CN115820816B - Multiple digital nucleic acid detection method, device and related medium based on deep learning - Google Patents
Multiple digital nucleic acid detection method, device and related medium based on deep learning Download PDFInfo
- Publication number
- CN115820816B CN115820816B CN202211516857.XA CN202211516857A CN115820816B CN 115820816 B CN115820816 B CN 115820816B CN 202211516857 A CN202211516857 A CN 202211516857A CN 115820816 B CN115820816 B CN 115820816B
- Authority
- CN
- China
- Prior art keywords
- reagent
- detection
- liquid drop
- sample
- droplet
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
The invention discloses a multiple digital nucleic acid detection method, a device and a related medium based on deep learning, wherein the method comprises the following steps: preparing multiple detection reagents, adding dyes with different colors for each detection reagent, and then coding; preparing sample reagents with the same quantity as the detection reagents, and uniformly mixing the sample reagents according to a preset target concentration gradient; generating a reagent droplet from the encoded detection reagent by a micro droplet generation device, and generating a sample droplet from the uniformly mixed sample reagent; pairing, fusing and amplifying the reagent liquid drops and the sample liquid drops, and generating liquid drop images; quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology; based on the quantitative analysis result, the detection result of the detection reagent is obtained through poisson distribution calculation. The invention can complete the aim of multiple digital nucleic acid detection under the condition of bright field, and avoids the dependence of digital nucleic acid detection on fluorescent signals.
Description
Technical Field
The invention relates to the technical field of nucleic acid detection, in particular to a multiple digital nucleic acid detection method and device based on deep learning and a related medium.
Background
Infectious diseases are diseases which are caused by various pathogens and can be transmitted mutually between people, have strong transmissibility, are easy to mutate, have great difficulty in prevention and control, and have extremely serious influence on the health and life of people. The nucleic acid detection can accurately detect whether genetic materials of pathogens exist in a patient, so that the infection type of the patient is judged, the patient infected by the pathogens with strong infectivity can be timely found, the patient can be effectively managed, and social transmission is prevented. In addition, nucleic acid detection plays an important role in applications such as diagnosis of serious diseases, food safety detection, and detection of water environment.
In biological samples, which are generally used for detection, the nucleic acid content is often very low, and it is necessary to increase the amount of target nucleic acid by a nucleic acid amplification technique. Among them, polymerase Chain Reaction (PCR) is the most common nucleic acid amplification technique. Based on this reaction, researchers have invented real-time fluorescent quantitative PCR, which is the currently mainstream method of nucleic acid detection, and has a certain quantitative capacity relative to PCR technology. However, the method is a relative quantitative technology which depends on a standard curve or a standard substance to determine the target nucleic acid amount, is time-consuming and labor-consuming in establishing the standard curve, is extremely easy to be interfered by human factors and environmental factors, and can cause certain waste of reagents and biological samples.
In recent years, digital PCR has emerged in human vision as a nucleic acid detection technique that can achieve absolute quantification without the need for a standard curve. With the development of microfluidic technology, the generation of digital droplet PCR (ddPCR) was promoted. ddPCR, however, requires absolute quantitative detection by fluorescence, requires stringent experimental conditions and expensive experimental equipment, and has limited available fluorescent channels, which makes multiplex digital nucleic acid detection difficult. Thus, how to achieve multiplex digital nucleic acid detection under brightfield conditions is a problem that needs to be addressed by those skilled in the art.
Disclosure of Invention
The embodiment of the invention provides a multiple digital nucleic acid detection method, a device, computer equipment and a storage medium based on deep learning, aiming at realizing multiple digital nucleic acid detection under the condition of bright field.
In a first aspect, an embodiment of the present invention provides a multiple digital nucleic acid detection method based on deep learning, including:
preparing multiple detection reagents, adding dyes with different colors for each detection reagent, and then coding;
preparing sample reagents with the same quantity as the detection reagents, and uniformly mixing the sample reagents according to a preset target concentration gradient;
generating a reagent droplet from the encoded detection reagent by a micro droplet generation device, and generating a sample droplet from the uniformly mixed sample reagent;
pairing, fusing and amplifying the reagent liquid drops and the sample liquid drops, and generating liquid drop images;
quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology;
based on quantitative analysis results, detecting results of the detection reagents are obtained through poisson distribution calculation, a multiple digital nucleic acid detection model is constructed, and the multiple digital nucleic acid detection model is utilized to detect the specified reagents.
In a second aspect, embodiments of the present invention provide a multiple digital nucleic acid detection apparatus based on deep learning, including:
the first configuration unit is used for configuring a plurality of detection reagents, and coding after adding dyes with different colors for each detection reagent;
the second configuration unit is used for configuring sample reagents with the same quantity as the detection reagents and uniformly mixing the sample reagents according to a preset target concentration gradient;
a droplet generation unit for generating a droplet of the encoded detection reagent by the micro droplet generation device and a sample droplet by the uniformly mixed sample reagent;
the image generation unit is used for carrying out pairing fusion and amplification on the reagent liquid drops and the sample liquid drops and generating liquid drop images;
the quantitative analysis unit is used for quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology;
and the model detection unit is used for obtaining the detection result of the detection reagent through poisson distribution calculation based on the quantitative analysis result, constructing a multiple digital nucleic acid detection model, and detecting the specified reagent by using the multiple digital nucleic acid detection model.
