then amplified so that some of the droplets produce precipitates. The droplets are then transferred to a microchannel, and an image of the droplets in the microchannel is obtained. The size of the droplets in the droplet image and the optical density value of the droplet precipitate are extracted through an image analysis algorithm, and the type of the droplets is determined in combination with a pre-fitted droplet characteristic curve. Finally, the proportions of droplets of different types are counted, and the concentrations of different target nucleic acids in the droplets are determined based on the proportions.

[0072] This embodiment realizes label-free, multi-target digital nucleic acid detection by utilizing the precipitation effect of LAMP amplification products and the generated non-uniform droplets. It not only maintains high sensitivity but also significantly reduces the dependence on complex equipment, while further improving the accuracy and breadth of application of detection.

[0073] In actual application scenarios, this embodiment verifies the above effects through experiments. Specifically, firstly, relevant parameters and conditions suitable for the experiment are determined.

[0074] (1) Effect of heating on droplets

[0075] Since the volume of the droplets is a key parameter in the quantitative formula for nucleic acid inhomogeneity, the accuracy of the droplet volume is an important prerequisite for accurate quantitative results. If the droplets evaporate or fuse during the amplification process, it will directly affect the accuracy of the volume, thereby causing deviations in the quantitative results. Therefore, in this embodiment, droplets of uniform size are prepared and incubated at 63 degrees Celsius for 45 minutes. By comparing the droplet sizes before and after heating, it is analyzed whether there is evaporation or fusion.

[0076] By measuring the droplet size before and after heating, this embodiment found that the radius of the droplet before and after expansion was 42.89±0.70 and 42.22±1.10 μm respectively. Figure 2 As shown, Figure 2 It was investigated whether there was any droplet fusion or evaporation during the heating process. After statistical analysis, the p value was 0.57, p>0.05, indicating that there was no significant difference in the droplet radius before and after droplet heating. This data clearly shows that in droplet amplification, there was no droplet fusion or evaporation. Therefore, when performing quantitative analysis on non-uniform droplets, this embodiment does not need to consider the volume change during droplet heating, which greatly simplifies the analysis process and improves accuracy. This may be because during amplification, this embodiment sets the hot cover temperature of the qPCR instrument to 105°C, which is much higher than the 63°C required for droplet heating, which effectively inhibits the evaporation of the reaction solution, ensures the stability and independence of the droplets, and provides more accurate and reliable experimental data for this embodiment.

[0077] (2) Effect of refrigeration on precipitation

[0078] The stability of the precipitation is crucial for subsequent analysis. Therefore, the present embodiment designs the following experiment to explore the effect of storage temperature on precipitation. First, the present embodiment prepares the LAMP amplification reagent, divides it into two parts, and generates two batches of uniform droplets. The droplets collected in the two tubes are placed in a PCR instrument for amplification. After the amplification program is completed, one tube of droplets is stored in a 4°C refrigerator, and the other tube of droplets is stored at room temperature (25°C). After 30 minutes, the present embodiment draws droplets from the two tubes respectively and injects them into the observation chip for imaging. In order to quantify the amount of precipitation, the present embodiment uses ImageJ software to manually measure the area of ​​the precipitation area.

[0079] Figure 3 The precipitation at different storage temperatures is shown. The precipitation area is measured and the precipitation area is obtained when the temperature is stored at room temperature (such as Figure 3 as shown in part a of the figure), 4°C refrigeration conditions (as shown in part a of the figure), Figure 3 The precipitation imaging areas in the droplets (as shown in part b) were 1815.00±174.99 and 1715.88±149.32 pixels respectively. The two sets of data were statistically analyzed, and the p value was 0.19, that is, p>0.05, indicating that there was no significant difference between the two sets of data (as shown in part b). Figure 3 Therefore, it can be concluded from this example that the precipitation area of ​​the droplets did not change significantly before and after refrigeration, and the precipitation amount was not affected by the storage temperature, which confirmed the stability of the precipitation at different storage temperatures.

[0080] (3) Imaging test of precipitation

[0081] Since the precipitation yield inside the droplet is the basis for droplet classification in this embodiment during droplet typing. Therefore, it is crucial to ensure clear imaging of the precipitation, which is directly related to the accuracy and reliability of typing. First, in this embodiment, LAMP amplification reagents are prepared, and negative pressure is used to generate uniform droplets with a diameter of about 85 μm. After amplification, the droplets are placed on an ordinary glass slide and in an observation chip with a height of 46.5 μm for imaging observation. Through the comparison of imaging effects, this embodiment aims to obtain a droplet reading method with excellent droplet and precipitation imaging effects, ensure the clarity and imaging quality of the precipitation, and provide strong support for subsequent droplet typing work.

[0082] In this embodiment, the amplified droplets are placed on glass slides and imaged inside an observation chip at a height of 46.5 μm. Figure 4 As shown in part a of the figure, when a droplet is placed on a glass slide, it is difficult for the precipitate in the droplet to remain on the same horizontal focal plane, causing some areas of the precipitate to appear out of focus during imaging, and there are also differences in the precipitate focusing effects between different droplets. This is because it is difficult for a droplet to remain still on a glass slide, causing the precipitate to tumble continuously inside the droplet as the droplet moves. Some of the precipitate even tumbles to the side of the droplet, perpendicular to the imaging focal plane, making it impossible to fully present the morphological characteristics of the precipitate. In addition, fluorinated oil is volatile, and when the droplet is placed on a glass slide and in direct contact with the air, it evaporates within 5 minutes, making it difficult to image for a long time. Figure 4 As shown in part b of the figure, when the droplet is placed in an observation chip with a height of 46.5 μm, while the sediment remains stationary, a clearer sediment imaging effect is presented, so it is used as a device for reading droplets in this embodiment. On the one hand, the droplet can remain stationary in the chip, which is conducive to the stable imaging of the sediment on the horizontal plane and avoids the sediment from flipping; on the other hand, since the chip height is minimal, the sediment in the droplet is better confined to the same focal plane, ensuring the consistency of imaging between droplets.