In a third aspect, an embodiment of the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the deep learning-based multiple digital nucleic acid detection method according to the first aspect when the computer program is executed.
In a fourth aspect, embodiments of the present invention provide a computer readable storage medium having a computer program stored thereon, which when executed by a processor implements the deep learning-based multiple digital nucleic acid detection method according to the first aspect.
The embodiment of the invention provides a multiple digital nucleic acid detection method, a device, computer equipment and a storage medium based on deep learning, wherein the method comprises the following steps: preparing multiple detection reagents, adding dyes with different colors for each detection reagent, and then coding; preparing sample reagents with the same quantity as the detection reagents, and uniformly mixing the sample reagents according to a preset target concentration gradient; generating a reagent droplet from the encoded detection reagent by a micro droplet generation device, and generating a sample droplet from the uniformly mixed sample reagent; pairing, fusing and amplifying the reagent liquid drops and the sample liquid drops, and generating liquid drop images; quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology; based on quantitative analysis results, detecting results of the detection reagents are obtained through poisson distribution calculation, a multiple digital nucleic acid detection model is constructed, and the multiple digital nucleic acid detection model is utilized to detect the specified reagents. According to the embodiment of the invention, the color dye is added into the detection reagent, and the detection reagent is quantitatively analyzed by combining with the deep learning image recognition technology, so that the aim of completing multiple digital nucleic acid detection under the condition of bright field is fulfilled, the dependence of digital nucleic acid detection on fluorescent signals is solved, the detection cost is greatly saved, and the nucleic acid detection efficiency is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 2 is a schematic block diagram of a multiple digital nucleic acid detection device based on deep learning according to an embodiment of the present invention;
FIG. 3 is a schematic diagram showing the color dye concentration in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 4 is a schematic diagram showing excessive color dye concentration in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of dye added in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a droplet generator in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of droplet pairing, fusion and amplification in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 8 is a schematic diagram of a droplet marker in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 9 is a diagram showing the effect of weight training in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
FIG. 10 is a schematic diagram showing droplet classification in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention
FIG. 11 is a diagram showing the verification effect in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention;
fig. 12 is a network structure diagram of YOLOv5 network model in a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the terms "comprises" and "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Referring to fig. 1, fig. 1 is a schematic flow chart of a multiple digital nucleic acid detection method based on deep learning according to an embodiment of the present invention, which specifically includes: steps S101 to S106.
S101, preparing a plurality of detection reagents, and adding dyes with different colors for each detection reagent to code;
s102, preparing sample reagents with the same quantity as the detection reagents, and uniformly mixing the sample reagents according to a preset target concentration gradient;
s103, generating a reagent droplet from the coded detection reagent by a micro droplet generation device, and generating a sample droplet from the uniformly mixed sample reagent;
s104, carrying out pairing fusion and amplification on the reagent liquid drops and the sample liquid drops, and generating liquid drop images;
s105, quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology;
s106, based on quantitative analysis results, obtaining detection results of the detection reagents through poisson distribution calculation, constructing a multiple digital nucleic acid detection model, and detecting the specified reagents by using the multiple digital nucleic acid detection model.
In this embodiment, on one hand, detection reagents are configured by combining dyes of different colors and are generated into reagent droplets, on the other hand, sample reagents of the same number are configured and are uniformly mixed to generate sample droplets, then the reagent droplets and the sample droplets are sequentially paired, fused and amplified, then quantitative analysis is performed on amplified droplet images based on a deep learning image recognition technology, and corresponding detection results are calculated by combining poisson distribution. After the above process, a multiple digital nucleic acid detection model for multiple digital nucleic acid detection can be established, and the multiple digital nucleic acid detection model can be used for rapidly and accurately detecting the specified reagent in the open field.
According to the embodiment, the color dye is added into the detection reagent, and the detection reagent is quantitatively analyzed by combining the deep learning image recognition technology, so that the aim of completing multiple digital nucleic acid detection under the condition of bright field is fulfilled, the dependence of digital nucleic acid detection on fluorescent signals is solved, the detection cost is greatly saved, and the efficiency of nucleic acid detection is improved.
In the prior art, a fluorescent quantitative PCR technology is adopted for common nucleic acid detection, three basic reactions of denaturation, annealing and extension are required to be carried out in a PCR reaction system, and the required reaction time is about 2 hours. The embodiment is based on loop-mediated isothermal amplification (LAMP), and is characterized in that 4 specific primers are designed for 6 regions of a target gene, and under the action of a strand displacement DNA polymerase, the amplification is carried out at the constant temperature of 60-65 ℃ for less than 1 hour, so that hundreds of millions of times of nucleic acid amplification can be realized, and the method has the characteristics of simplicity in operation, strong specificity, easiness in detection of products and the like. On the other hand, in the LAMP reaction, pyrophosphate ions precipitated from dNTPs of a deoxyribonucleic acid triphosphate substrate react with magnesium ions in a reaction solution during DNA synthesis, a large amount of magnesium pyrophosphate precipitates are generated, and the magnesium pyrophosphate precipitates can be used as a visual index of the reaction, that is, whether nucleic acid amplification exists or not can be identified by visually observing whether turbid precipitates exist or not, so that fluorescent observation is not needed. In addition, the embodiment also enables the liquid drop to generate color under the bright field by means of the color dye, and the color information of the liquid drop corresponds to the target gene information contained in the liquid drop. For example, the shigella-containing liquid drops are enabled to correspond to red, so that the number of positive liquid drops and negative liquid drops containing shigella can be respectively counted by combining whether the liquid drops contain magnesium pyrophosphate sediment or not in a bright field, and then the copy number of the liquid drops is calculated through poisson distribution, so that the digital nucleic acid detection can be realized. And by analogy, the digital nucleic acid detection of different pathogens can be realized by using different color dyes to code the liquid drops, so as to achieve the aim of multiple digital nucleic acid detection. However, using poisson distribution to calculate the copy number of a target droplet requires that the number of droplets be large enough, which is clearly not feasible if manual counting is used. Therefore, the method of deep learning image recognition can judge whether a large number of liquid drops have sediments in a very short time, then judge the color information of the liquid drops to further obtain which target genes the liquid drops contain, finally count the number of positive liquid drops and negative liquid drops with different colors, and calculate the copy number of target molecules by poisson distribution, so as to achieve the effect of digital quantification.
In general, the embodiment of the invention innovatively proposes that a precipitation product of isothermal amplification is taken as a marker, bright field imaging is used for replacing fluorescence imaging, whether precipitation in liquid drops exists or not is recognized through deep learning, and the yin-yang of each target in each liquid drop is judged, so that absolute quantification of multiple targets is realized on the premise of no fluorescence mark. Meanwhile, the present embodiment also has the following benefits: (1) The bright field imaging is used for replacing fluorescent imaging, so that the cost of liquid drop reading in liquid drop digital nucleic acid detection is greatly reduced. (2) The precipitation identification of the liquid drops is performed by deep learning, the liquid drops are classified again by combining the color dyes, and multi-target detection is realized on the premise of no hardware addition.
In one embodiment, the step S101 includes:
preparing a plurality of detection reagents by adopting specific primers and DNA amplification materials; wherein the DNA amplification material comprises DNA polymerase and dNTP material;
each detection reagent is added with a different color dye and coded in sequence.
In this embodiment, when the detection reagents are configured, each detection reagent contains a specific primer, and materials such as DNA polymerase and dNTP required for DNA amplification, and different color dyes are added respectively. Further, when preparing the sample reagent, the DNA containing the pathogen is adopted for preparation, and then the multitube sample reagent is uniformly mixed according to a preset target concentration gradient.
In a specific embodiment, red food dye, blue food dye and Orange G yellow dye are adopted, and by exploring the influence of the color dye at different concentrations on the subsequent amplification step, part of liquid drops shown in fig. 3 (a) are obtained, wherein the part of liquid drops are liquid drops made by oscillating at different concentrations of the red food dye, and it can be found that precipitation does not aggregate when the dye concentration is high, and precipitation is more obvious when the dye concentration is low. The effect of the color dye on the subsequent amplification step was also investigated, as shown in FIG. 3 at parts (b) and (c). Finally, 0.5% of red food dye, 0.5% of Orange X yellow dye and 0.3% of blue food dye are selected and no dye is added for carrying out multiple digital nucleic acid detection experiments.
Referring to fig. 3, it can be seen that adjusting the dye concentration to have little effect on the subsequent amplification step. Since too high a dye concentration would not allow precipitation to be observed, as shown in FIG. 4, too low a concentration would not allow the droplet color to be distinguished. Because the color of the liquid drops is in one-to-one correspondence with the detection reagent before the liquid drops are generated, for example, the shigella corresponds to the red liquid drops, the concentration of the shigella can be quantified through the number of the positive and negative red liquid drops, and in the same way, multiple digital nucleic acid detection can be realized by combining other pathogens corresponding to other colors, as shown in fig. 5.
In one embodiment, the step S102 includes:
generating the reagent droplets with a diameter of 85-95 μm from the encoded detection reagent by using an amplification reagent generating device;
the sample droplet having a diameter of 45 μm to 55 μm is produced by the sample package producing device by uniformly mixing the sample reagents.
In this embodiment, the micro-droplet generating device may include an experiment device such as an injection pump, an inverted optical microscope, and a PDMS microfluidic chip, and by using the experiment device, micro-droplets with equal volumes may be generated and collected. Wherein the color dye encodes a reagent having a droplet diameter of 85 μm to 95 μm, for example 90 μm, and the sample reagent has a droplet diameter of 45 μm to 55 μm, for example 50 μm.
In a specific embodiment, three microfluidic devices are employed, specifically including two droplet generation devices for LAMP reagent and sample encapsulation, respectively, and one microwell array device for droplet pairing. Wherein, the microfluidic chip for droplet generation is prepared by adopting SU-8 photoetching technology and a standard process of Polydimethylsiloxane (PDMS) reverse mould. Then, the SU-8 mold was subjected to oxygen plasma treatment and treated in a vacuum chamber filled with silane vapor for 10 hours. 10:1, are mixed, degassed, poured onto SU-8 molds and then baked at 60 ℃ for 10 hours. The cured PDMS was then peeled off, cut into the desired shape, and then perforated at the gate. The PDMS chips and glass slides were treated with oxygen plasma for 1 minute, placed in contact, and briefly baked at 110 ℃ for adhesion. These devices were then baked at 60 ℃ for 24 hours to render the channel surface hydrophobic.