[0083] Secondly, verify that the size of the precipitate generated in the droplets after amplification is related to the droplet size

[0084] In a non-uniform droplet system, the amount of precipitation generated is not only affected by the primer concentration, but also closely related to the size of each droplet, because droplets of different sizes will contain different amounts of amplification materials. In-depth research and clarification of the quantitative law or functional relationship between precipitation amount and droplet size will provide important theoretical basis and practical guidance for subsequent droplet judgment and analysis.

[0085] In order to systematically study the relationship between droplet size and precipitation yield, this example produced a batch of uniform droplets of different sizes and measured the amount of precipitation produced by the droplets. In the experiment, this example first prepared a LAMP reagent containing only target 1 and primer 1 as the aqueous phase, and BioRad droplet generation oil as the oil phase. A flow focusing droplet generation chip was used to successfully produce uniform droplets of different sizes by adjusting the flow rates of the aqueous phase and the oil phase. A small amount of droplets was taken on a glass slide to measure their true diameters, and the droplet diameters were found to be 27, 44, 55, 73, 77 and 89 μm, respectively. Subsequently, these droplets were subjected to LAMP amplification and injected into the observation chip for imaging, as shown in FIG. Figure 5 As shown in part (A) of the figure, it shows the precipitation in droplets of different sizes. The scale bar in the large figure is 100 μm, and the scale bar in the small figure is 20 μm.

[0086] In the experiment, this example observed that the precipitate not only expanded in the horizontal plane, but also accumulated significant thickness in the vertical direction. In order to more comprehensively evaluate the yield of precipitate, this example introduced the optical density value (OD) as a characteristic value, which is usually used to quantify the attenuation of light when passing through a substance. The optical density value calculation formula is as follows:

[0087]

[0088] Among them, pixel value indicates the grayscale value of the pixel at a certain position in the field image, that is, one pixel can get a light density value.

[0089] In this embodiment, the area of ​​the precipitation region was manually measured using ImageJ software. The imaging areas of the precipitation regions corresponding to the droplets with diameters of 27, 44, 55, 73, 77 and 89 μm from small to large were 67.57±9.41, 190.78±26.20, 333.49±36.53, 745.31±86.62, 876.41±63.31, 1456.90±113.61 μm 2 , and converted into optical density value by calculation formula for consideration. After analysis, it is found in this embodiment that the binomial fitting effect between droplet diameter and precipitation area is good, R 2 The value reaches 0.993, and the binomial fitting curve equation is y = 0.32x 2 -16.24x+174.94, where R 2 It is an indicator to measure the goodness of curve fitting. The closer to 1, the better. y is the precipitation area on the ordinate and x is the droplet diameter on the abscissa. Figure 5As shown in part (B) of FIG. 1 , it represents a quadratic fitting curve of the droplet diameter and the precipitation area value. This result shows that there is a quadratic mathematical relationship between the droplet area and the precipitation area, which provides an in-depth understanding of the relationship between the droplet size and the precipitation amount for this embodiment, and also provides an idea for the subsequent droplet classification for this embodiment.

[0090] By using the possible relationship between the amount of precipitation and the size of the droplets, this embodiment further explores the feasibility of the precipitation classification theory in the non-uniform droplet dual detection system. This embodiment measures the optical density of the precipitation in the three types of droplets, and also records the imaging area of ​​the droplets where the precipitation is located in the observation chip. Among the three types of droplets, this embodiment collects a series of optical density data of precipitation generated by droplets of different radii. In order to eliminate the deviation that may be caused by the measurement, this embodiment processes the data with an interval of 4μm, and takes the average value of the center of the interval as the radius. Since the primer concentration has been adjusted in advance, ideally, the amount of precipitation produced in the droplets containing only target 1 is the least, the droplets containing only target 2 produce a medium amount of precipitation, and the droplets containing both target 1 and target 2 produce the most precipitation.

[0091] After analysis, the data showed a good fit to a second-order polynomial.

[0092] In a specific embodiment, the determining the type of the corresponding droplet by combining the droplet characteristic curve of each type of droplet, the size of the droplet, and the optical density value of the precipitate includes:

[0093] A second-order polynomial is used in advance to fit the relationship between the optical density value of the precipitate in each type of droplet and the size of the droplet, so as to obtain a droplet characteristic curve of each type of droplet;

[0094] The size of the droplets extracted from the droplet image and the optical density value of the precipitate are substituted into the droplet characteristic curves of each type of droplets for matching, so as to determine the type of each droplet in the droplet image.

[0095] Specifically, the R values ​​of the quadratic fits of the droplet containing only target 1, the droplet containing only target 2, and the droplet containing both target 1 and target 2 are 2 The values ​​are 0.958, 0.981 and 0.972 respectively. Therefore, this embodiment selects a second-order polynomial as a mathematical model to describe the relationship between the OD of the sediment and the equivalent radius of the imaged droplet, which is called the droplet characteristic curve in this embodiment and will be used for subsequent droplet classification. It is worth noting that although this embodiment tried a third-order polynomial fitting, the effect was not ideal. This may be because when the droplets are imaged in the microfluidic observation chip, their morphology is mostly drum-shaped rather than an ideal sphere. Figure 6As shown in the figure, three fitting curves were obtained in the three types of droplets in this embodiment. Not only the fitting effect was good, but also there was a clear distinction between them. This verifies that the precipitation classification regulated by the primer still exists in the non-uniform droplet system. Among them, the blue curve represents the "droplet containing only target 1", and its fitting equation is y=0.002x 2 +0.106x–3.308, the red curve represents the “droplet containing only target 2”, and its fitting equation is y=0.003x 2 –0.139x–2.929, the black curve represents the “droplet containing both target 1 and target 2”, and its fitting equation is y=0.004x 2 +0.178x+5.44. In summary, the following conclusions can be drawn from this embodiment: In the three types of droplets, there is a highly fitted binomial relationship between the optical density value of the precipitate and the equivalent radius of the droplet. This result not only once again verifies the relationship between the amount of precipitation and the droplet size, but also verifies the feasibility of the concept of precipitation grading in a non-uniform system, laying the foundation for subsequent research in this embodiment. Therefore, this embodiment can extend the precipitation amount grading to non-uniform droplets and construct a multiple digital nucleic acid detection technology based on non-uniform droplets.

[0096] Secondly, the effect of different primer concentration conditions on the regulation of precipitation yield was explored

[0097] In order to construct a dual detection system with excellent discrimination, in the dual amplification system containing target 1 (buffalo-derived DNA) and target 2 (Proteus mirabilis DNA), this example adjusted the concentrations of target 1-specific primer 1 and target 2-specific primer 2, and set 7 sets of different primer concentration combinations to explore their regulatory effects on the precipitation yield.

[0098] In a specific embodiment, the concentration of each specific primer in the primer sample is determined as follows:

[0099] Presetting multiple sets of combinations of specific primers with different concentrations and adding them to the amplification reaction system respectively to obtain multiple sets of amplification reaction solutions that do not contain target nucleic acid;

[0100] Each group of amplification reaction solutions not containing target nucleic acid is divided into multiple portions, and then an excess of a single target nucleic acid and multiple target nucleic acids are added to each amplification reaction solution respectively;

[0101] Each group of multiple amplification reaction solutions is used as an aqueous phase to generate uniform droplets with known contents, and the droplets are collected, amplified, and imaged, and finally droplet images of multiple groups of uniform droplets under different primer concentration conditions are obtained;

[0102] Extracting the optical density value of the precipitate in the droplet image as a characteristic value, and drawing a characteristic value distribution histogram of various types of precipitates;

[0103] A threshold is set according to the characteristic value distribution histogram, and the droplets are classified using the threshold, and the accuracy of the droplets corresponding to each type is calculated in combination with the gold standard;

[0104] According to the accuracy of each type of droplet and the size of the precipitate, a primer concentration combination with high accuracy and large precipitate is selected.

[0105] In the experiment, this embodiment mixed 7 groups of primers 1 and primers 2 of different concentrations and added them to the LAMP reaction system to obtain an amplification reaction solution without a target. The amplification reaction solution was then divided into three parts, and an excess of primer 1-specific target 1 (corresponding to the Low category), primer 2-specific target 2 (corresponding to the Medium category), target 1 and target 2 (corresponding to the Large category) were added respectively. Then this embodiment divided the reagent into three parts, and the addition of primer 1-specific target 1 to one reagent generated a smaller amount of precipitation (corresponding to the "Low" category); the addition of primer 2-specific target 2 to one reagent was intended to generate a medium amount of precipitation (corresponding to the "Medium" category); and the addition of target 1 and target 2 to another reagent was intended to generate the largest amount of precipitation (corresponding to the "Large" category). In this embodiment, three amplification reaction solutions of each group are used as the aqueous phase to generate three uniform droplets with known contents, which are collected in different PCR tubes, respectively, and the droplets are subjected to LAMP amplification and imaging, and finally 7 groups of uniform droplets under different primer concentration conditions are obtained, such as Figure 7 shown.

[0106] Subsequently, the embodiment extracts the characteristic value of the precipitation region. In order to quantify the precipitation overlap effect of the droplets, the primer concentration with the best discrimination is screened out and applied to the non-uniform droplet system.

[0107] This example uses the following threshold classification method to evaluate the effect of the primer concentration combination:

[0108] (1) Select the optical density value of the precipitation after background removal as the feature and draw the characteristic value distribution histogram of the three types of precipitation.

[0109] (2) Figure 8 As shown in the figure, two thresholds are manually set according to the histogram distribution: threshold 1 (t1) and threshold (t2). t1 distinguishes the "Low" and "Medium" categories, and t2 distinguishes the "Medium" and "Large" categories. Therefore, the sediment less than t1 is the "Low" category, the sediment greater than t1 and less than t2 is the "Medium" category, and the sediment greater than t2 is the "Large" category.

[0110] (3) The droplets are classified according to the threshold value and combined with the gold standard to calculate the accuracy of each type of droplet.

[0111] (4) Based on the three types of accuracy, the primer concentration combination with better performance was selected to prepare the amplification reagent, and the non-uniform droplets containing only target 1, only target 2, and both target 1 and target 2 were vortexed to generate the droplets, which were then amplified, imaged, and analyzed to analyze the feasibility of implementing the concentration in the non-uniform droplet system.