Further, in consideration of the situation that liquid drops evaporate and bubbles are generated during amplification by using a traditional full PDMS chip, the processing technology of the chip is optimized according to the mass transfer principle. The micropore array device is formed by manufacturing a glass-PDMS micropore chip of a top layer and a middle layer, and mainly comprises three parts: 1. manufacturing a micropore array glass substrate; 2. manufacturing a micropore array mold; 3. and (3) manufacturing a flowing layer. Specifically, the slide was drilled and sealed with PDMS, and the cured PDMS was drilled with a punch as an access port. And secondly, the mould of the micropore array chip is obtained by overturning a mould of SU-8 mould through PDMS, and the thickness of the photoresist is 50 micrometers for the first layer and 20 micrometers for the second layer. Then, the SU-8 mold was subjected to oxygen plasma treatment and treated in a vacuum chamber filled with silane vapor for 10 hours. 10:1, mixing and degassing the PDMS prepolymer, pouring the mixture on an SU-8 mould, and stripping the mixture to obtain the PDMS micropore array mould. And (3) silanizing the obtained mold, pouring PDMS, attaching the punched slide to the PDMS, degassing, heating for curing, and stripping to obtain the PDMS micropore array. The macropores have a diameter of 90 microns and a depth of 70 microns, and the micropores have a diameter of 50 microns and a depth of 50 microns. And thirdly, pressing two layers of double faced adhesive tape and two layers of sealing films together (300-400 mu m thick), and cutting into a runner shape by using a laser cutting machine. And finally, fixing the glass-PDMS microporous chip, the runner layer and a new glass slide by using a clamp, wherein the upper layer of the clamp is made of acrylic material, the upper layer of the clamp is made of aluminum alloy by laser cutting, and the lower layer of the clamp is made of CNC (computerized numerical control) by using CNC (computer numerical control) processing, and the manufacturing schematic diagram is shown in figure 6.
In one embodiment, the step S103 includes:
loading the reagent liquid drop and the sample liquid drop into a PDMS micro-pore array chip for pairing;
fusing the reagent droplets and the sample droplets with different volumes into target droplets by using a corona machine;
amplifying the target liquid drops in a water bath amplification mode; wherein the water bath amplification temperature is 60-65 ℃ and the amplification time is 45-60 minutes;
and imaging the amplified target liquid drop to obtain the liquid drop image.
In this example, as shown in fig. 7, the reagent droplet and the sample droplet are sequentially subjected to droplet capturing, pairing, and fusion. Specifically, reagent droplets and sample droplets are sequentially loaded into a PDMS microporous array chip, so that the two droplets are paired, then, the droplets with two different volumes are fused into target droplets by using a corona machine for treatment, and then, the target droplets are subjected to water bath amplification, wherein in the amplification process, the temperature is controlled to be 60-65 ℃, and the time is controlled to be 45-60 minutes. After amplification was completed, the droplets in the chip were imaged with a microscope with a camera, and the total number of droplets was 15000.
In one embodiment, the step S104 includes:
identifying whether a target liquid drop corresponding to the liquid drop image has sediment or not in a bright field by utilizing a deep learning image identification technology, and identifying the liquid drop color of the target liquid drop;
judging whether the target liquid drop is a positive liquid drop according to the existence of sediment;
classifying the target liquid drops according to the liquid drop colors of the target liquid drops, and judging the negative probability of the target liquid drops according to the classified target liquid drop numbers.
In this embodiment, a deep learning method is used to identify whether there is a turbid precipitate in the droplets, that is, determine whether the droplets are positive droplets (droplets with precipitate) or negative droplets (droplets without precipitate), and then count the number of 8 droplets by determining the color of the droplets (2x4=8). It should be noted that, in this embodiment, the deep learning method is not adopted to directly identify 8 droplets, and this is because the difficulty of training weights is great, and compared with the method of identifying only whether there is a precipitate, a large amount of data is required, and the training effect is not necessarily ideal. Specifically, (1) the training set, the verification set and the training weight are marked, for example, 30 pictures are selected from the collected pictures, the pictures are randomly distributed according to the ratio of the training set to the verification set of 2:1, namely, 20 pictures are obtained from the training set, 10 pictures are obtained from the verification set, and then the marking drops are divided into yin and yang types by using an ImageLabel (picture area marking plug-in), as shown in fig. 8. The marked data is input into a deep learning network to train weights, and weight files with ideal effects can be obtained by training about 400 epochs. (2) Image recognition and drop number statistics based on model and training weights: and carrying out liquid drop identification on the collected pictures by using the trained weight file and the YOLOv5 network model, wherein the set conf-thres value is 0.75, and the set iou-thres value is 0.25, and the effect is shown in fig. 9. Because many pictures are detected by rewriting the YOLOv5 code, two txt files are created each time a picture is detected, wherein one txt file stores the position information of each positive liquid drop, and the other txt file stores the position information of each negative liquid drop; it will also limit the conditions of the detection frame, such as the aspect ratio of 0.8 to 1.2, to obtain the position information of the detection frame, since the droplet is substantially circular, the detection frame should be square-like. The method comprises the steps of detecting liquid drops through a YOLOv5 network model to obtain coordinate information, identifying the colors of the liquid drops, classifying and counting, so that the coordinate information of positive liquid drops and the coordinate information of negative liquid drops can be obtained respectively, judging the colors of the liquid drops on a liquid drop image through the coordinate information, classifying the liquid drops into 8 types, and counting the number and the negative probability of the liquid drops, as shown in fig. 10.