[0112] According to the accuracy calculation method mentioned above, this example can obtain the estimated error rate under each primer concentration condition, as shown in Table 1. Taking all factors into consideration, this example believes that combination e shows the best accuracy, followed by combination b. Next, this example will evaluate the precipitation effect of the two groups in a non-uniform system. This example observed that the small precipitation produced by combination e is very small in the smaller droplets, which is not conducive to the positive and negative judgment of the droplet detection model. Fig. 9 As shown, the droplet detection model will misclassify positive droplets as negative, which will greatly affect the quantitative ability of the method. This example attempts to apply combination b to a non-uniform droplet system and finds that combination e does not appear in droplets with low precipitation. Therefore, this example subsequently uses combination b to construct a dual detection system.

[0113]

[0114] Table 1 Classification accuracy based on threshold under seven different primer concentration conditions

[0115] In one embodiment, extracting the size of the droplet and the optical density value of the precipitate in the droplet image by an image analysis algorithm includes:

[0116] Using a first deep learning model to detect and segment the droplets in the droplet image, and obtain the position and attribute of each droplet in the droplet image; wherein the attribute of the droplet includes positive and negative, a positive one represents that the droplet contains sediment, and a negative one represents that the droplet does not contain sediment;

[0117] Screening out positive droplets according to the attributes, and cutting out the screened positive droplets according to the positions to obtain the areas and / or volumes of the corresponding droplets;

[0118] Using a second deep learning model to detect and segment the sediment in the cut-out droplet to extract sediment area information; wherein the sediment area information includes the location and size of the sediment;

[0119] The precipitation feature is extracted according to the precipitation area information; wherein the precipitation feature at least includes the optical density value of the precipitation in the droplet.

[0120] In this embodiment, when performing image analysis on the droplet image to extract the size of the droplet and the optical density value of the sediment, the droplet image is first subjected to target detection and image segmentation through a deep learning model to obtain the position and properties of the droplets, i.e., positive droplets and negative droplets. The positive droplets are then cropped, and the sediment of the cropped droplets is further detected and segmented through a deep learning model to obtain sedimentation area information, and then sedimentation features including optical density values ​​are obtained based on the sedimentation area information.

[0121] In practical application scenarios, in order to achieve in-depth research on the amount of precipitation in non-uniform droplets, this embodiment first establishes a reliable image analysis process, such as Fig.10 The specific steps are as follows:

[0122] (1) First, this embodiment uses a deep learning model to accurately detect and segment non-uniform droplets, which not only effectively realizes the positive and negative classification of droplets, but also accurately measures the droplet size as a key parameter in subsequent quantification, such as Fig.10 As shown in part (a) of

[0123] (2) In order to deeply analyze the precipitation amount in the non-uniform droplets, this embodiment accurately locates each droplet according to the droplet detection results and performs detailed cropping processing on it, such as Fig.10 As shown in part (b) of ;

[0124] (3) Subsequently, this embodiment uses another deep learning model to detect and segment the precipitation in the droplet, such as Fig.10 As shown in part (c) of the figure;

[0125] (4) According to the specific requirements of this embodiment, this embodiment extracts the characteristic value of the precipitation area, such as Fig.10 As shown in part (d) of .

[0126] Secondly, a droplet detection model was built. Specifically, first, 381 droplet bright field images were collected in this example. In the labeling software Labelme, a polygon tool was used to segment the droplets along the edge, and droplets containing visible precipitation were marked as "positive", and droplets without precipitation were marked as "negative". Droplets that were not fully displayed at the edge of the image and those with a small droplet size were not marked. Fig.11 As shown, Fig.11 In the analysis, droplets containing precipitates are labeled as “positive” and droplets without precipitates are labeled as “negative”.

[0127] The images were randomly assigned in a ratio of 9:1 between the training set and the validation set. The training set contained 342 images and the validation set contained 39 images. A total of 12,785 droplets were included, of which 6,794 were positive droplets and 5,991 were droplets without precipitation. When the droplet diameter is too small, the algorithm needs to use a very small scanning frame to scan it, and the program's computational cost is too high. Discarding these droplets can greatly improve the program's efficiency. When the droplet diameter is too large, the image is irregular in shape and difficult to detect, and the algorithm has difficulty detecting such droplets.

[0128] Next, this embodiment uses the Mask R-CNN model to detect droplets, obtain the detection category and detection frame information segmentation and positive and negative classification. The algorithm structure is as follows: Fig.12 As shown in the figure. First, the image is processed by a feature extraction network (Feature Pyramid Network FPN based on ResNet 50) to generate a series of feature maps of different scales. These feature maps not only contain rich semantic information, but also effectively solve the problem of target scale change through multi-scale fusion strategy. Then, the region proposal network (RPN) scans these feature maps to generate a series of candidate regions of interest (RoI). RPN preliminarily screens the candidate RoIs through binary classification (foreground or background) and bounding box regression operations, and eliminates a large number of candidate regions that do not meet the requirements. Subsequently, the remaining RoIs are aligned, and the spatial misalignment problem caused by the quantization operation is avoided by adopting the bilinear interpolation algorithm, thereby improving the accuracy of feature extraction. Finally, these precisely aligned RoIs are sent to the fully connected layer and the convolutional layer for further classification, regression and mask generation. In the mask generation stage, the network performs a full convolutional network operation on each RoI, predicts the segmentation mask with pixel-level accuracy, and achieves accurate segmentation of the target instance.

[0129] According to the actual needs of this embodiment, this embodiment adjusts the following parameters:

[0130] (1) Setting of training parameters. In order to improve the model training speed, this embodiment reduces the original image size from 1608×1608 pixels to 768×768 pixels. In order to accurately segment droplets of different sizes, this embodiment sets the size of the RPN anchor frame of the scanned droplets to 32, 64, 128, 256 and 512 pixels, with a side length range of 4, 8, 16, 32 and 64 pixels, and an aspect ratio of 2, 1 and 0.5. According to the number of droplets in the image, this embodiment sets the number of RoIs in the image to 512. The non-maximum suppression threshold in RPN is set to 0.1, which effectively eliminates candidate frames with high overlap and reduces false subtraction.

[0131] (2) Stochastic gradient descent optimization algorithm: During the model training process, this embodiment uses the stochastic gradient descent optimization algorithm and sets a learning rate of 0.0005.

[0132] (3) Data enhancement: This embodiment uses the imgaug tool to enrich the training data through a series of complex image transformations, thereby enhancing the robustness of the model so that it can maintain stable performance when facing various complex scenarios and changes.

[0133] Next, build a precipitation detection model. Input the image containing droplets into the trained droplet detection model, and cut out the positive droplets according to the boundary coordinates and predicted category of each droplet in the output file (i.e., the output file of the droplet detection model, which contains the predicted category of the droplet and the coordinates of the droplet detection frame). Then, segment and annotate the precipitation in the positive droplets in the annotation software Labelme, outline the precipitation area with a polygon tool and assign a "precipitate" label, such as Fig.13 In this embodiment, 954 precipitation images are annotated, and the images are randomly allocated according to the ratio of 8:1 between the training set and the validation set. The training set contains 846 images and the validation set contains 108 images.

[0134] Considering the large number of droplet images and the demand for detection speed, this embodiment selects the YOLOv5 model to implement precipitation segmentation. YOLOv5 performs well in both accuracy and speed, and the model file size is small. With effective pre-trained weights and flexible deployment capabilities, it is possible to quickly start and run on various hardware platforms. The network architecture of YOLOv5 mainly consists of the following key parts: input layer, backbone network, neck network and head network. The input layer is responsible for preprocessing image data, and ensures that the model can stably extract features through size adjustment and normalization operations. The backbone network adopts the advanced CSP Darknet53 structure, and extracts deep features in the image through layer-by-layer convolution and pooling, providing strong support for subsequent detection. The neck network uses feature pyramid technologies such as FPN to fuse feature information at different levels and enhance the multi-scale detection capability of the model. Finally, the head network generates detection results based on the feature map of the neck network, and predicts bounding boxes, category probabilities and target confidences of different scales and aspect ratios by processing multiple detection layers in parallel.

[0135] In the training model, this embodiment uses the pre-trained yolov5l-seg.pt weight file to initialize the model parameters. The stochastic gradient descent algorithm is used for optimization, and the initial learning rate and the final learning rate are both 0.01. The intersection-over-union ratio threshold is set to 0.2, which helps to remove bounding boxes with high overlap. The batch size is set to 8, and the input image is normalized to 640. The total number of training rounds is set to 500 rounds. At the end of each round, the model is evaluated using the validation set and the current optimal model weights are saved. In addition, data enhancement techniques are introduced, including color space transformations, such as adjusting the hue, saturation, and brightness of the image; geometric transformations, such as translation, scaling, and left-right flipping. And Mosaic data enhancement technology is used to mix multiple training images into one image. These operations are randomly applied to the input images during the training process to generate a large number of new and diverse training samples, helping the model learn more robust feature representations to improve its performance in actual scenarios.

[0136] Through Python code, this embodiment automates the process of measuring the optical density of precipitation. The program first removes the background of the droplet image, and then extracts the optical density value of the precipitation based on the precipitation area information obtained by the precipitation segmentation algorithm as the key characteristic value to characterize the precipitation amount. In addition, this embodiment also extracts the precipitation area and the total gray value of the precipitation area as alternative characteristic values. These characteristic values ​​are saved together with their corresponding non-uniform droplet area.

[0137] In one embodiment, the precipitate contained in the interior of the amplified droplet is an amplification product of a single target nucleic acid or an amplification product of at least two target nucleic acids;

[0138] The counting of the proportion of each type of droplets to the total number of droplets, and calculating the concentration of at least two target nucleic acids in all droplets according to the proportion of each type of droplets to the total number of droplets, includes:

[0139] Counting the types of each droplet, and obtaining the ratio of each type of droplet to the total number of droplets and the size of each droplet; wherein the types of droplets include droplets without target nucleic acid, droplets containing only one target nucleic acid, and droplets containing at least two target nucleic acids;

[0140] Based on the proportion of each type of droplets in the total number of droplets and the size of each droplet, the concentration of at least two target nucleic acids is calculated using the maximum likelihood estimation method.

[0141] Specifically, the concentrations of at least two target nucleic acids are calculated using the maximum likelihood estimation method based on the proportion of each type of droplets to the total number of droplets and the size of each droplet, including:

[0142] The concentrations of the first target nucleic acid and the second target nucleic acid are calculated according to the following formula:

[0143]

[0144] Wherein, N0, N1, N2, and N3 are the number of droplets without target nucleic acid, droplets containing only the first target nucleic acid, droplets containing only the second target nucleic acid, and droplets containing both the first target nucleic acid and the second target nucleic acid, respectively; V 0,i 、V 1,i 、V 2,i 、V 3,i are the volumes of droplets without target nucleic acid, droplets containing only the first target nucleic acid, droplets containing only the second target nucleic acid, and droplets containing both the first target nucleic acid and the second target nucleic acid, respectively; i is a number; C1 and C2 are the concentrations of the first target nucleic acid and the second target nucleic acid, respectively.