Further, the identifying whether the target droplet corresponding to the droplet image has a deposit or not and identifying the droplet color of the target droplet by using a deep learning image identification technology under a bright field includes:
inputting the liquid drop image into a YOLOv5 network model, and sequentially performing mosaic data enhancement, self-adaptive picture scaling and self-adaptive anchor frame calculation on the liquid drop image by utilizing an input end of the YOLOv5 network model;
reducing information loss and learning characteristics of the liquid drop image through a Focus structure and a CSP structure in a back bone end of the Yolov5 network model;
and predicting a feature map for the droplet image through a FPN structure and a PAN structure in a Neck end of the YOLOv5 network model, and judging whether sediment and droplet color exist or not based on the feature map.
The YOLOv5 network used for deep learning image recognition in this embodiment includes an input terminal, a backup terminal, a lock terminal, and a Head terminal, as shown in fig. 12. The input end comprises three parts of mosaic data enhancement, adaptive picture scaling and adaptive anchor frame calculation. Mosaic data enhancement: the 4 images are randomly cut and scaled, and then the images are randomly arranged and spliced to form one image for training, so that the data required by training is increased, meanwhile, the target data of a small sample is increased, and the robustness of the model is improved. Adaptive picture scaling: the size of the input image is adaptively scaled to 640 x 640 pixels and the amount of black edges filled is guaranteed to be minimal, reducing information redundancy. Self-adaptive anchor frame calculation: and comparing the output prediction frame with the groudtluth of the real frame, calculating the difference between the output prediction frame and the groudtluth of the real frame, reversely updating the prediction frame, and calculating the most suitable anchor frame through multiple iterations.
The backhaul includes a Focus structure and a CSP structure. Focus structure: the large-size picture is split into a plurality of small-size pictures by adopting slicing operation, and different features are extracted in a convolution mode, so that information loss caused by downsampling can be reduced. CSP structure: the input image information is divided into two parts, convolution operation is carried out to halve the number of channels, and then the two parts are connected to enable the model to learn more features.
The Neck end adopts an FPN+PAN structure, the FPN structure transmits and fuses high-level characteristic information from top to bottom in an up-sampling mode to obtain a predicted characteristic diagram, the PAN structure generates a characteristic pyramid, strong positioning characteristics are transmitted from bottom to top, and the detection of different-size targets by a model is enhanced; the Head end adopts CIOU_Loss as a Loss function of the target frame, and the nms non-maximum suppression operation is used when screening a large number of target frames, so that the inaccurate target frames are reduced.
In one embodiment, the step S105 includes:
calculating a positive command for the target droplet based on a negative probability of the target droplet according to:
where λ represents the average copy number of the DNA template molecule in each reaction unit, and k represents the number of the DNA template molecules in the reaction unit.
In this embodiment, the positive finger concentration is calculated from poisson distribution and compared with a preset concentration to verify reliability, as shown in fig. 11.
In the poisson distribution calculation formula, λ is the average copy number of the DNA template molecule in each reaction unit, and p is the probability that k copies of the DNA template molecule are contained in the reaction unit. Lambda corresponds to the initial copy number (c) of the DNA template molecule in the sample diluted v-fold (v is the dilution, corresponds to the volume of the independent reaction unit), lambda = cv; when k=0, i.e. in the absence of DNA template molecules in the reaction unit, the above is the caseThe poisson distribution calculation formula can be simplified as: p (x=0) =e -λ C is the desired value and v is calculated by the sphere volume, e.g. the droplet diameter of the sample droplet produced is 50um, from which the volume of the sample droplet (0.065 nl) is calculated. Because p (x=0) =e -λ＝-cv Since negative probability is calculated, the negative probability can be calculated from the negative and positive properties of the droplets, and thus c= -ln (P Yin type vagina ) V. Preferably, c (detected value) and c 0 The feasibility and reliability of the present embodiment can be verified by comparing (preset values).
Fig. 2 is a schematic block diagram of a multiple digital nucleic acid detection apparatus 200 based on deep learning according to an embodiment of the present invention, where the apparatus 200 includes:
a first configuration unit 201, configured to configure multiple detection reagents, and encode after adding different color dyes for each detection reagent;
a second configuration unit 202, configured to configure the same number of sample reagents as the detection reagents, and uniformly mix the sample reagents according to a preset target concentration gradient;
a droplet generation unit 203 for generating a droplet of the encoded detection reagent by the micro droplet generation device and a sample droplet by the uniformly mixed sample reagent;
an image generating unit 204, configured to pair, fuse and amplify the reagent droplet and the sample droplet, and generate a droplet image;
a quantitative analysis unit 205 for quantitatively analyzing the droplet image in the bright field by using a deep learning image recognition technique;
and a model detection unit 206 for obtaining the detection result of the detection reagent by poisson distribution calculation based on the quantitative analysis result, thereby constructing a multiple digital nucleic acid detection model, and detecting the specified reagent by using the multiple digital nucleic acid detection model.