[0145] Combination Figure 6 , this embodiment uses the three fitting curves as the characteristic curves of the subsequent droplet classification in this embodiment. For each positive droplet sample point to be detected, this embodiment extracts its droplet size and precipitation optical density value through image analysis algorithm, and calculates the point Figure 6 The distance between the three fitting curves in the y direction is calculated, and the curve with the closest distance is found, and the category of the curve is assigned to the sample point.

[0146] In droplet digital PCR (ddPCR), the distribution of target nucleic acids in droplets can be described by Poisson distribution. Poisson distribution assumes that the packaging of nucleic acid molecules in each droplet is random and the number of nucleic acid molecules contained in each droplet is an independent random variable. The formula is as follows:

[0147]

[0148] Where k is the number of target nucleic acid molecules contained in the droplet, λ = C·V i represents the average copy number of the target nucleic acid in the droplet, C is the nucleic acid concentration, V i is the volume of the droplet.

[0149] Based on the above formula, the probability that a droplet contains k (k = 0, 1, 2...) nucleic acid molecules is:

[0150]

[0151] The amplification reaction causes changes in the partitions containing one or more nucleic acid molecules, distinguishing these cells from empty cells. Based on the above formula, the probabilities of empty partitions (k = 0) and positive partitions (k>0) are:

[0152]

[0153] In a non-uniform reaction system, assuming that the number of negative partitions is M, the partition volume is The number of positive partitions is N, and its volume is Then we have:

[0154]

[0155] The left side of the equation is the total volume of all partitions in the system, and the right side of the equation is the total number of positive partitions divided by the probability of positive partitions, which is also the total volume of the partitions, so the equation holds. Substituting the volume of the partitions and the yin-yang state, solving the equation to obtain the best estimate of C, you can achieve quantification of nucleic acid samples.

[0156] Extended to the digital quantification of dual-target nucleic acids in heterogeneous systems, assuming that the number of empty droplets, droplets containing only target 1, droplets containing only target 2, and droplets containing both targets 1 and 2 observed in the experimental results are N0, N1, N2, and N3, respectively. The volume of each droplet in the four types of droplets is V 0,i 、V 1,i 、V 2,i 、V 3,i (i is a number). Using the maximum likelihood estimation method, the target 1 concentration C1 and the target 2 concentration C2 satisfy the following equation:

[0157]

[0158] By solving the above two equations separately, we can get the estimation of the dual target concentrations C1 and C2.

[0159] In one embodiment, generating non-uniform droplets of the amplification reagent comprises:

[0160] Vortexing is used to generate non-uniform droplets of the amplification reagent.

[0161] Specifically, the step of generating non-uniform droplets of the amplification reagent by vortex oscillation includes:

[0162] Inject droplet generation oil as the oil phase into the loading tube;

[0163] Adding the prepared amplification reagent into the loading tube as an aqueous phase;

[0164] The loading tube was placed on a vortex shaker for shaking and standing to generate droplets of different sizes.

[0165] Combination Fig.14 In this embodiment, when analyzing droplets through deep learning and constructing droplet characteristic curves to realize droplet typing, the dual digital droplet nucleic acid detection method specifically includes:

[0166] (1) Reagent preparation: Prepare a tube of detection reagent, including LAMP amplification reagent, nucleic acid sample and fluorinated oil containing surfactant (Biorad EvaGreenI).

[0167] (2) Droplet generation: Vortex oscillation is used to generate non-uniform water-in-oil droplets through the difference in shear force at the oil-water interface. The aqueous phase is the mixed reagent in (1), and the oil phase is Biorad EvaGreen I oil. The droplet inclusion conditions are divided into four categories: no nucleic acid, nucleic acid 1, nucleic acid 2, and nucleic acid 1+2, corresponding to empty droplets, droplets with a small amount of precipitation, droplets with a medium amount of precipitation, and droplets with a large amount of precipitation, respectively.

[0168] (3) Isothermal amplification: The collected water-in-oil droplets were transferred to a qPCR instrument, the constant temperature was set at 63°C, and amplification was performed for 45 minutes.

[0169] (4) Reading chip preparation: The droplet reading adopts the imaging method in the microchannel. In this embodiment, photolithography and soft etching technology are used to process a microchannel with a height of about 46.5 μm and a width of about 800 μm. The channel is designed with one inlet and one outlet.

[0170] (5) Image acquisition of droplets: First, soak the chip channel with oil phase, and then inject droplets from the chip inlet. If the droplets in the chip are too crowded, a small amount of oil phase can be injected, and the droplet image is captured using a 10x bright field microscope.

[0171] (6) Droplet classification and precipitation segmentation: After the droplet image is input into the algorithm, the system will classify the droplets based on the characteristic curve, identify empty droplets, droplets with small precipitation, droplets with medium precipitation, and droplets with large precipitation, and calculate the proportion of each type of droplet in the total number of droplets and the volume of each droplet.

[0172] (7) Concentration calculation: Based on the results output by the algorithm, substitute them into the formula to calculate the nucleic acid concentration.

[0173] Specifically, we first build a dual detection system with discrimination:

[0174] According to the formula in Table 2, a total amount of 10 μL of reaction reagents was prepared. In the verification experiment of this embodiment, a commercial kit was used for experimental operation. Target 1 and primer 1 correspond to buffalo DNA and its specific LAMP primer, and target 2 and primer 2 correspond to Proteus mirabilis DNA and its LAMP primer.

[0175]

[0176] Table 2 Preparation of LAMP reagent

[0177] Secondly, the processing and production of microfluidic chips: The manufacturing process of microfluidic chips involves precise photolithography and replication molding technology. First, design the chip structure and determine the parameters in AutoCAD software. Fig.15 As shown in part (a) of the diagram, the droplet generation chip includes an oil phase (continuous phase) inlet, a water phase (dispersed phase) inlet, and an outlet, and the chip height is 30 μm. Fig.15 As shown in part (b) of the figure, after drilling holes at both ends, droplets can be loaded to image the droplets. The width of the imaging channel neck is 800μm, the width of the middle of the imaging channel is 5mm, and the channel height is 46.5μm. In practical applications, the mold is processed by SU-8 photolithography technology. The SU-8 photolithography process includes processing film mask, coating, pre-baking, exposure, post-baking, development and curing.

[0178] Next, this embodiment uses polydimethylsiloxane (PDMS) soft lithography technology to replicate the pattern on the SU-8 mold. The specific processing operations of PDMS soft lithography are as follows:

[0179] (1) SU-8 mold surface silanization treatment: After treating the SU-8 mold with an oxygen plasma cleaner for 1 minute, the SU-8 mold was placed in a vacuum pump filled with silane gas for 10 hours to ensure that the SU-8 mold was fully in contact with the silane gas.

[0180] (2) Preparation of PDMS prepolymer: PDMS prepolymer and curing agent were mixed in a mass ratio of 10:1, stirred for about 2 minutes to mix evenly, and placed in a vacuum pump for about 30 minutes to preliminarily remove bubbles in the mixture.

[0181] (3) PDMS soft lithography. Use an air compressor gun to blow away dust and debris on the surface of the silicon mold, pour the PDMS prepolymer into the mold, ensure that the PDMS fully covers the mold surface and fits tightly with the pattern on the mold, and place it in the vacuum pump for 20 minutes again. After observing that the bubbles are completely removed, place the mold with the PDMS poured in a horizontal position in a 65°C oven and bake for 10 hours for curing.

[0182] (4) Demolding, cutting and punching of the PDMS chip. After taking out the solidified silicon mold, use a scalpel to peel off the PDMS printed with the pattern from the silicon mold. The pattern on the mold has been accurately copied to the PDMS block. Use a cutting machine to perform necessary cutting and trimming on the PDMS block to obtain a PDMS microfluidic chip with the desired pattern. Use a puncher to punch holes at the outlet and inlet. The oil phase inlet and water phase inlet of the droplet generation chip are punched with a puncher with a diameter of 1.2 mm, and the droplet collection port is punched with a puncher with a diameter of 0.8 mm. Both ends of the droplet observation chip are punched with a 1.2 mm puncher.

[0183] (5) Chip bonding. The bonding surface of the PDMS block and the glass slide was treated with oxygen plasma for 1 minute. After the bonding was completed, the two surfaces were bonded tightly and then placed in an oven at 110 degrees Celsius for 20 minutes.

[0184] (6) Hydrophobic treatment: The bonded PDMS chip is placed in a 65°C oven and baked for 1-2 days to achieve hydrophobic treatment of the internal flow channel of the PDMS chip.

[0185] Again, non-uniform droplet generation and imaging: In a 0.2 ml PCR eight-tube, first inject 20 μL of BioRad droplet generation oil as the oil phase, and then add the LAMP amplification reagent prepared in the previous step as the water phase. Place the eight-tube on a vortex oscillator and oscillate at 3200 rpm for 10 seconds, then let it stand for a few seconds to allow droplets to gradually form. In order to further break up and evenly distribute the droplets, oscillate again at the same speed for 5 seconds. After these two oscillation operations, a non-uniform droplet amplification system consisting of droplets of different volumes was generated.

[0186] After the droplet generation is completed, the PCR tube containing the droplets is placed in the PCR instrument for LAMP amplification reaction. The temperature of the amplification reaction is set to 63°C, the top cover temperature is set to 105°C, and the reaction duration is 45 minutes to ensure sufficient amplification reaction.

[0187] After amplification, the droplets are imaged in bright field using a specially designed observation chip. First, BioRad droplet generation oil is injected into the microchannel of the chip for full infiltration to reduce the adhesion and friction of the droplets in the channel. Subsequently, the droplets are injected into the microchannel through the inlet of the chip to ensure that the droplets are evenly distributed in the channel. If the droplets are too dense in the channel, a small amount of oil phase can be injected to assist the flow and dispersion of the droplets. As the droplets flow through the channel, a 10x bright field microscope is used to image the droplets and take bright field images of each droplet to ensure that there is sufficient image data support for subsequent analysis and processing.

[0188] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of this application.