In an embodiment, the first configuration unit 201 includes:
a material adding unit for preparing a plurality of tubes of the detection reagent by using a specific primer and a DNA amplification material; wherein the DNA amplification material comprises DNA polymerase and dNTP material;
and the dye adding unit is used for adding different color dyes to each detection reagent and coding the detection reagents in sequence.
In an embodiment, the droplet generation unit 203 includes:
a first generation unit for generating the reagent droplets having a diameter of 85 μm to 95 μm from the encoded detection reagent by an amplification reagent generation device;
and a second generation unit for generating the sample droplets with the diameter of 45-55 μm by using the sample package generation device.
In an embodiment, the image generating unit 204 includes:
the pairing unit is used for loading the reagent liquid drops and the sample liquid drops into the PDMS microporous array chip for pairing;
the fusion unit is used for fusing the reagent liquid drops with different volumes and the sample liquid drops into target liquid drops by using a corona machine;
the amplification unit is used for amplifying the target liquid drops in a water bath amplification mode; wherein the water bath amplification temperature is 60-65 ℃ and the amplification time is 45-60 minutes;
and the imaging unit is used for carrying out imaging processing on the amplified target liquid drops to obtain liquid drop images.
In one embodiment, the quantitative analysis unit 205 includes:
an image recognition unit for recognizing whether a deposit exists in a target droplet corresponding to the droplet image in a bright field by using a deep learning image recognition technology, and recognizing a droplet color of the target droplet;
a positive judging unit for judging whether the target liquid drop is a positive liquid drop according to the existence of sediment;
the probability judging unit is used for classifying the target liquid drops according to the liquid drop colors of the target liquid drops and judging the negative probability of the target liquid drops by combining the classified target liquid drop numbers.
In one embodiment, the modeled detection unit 206 includes:
a positive command unit for calculating a positive command of the target droplet based on a negative probability of the target droplet according to the following formula:
where λ represents the average copy number of the DNA template molecule in each reaction unit, and k represents the number of the DNA template molecules in the reaction unit.
In an embodiment, the image recognition unit includes:
the image input unit is used for inputting the liquid drop image into a YOLOv5 network model, and sequentially carrying out mosaic data enhancement, self-adaptive picture scaling and self-adaptive anchor frame calculation on the liquid drop image by utilizing an input end of the YOLOv5 network model;
a feature learning unit, configured to reduce information loss and learn features of the droplet image through a Focus structure and a CSP structure in a backhaul end of the YOLOv5 network model;
and the characteristic diagram prediction unit is used for predicting a characteristic diagram of the liquid drop image through the FPN structure and the PAN structure in the Neck end of the Yolov5 network model, and judging whether sediment and liquid drop color exist or not based on the characteristic diagram.
Since the embodiments of the apparatus portion and the embodiments of the method portion correspond to each other, the embodiments of the apparatus portion are referred to the description of the embodiments of the method portion, and are not repeated herein.
The embodiment of the present invention also provides a computer readable storage medium having a computer program stored thereon, which when executed can implement the steps provided in the above embodiment. The storage medium may include: various media capable of storing program codes, such as a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.
The embodiment of the invention also provides a computer device, which can comprise a memory and a processor, wherein the memory stores a computer program, and the processor can realize the steps provided by the embodiment when calling the computer program in the memory. Of course, the computer device may also include various network interfaces, power supplies, and the like.
In the description, each embodiment is described in a progressive manner, and each embodiment is mainly described by the differences from other embodiments, so that the same similar parts among the embodiments are mutually referred. 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 points refer to the description of the method section. It should be noted that it would be obvious to those skilled in the art that various improvements and modifications can be made to the present application without departing from the principles of the present application, and such improvements and modifications fall within the scope of the claims of the present application.
It should also be noted that in this specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Claims (7)
1. A multiple digital nucleic acid detection method based on deep learning, comprising:
preparing multiple detection reagents, adding dyes with different colors for each detection reagent, and then coding;
preparing sample reagents with the same quantity as the detection reagents, and uniformly mixing the sample reagents according to a preset target concentration gradient;
generating a reagent droplet from the encoded detection reagent by a micro droplet generation device, and generating a sample droplet from the uniformly mixed sample reagent;
pairing, fusing and amplifying the reagent liquid drops and the sample liquid drops, and generating liquid drop images;
quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology;
based on quantitative analysis results, obtaining detection results of the detection reagents through poisson distribution calculation, constructing a multiple digital nucleic acid detection model, and detecting the designated reagents by using the multiple digital nucleic acid detection model;
the pair fusion and amplification of the reagent droplets and the sample droplets and the generation of droplet images comprise:
loading the reagent liquid drop and the sample liquid drop into a PDMS micro-pore array chip for pairing;
fusing the reagent droplets and the sample droplets with different volumes into target droplets by using a corona machine;
amplifying the target liquid drops in a water bath amplification mode; wherein the water bath amplification temperature is 60-65 ℃ and the amplification time is 45-60 minutes;
imaging the amplified target liquid drop to obtain a liquid drop image;
the quantitative analysis of the drop image in the bright field by using the deep learning image recognition technology comprises the following steps:
identifying whether a target liquid drop corresponding to the liquid drop image has sediment or not in a bright field by utilizing a deep learning image identification technology, and identifying the liquid drop color of the target liquid drop;
judging whether the target liquid drop is a positive liquid drop according to the existence of sediment;
classifying the target liquid drops according to the liquid drop colors of the target liquid drops, and judging the negative probability of the target liquid drops by combining the classified target liquid drop numbers;
the identifying whether the target liquid drop corresponding to the liquid drop image has sediment or not in the bright field by utilizing the deep learning image identification technology, and identifying the liquid drop color of the target liquid drop, comprises the following steps:
inputting the liquid drop image into a YOLOv5 network model, and sequentially performing mosaic data enhancement, self-adaptive picture scaling and self-adaptive anchor frame calculation on the liquid drop image by utilizing an input end of the YOLOv5 network model;
reducing information loss and learning characteristics of the liquid drop image through a Focus structure and a CSP structure in a back bone end of the Yolov5 network model;
and predicting a feature map for the droplet image through the FPN structure and the PAN structure in the Neck end of the Yolov5 network model, and judging whether sediment and droplet color exist or not based on the feature map.