[0189] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

Claims

1. A dual digital nucleic acid detection method based on non-uniform droplets, characterized in that: include: Adding a nucleic acid sample containing at least two target nucleic acids and a primer sample containing at least two corresponding specific primers into an amplification reaction system to obtain an amplification reagent; generating non-uniform droplets of the amplification reagent; Amplifying the droplets; wherein at least a portion of the amplified droplets contain precipitates; Injecting the amplified droplets into a microchannel and imaging the droplets in the microchannel using an imaging device to obtain a droplet image; The size of the droplets and the optical density value of the precipitate in the droplet image are extracted by an image analysis algorithm, and the type of the corresponding droplet is determined by combining the droplet characteristic curve of each type of droplet, the size of the droplet and the optical density value of the precipitate; wherein the droplet characteristic curve is a pre-fitted mathematical model that describes the relationship between the optical density value of the precipitate and the droplet size; The ratio of each type of droplets to the total number of droplets is counted, and the concentration of at least two target nucleic acids in all droplets is calculated according to the ratio of each type of droplets to the total number of droplets.

2. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The extracting the size of the droplet and the optical density value of the precipitate in the droplet image by using an image analysis algorithm comprises: Using a first deep learning model to detect and segment the droplets in the droplet image, and obtain the position and attribute of each droplet in the droplet image; wherein the attribute of the droplet includes positive and negative, a positive one represents that the droplet contains sediment, and a negative one represents that the droplet does not contain sediment; Screening out positive droplets according to the attributes, and cutting out the screened positive droplets according to the positions to obtain the areas and / or volumes of the corresponding droplets; Using a second deep learning model to detect and segment the sediment in the cut-out droplet to extract sediment area information; wherein the sediment area information includes the location and size of the sediment; The precipitation feature is extracted according to the precipitation area information; wherein the precipitation feature at least includes the optical density value of the precipitation in the droplet.

3. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The step of generating non-uniform droplets of the amplification reagent comprises: Vortexing is used to generate non-uniform droplets of the amplification reagent.

4. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 3, characterized in that: The step of generating non-uniform droplets of the amplification reagent by vortex oscillation comprises: Inject droplet generation oil as the oil phase into the loading tube; Adding the prepared amplification reagent into the loading tube as an aqueous phase; The loading tube was placed on a vortex shaker for shaking and standing to generate droplets of different sizes.

5. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The precipitate contained in the droplet after amplification is an amplification product of a single target nucleic acid or an amplification product of at least two target nucleic acids; The counting of the proportion of each type of droplets to the total number of droplets, and calculating the concentration of at least two target nucleic acids in all droplets according to the proportion of each type of droplets to the total number of droplets, includes: Counting the types of each droplet, and obtaining the ratio of each type of droplet to the total number of droplets and the size of each droplet; wherein the types of droplets include droplets without target nucleic acid, droplets containing only one target nucleic acid, and droplets containing at least two target nucleic acids; Based on the proportion of each type of droplets in the total number of droplets and the size of each droplet, the concentration of at least two target nucleic acids is calculated using the maximum likelihood estimation method.

6. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 5, characterized in that: The concentrations of at least two target nucleic acids are calculated using a maximum likelihood estimation method based on the proportion of each type of droplets to the total number of droplets and the size of each droplet, including: The concentrations of the first target nucleic acid and the second target nucleic acid are calculated according to the following formula: Wherein, N0, N1, N2, and N3 are the number of droplets without target nucleic acid, droplets containing only the first target nucleic acid, droplets containing only the second target nucleic acid, and droplets containing both the first target nucleic acid and the second target nucleic acid, respectively; V 0,i 、V 1,i 、V 2,i 、V 3,i are the volumes of droplets without target nucleic acid, droplets containing only the first target nucleic acid, droplets containing only the second target nucleic acid, and droplets containing both the first target nucleic acid and the second target nucleic acid, respectively; i is a number; C1 and C2 are the concentrations of the first target nucleic acid and the second target nucleic acid, respectively.

7. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The method of determining the type of corresponding droplets by combining the droplet characteristic curves of each type of droplets, the size of the droplets, and the optical density value of the precipitate includes: A second-order polynomial is used in advance to fit the relationship between the optical density value of the precipitate in each type of droplet and the size of the droplet, so as to obtain a droplet characteristic curve of each type of droplet; The size of the droplets extracted from the droplet image and the optical density value of the precipitate are substituted into the droplet characteristic curves of each type of droplets for matching, so as to determine the type of each droplet in the droplet image.

8. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The concentration of each specific primer in the primer sample is determined as follows: Presetting multiple sets of combinations of specific primers with different concentrations and adding them to the amplification reaction system respectively to obtain multiple sets of amplification reaction solutions that do not contain target nucleic acid; Each group of amplification reaction solutions not containing target nucleic acid is divided into multiple portions, and then an excess of a single target nucleic acid and multiple target nucleic acids are added to each amplification reaction solution respectively; Each group of multiple amplification reaction solutions is used as an aqueous phase to generate uniform droplets with known contents, and the droplets are collected, amplified, and imaged, and finally droplet images of multiple groups of uniform droplets under different primer concentration conditions are obtained; Extracting the optical density value of the precipitate in the droplet image as a characteristic value, and drawing a characteristic value distribution histogram of various types of precipitates; A threshold is set according to the characteristic value distribution histogram, and the droplets are classified using the threshold, and the accuracy of the droplets corresponding to each type is calculated in combination with the gold standard; According to the accuracy of each type of droplet and the size of the precipitate, a primer concentration combination with high accuracy and large precipitate is selected.

9. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The microchannel is a microchannel with a height of 46.5 μm and a predetermined floating value, and the microchannel has an inlet and an outlet.

10. The method for double digital nucleic acid detection based on non-uniform droplets according to claim 1, characterized in that: The top cover temperature during amplification is higher than the amplification temperature.

Images