2. The deep learning based multiplex digital nucleic acid detection method according to claim 1, wherein the configuring of a plurality of detection reagents and the encoding after adding different color dyes for each detection reagent comprises:
preparing a plurality of detection reagents by adopting specific primers and DNA amplification materials; wherein the DNA amplification material comprises DNA polymerase and dNTP material;
each detection reagent is added with a different color dye and coded in sequence.
3. The deep learning based multiplex digital nucleic acid detection method of claim 1, wherein the generating of the encoded detection reagent into the reagent droplets by the micro-droplet generation device and the generation of the sample droplets from the uniformly mixed sample reagents comprises:
generating the reagent droplets with a diameter of 85-95 μm from the encoded detection reagent by using an amplification reagent generating device;
the sample droplet having a diameter of 45 μm to 55 μm is produced by the sample package producing device by uniformly mixing the sample reagents.
4. The deep learning-based multiplex digital nucleic acid detection method according to claim 1, wherein the detection result of the detection reagent is obtained by poisson distribution calculation based on the quantitative analysis result, comprising:
calculating a positive command for the target droplet based on a negative probability of the target droplet according to:
where λ represents the average copy number of the DNA template molecule in each reaction unit, and k represents the number of the DNA template molecules in the reaction unit.
5. A multiple digital nucleic acid detection device based on deep learning, comprising:
the first configuration unit is used for configuring a plurality of detection reagents, and coding after adding dyes with different colors for each detection reagent;
the second configuration unit is used for configuring sample reagents with the same quantity as the detection reagents and uniformly mixing the sample reagents according to a preset target concentration gradient;
a droplet generation unit for generating a droplet of the encoded detection reagent by the micro droplet generation device and a sample droplet by the uniformly mixed sample reagent;
the image generation unit is used for carrying out pairing fusion and amplification on the reagent liquid drops and the sample liquid drops and generating liquid drop images;
the quantitative analysis unit is used for quantitatively analyzing the liquid drop image in the bright field by utilizing a deep learning image recognition technology;
the model detection unit is used for obtaining a detection result of the detection reagent through poisson distribution calculation based on a quantitative analysis result, constructing a multiple digital nucleic acid detection model, and detecting a specified reagent by using the multiple digital nucleic acid detection model;
the image generation unit includes:
the pairing unit is used for loading the reagent liquid drops and the sample liquid drops into the PDMS microporous array chip for pairing;
the fusion unit is used for fusing the reagent liquid drops with different volumes and the sample liquid drops into target liquid drops by using a corona machine;
the amplification unit is used for amplifying the target liquid drops in a water bath amplification mode; wherein the water bath amplification temperature is 60-65 ℃ and the amplification time is 45-60 minutes;
the imaging unit is used for carrying out imaging treatment on the amplified target liquid drops to obtain liquid drop images;
the quantitative analysis unit includes:
an image recognition unit for recognizing whether a deposit exists in a target droplet corresponding to the droplet image in a bright field by using a deep learning image recognition technology, and recognizing a droplet color of the target droplet;
a positive judging unit for judging whether the target liquid drop is a positive liquid drop according to the existence of sediment;
the probability judging unit is used for classifying the target liquid drops according to the liquid drop colors of the target liquid drops and judging the negative probability of the target liquid drops by combining the classified target liquid drop numbers;
the image recognition unit includes:
the image input unit is used for inputting the liquid drop image into a YOLOv5 network model, and sequentially carrying out mosaic data enhancement, self-adaptive picture scaling and self-adaptive anchor frame calculation on the liquid drop image by utilizing an input end of the YOLOv5 network model;
a feature learning unit, configured to reduce information loss and learn features of the droplet image through a Focus structure and a CSP structure in a backhaul end of the YOLOv5 network model;
and the characteristic diagram prediction unit is used for predicting a characteristic diagram of the liquid drop image through the FPN structure and the PAN structure in the Neck end of the Yolov5 network model, and judging whether sediment and liquid drop color exist or not based on the characteristic diagram.
6. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the deep learning based multiple digital nucleic acid detection method of any one of claims 1 to 4 when the computer program is executed.
7. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, which when executed by a processor, implements the deep learning based multiple digital nucleic acid detection method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211516857.XA CN115820816B (en) | 2022-11-29 | 2022-11-29 | Multiple digital nucleic acid detection method, device and related medium based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211516857.XA CN115820816B (en) | 2022-11-29 | 2022-11-29 | Multiple digital nucleic acid detection method, device and related medium based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115820816A CN115820816A (en) | 2023-03-21 |
| CN115820816B true CN115820816B (en) | 2023-07-04 |
Family
ID=85532928
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211516857.XA Active CN115820816B (en) | 2022-11-29 | 2022-11-29 | Multiple digital nucleic acid detection method, device and related medium based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115820816B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113999764A (en) * | 2021-10-20 | 2022-02-01 | 郑州大学 | Integrated micro-droplet digital LAMP nucleic acid detection chip and detection method |
| CN115093961A (en) * | 2022-07-29 | 2022-09-23 | 华中农业大学 | Multi-volume liquid drop digital LAMP nucleic acid absolute quantitative detection device and method and application |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11450121B2 (en) * | 2017-06-27 | 2022-09-20 | The Regents Of The University Of California | Label-free digital brightfield analysis of nucleic acid amplification |
| US20220411858A1 (en) * | 2019-11-29 | 2022-12-29 | Mgi Tech Co., Ltd. | Random emulsification digital absolute quantitative analysis method and device |
| CN112029653B (en) * | 2020-08-17 | 2022-04-12 | 浙江大学 | Digital nucleic acid amplification detection method and integrated detection system based on CRISPR and Cas |
| CN113789259A (en) * | 2021-09-09 | 2021-12-14 | 哈尔滨工业大学 | Machine learning-assisted microdroplet digital nucleic acid detection device and detection method |
| CN114540469A (en) * | 2022-01-11 | 2022-05-27 | 深圳大学 | Digital nucleic acid quantification method based on non-uniform volume liquid drops and image processing |
| CN114752657A (en) * | 2022-05-05 | 2022-07-15 | 中山大学 | Polydisperse liquid drop digital nucleic acid detection method and application thereof |
-
2022
- 2022-11-29 CN CN202211516857.XA patent/CN115820816B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113999764A (en) * | 2021-10-20 | 2022-02-01 | 郑州大学 | Integrated micro-droplet digital LAMP nucleic acid detection chip and detection method |
| CN115093961A (en) * | 2022-07-29 | 2022-09-23 | 华中农业大学 | Multi-volume liquid drop digital LAMP nucleic acid absolute quantitative detection device and method and application |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115820816A (en) | 2023-03-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230122742A1 (en) | System and method for determining copies-per-unit-volume using pcr and flow control of droplets | |
| CN111886074A (en) | Apparatus, system and method for high throughput single cell analysis | |
| CN104722342B (en) | Slip chip device and method | |
| CN107407691A (en) | Device and system for the molecular bar code of unicellular amplifying nucleic acid target | |
| JP2018509915A (en) | Methods and compositions for combinatorial bar coding | |
| CN103429997A (en) | Methods and systems for performing digital measurements | |
| CN103998394A (en) | Cell capture system and method of use | |
| US20190185909A1 (en) | Methods and systems for analyzing nucleic acids | |
| Merrin | Frontiers in microfluidics, a teaching resource review | |
| EP3341489B1 (en) | Screening of nucleic acid sequence variations via random loading of optically-encoded particles | |
| Filatov et al. | Negative pressure provides simple and stable droplet generation in a flow-focusing microfluidic device | |
| Garcia-Rey et al. | High-resolution 3D printing fabrication of a microfluidic platform for blood plasma separation | |
| CN115820816B (en) | Multiple digital nucleic acid detection method, device and related medium based on deep learning | |
| CN107406886A (en) | For system, method and the kit for expanding or cloning in drop | |
| WO2018075577A1 (en) | Methods of integrated microfluidic processing for preventing sample loss | |
| CN114540469A (en) | Digital nucleic acid quantification method based on non-uniform volume liquid drops and image processing | |
| Chou | Microfabricated Devices for Rapid DNA Diagnostics | |
| Duchamp et al. | Microfluidic Device for Droplet Pairing by Combining Droplet Railing and Floating Trap Arrays | |
| Hsieh et al. | Enhancing throughput of combinatorial droplet devices via droplet bifurcation, parallelized droplet fusion, and parallelized detection | |
| Costa et al. | The microfluidic trainer: Design, fabrication and validation of a tool for testing and improving manual skills | |
| Zhang et al. | A femtoliter droplet array for massively parallel protein synthesis from single DNA molecules | |
| Lammertse et al. | Magnetically Functionalized Hydrogels for High‐Throughput Genomic Applications | |
| CN207828268U (en) | It is a kind of based on micro-fluidic drop regular array chip | |
| Zec | DEVELOPMENT OF MICROFLUIDIC PLATFORMS AS A TOOL FOR HIGH-THROUGHPUT BIOMARKER SCREENING | |
| Crivellari | Microfluidics & AC Electrokinetics: Developing Biosensors for Portable Diagnostics and Autologous Blood Doping in Endurance Athletes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |