Marine mussel protein Western blot quantitative detection method based on deep learning

Abstract

The invention provides a deep learning-based marine mussel protein Western blot quantitative detection method. The deep learning-based marine mussel protein Western blot quantitative detection method comprises the following steps: preparing a marine mussel protein expression detection system: marine mussel collection, protein extraction, Western blot system optimization and marine mussel Western blot glue map collection; classification and identification model weight parameter training: constructing a database based on an existing Western blot glue chart, distinguishing target stripes in the Western blot glue chart, and training weight parameters of a target stripe identification model; and carrying out protein expression relative quantitative analysis based on the training model: carrying out original input Western blot gel chart preprocessing and identification, and carrying out protein expression quantitative calculation. According to the method, the stability and the credibility of the marine mussel protein Western blot quantitative detection are improved through the combination of an efficient protein extraction method and a Western blot detection system, and the analysis flux and the detection efficiency are greatly improved through a Western blot gel diagram processing analysis method based on deep learning.

Description

Technical Field

[0001] The present invention relates to the field of marine biotechnology, and in particular to a deep learning-based marine mussel protein Western blot quantitative detection method. Background Art

[0002] Marine mussels, as a widely distributed, highly adaptable, relatively weakly mobile, and easily collected organism, play an important role in various marine environmental monitoring projects. Their representativeness and superiority stem from their status as a cornerstone species of offshore ecosystems and their ability to reflect the pollution status and biological health of the marine environment. Because marine mussels perform filter feeding activities and have a low metabolic rate, they have a strong ability to enrich pollutants. Therefore, marine mussels are widely regarded as one of the ideal environmental monitoring organisms. By monitoring the changes in key proteins of marine mussels under pollutant stress conditions, the relationship between biological damage and the health of the marine environment can be revealed, as well as the potential mechanism of biological damage caused by marine environmental pollutants.

[0003] Western blot is a commonly used detection method based on the specific immune reaction between antibodies and antigens in molecular biology and immunology. This method has also been gradually applied to marine ecotoxicology to characterize the expression of specific proteins. The basic principle of Western blot is to use polyacrylamide gel electrophoresis to separate protein samples of different molecular weights. The separated proteins are transfected onto the hybridization membrane. Through the specific recognition of antibodies and antigens and the subsequent development process, the expression of the target protein is visualized, and then targeted research is conducted on it.

[0004] In marine mussels, when using Western blot to conduct protein response-related research, not all protein extraction methods combined with any Western blot system can obtain the target band. In addition, there are few specific antibodies for marine organisms, and antibodies against other mammals (such as mice, rabbits, humans, etc.) are often used directly. Due to factors such as non-specificity, antibody titer and protein epitopes, it is difficult to obtain clear Western blot bands and conduct quantitative analysis. Therefore, it is necessary to establish a stable and suitable Western blot detection system for marine mussels, including: protein extraction, gel electrophoresis system, membrane transfer system, antibody selection, internal reference selection and relative quantitative analysis, so as to obtain accurate information on the expression of the target protein in specific tissues of marine mussels.

[0005] Traditionally, after obtaining the Western blot gel image, image processing software (such as ImageJ, ImageStudio, etc.) is used to load the digital image of the gel image, and then the target protein band area is selected for intensity measurement. Before performing the intensity measurement, the gel image is also subjected to background correction, homogenization and other operations. This process completely relies on the manual selection of the target protein band and the internal reference protein band to carry out the relative expression calculation of the protein, which is not only complicated to operate, but also has a certain degree of subjectivity. At present, with the development of deep learning image analysis technology, the detection throughput can be greatly improved and unbiased results can be output. Therefore, it is very important to carry out interdisciplinary cross-disciplinary research, and to carry out quantitative analysis of protein expression using an image processing method based on a deep learning algorithm on the basis of establishing a stable and effective Western blot detection system for marine mussels, in order to promote relevant research on the protein level in marine mussels. Therefore, the present invention proposes a Western blot quantitative detection method for marine mussel protein based on deep learning. Summary of the invention

[0006] The purpose of the present invention is to provide a quantitative detection method for protein expression in marine mussels, so as to apply it to the study of biological effects of marine pollution. By using the established stable and effective marine mussel Western blot detection system, combined with deep learning image processing technology, the marine mussel protein response can be revealed, the analysis efficiency and accuracy can be improved, and the protein research of marine mussels can be promoted.

[0007] The technical solution adopted by the present invention to solve its technical problem is: A deep learning-based Western blot quantitative detection method for marine mussel protein, comprising the following steps: S1. Marine mussel protein expression detection system: marine mussel collection, protein extraction and Western blot system optimization and marine mussel Western blot gel image collection; S2. Classification and recognition model weight parameter training: Build a database based on existing Western blot images, distinguish the target bands in the Western blot images, remove background interference, and train the weight parameters of the target band recognition model; S3. Relative quantitative analysis of protein expression based on the training model: original input Western blot gel image preprocessing, recognition and quantitative calculation of protein expression.

[0008] As a further technical solution of the present invention, step S1 includes: Total protein extraction and concentration detection: Weigh 30 mg of mussel tissue into a 1.5 mL centrifuge tube, chop the tissue with scissors, add 150 μL of tissue lysis solution (Western and IP cell lysis solution: PMSF = 100:1, and pre-cool to 4 ° C for standby use) to the chopped tissue, use a handheld homogenizer to homogenize until there is no obvious large piece of tissue, place the homogenate on ice for extraction for 30 minutes, and vortex mix every 10 minutes to fully extract the mussel tissue protein, then place the mixture in a centrifuge for 25 minutes (12000 rpm, 4 ° C), carefully aspirate the supernatant with a pipette, and place it in a pre-marked 1.5 mL centrifuge tubes, use the BCA protein concentration determination kit to detect the concentration of the extracted protein, dilute each tube of protein extract 10 times with ultrapure water, and dilute the BSA (2μg / μL) protein standard by half times into 6 concentration gradients (2, 1, 0.5, 0.25, 0.125, 0.0625, 0) to avoid multiple freeze-thaw cycles of the protein sample to affect the activity of the protein. Pipette 5μL of the protein to be tested and the protein standard into a 96-well plate, then add 200μL of the BCA working solution, incubate at room temperature for 5min, use an enzyme reader to measure the absorbance at a wavelength of 562nm, and draw a standard curve with the protein standard. Calculate the concentration of the protein to be tested based on the standard curve; Preparation of protein samples: According to the calculation results, each group of proteins was diluted to 3 μg / μL to ensure the consistency of the initial protein loading amount for Western blot (8 μL, 24 μg). After the concentration was unified, the protein samples were mixed with SDS loading buffer at a ratio of 5:1 and boiled in a 100°C water bath for 3-5 minutes to destroy the quaternary structure of the protein and neutralize the surface charge, ensuring that the migration speed of the protein to be tested in the SDS-PAGE gel is only related to the molecular weight of the protein to be tested, and has nothing to do with its spatial structure and other factors. The obtained protein samples were stored in a -80°C refrigerator for later use; SDS-PAGE gel protein separation: After adding concentrated gel premix to the gel interlayer, quickly add overlying water on top to isolate the air. When a clear boundary appears between the overlying water and the separation gel (at least 30 minutes), discard the overlying water and dry it with filter paper strips to prevent the separation gel and concentrated gel from being loosely combined. Continue to add concentrated gel premix to the separation gel that has been drained of water, and insert a gel comb. After the concentrated gel is completely solidified, carefully remove the gel comb, fix the gel plate in the electrophoresis tank, and add SDS-PAGE electrophoresis fluid to the electrophoresis tank. Add the protein to be tested and the prestained protein molecular weight marker to the spotting wells in the established order, perform electrophoresis at a constant voltage of 80V for 30 minutes, concentrate the protein, ensure that the starting point of protein migration in the separation gel is consistent, and then perform electrophoresis at a constant voltage of 120V for 70 minutes to allow the protein to migrate and separate in the separation gel. Transfer: Add SDS-PAGE transfer solution to the tray, put 4 sponges, 6 transfer filter papers, and 1 nitrocellulose membrane in it, take 2 sponges and place them on the black plywood of the sandwich transfer clamp, and place three filter papers soaked in transfer solution on them, stack them one by one, and use a scraper to remove bubbles, place the SDS-PAGE gel on the filter paper, place the nitrocellulose membrane on the gel, and then stack 3 transfer filter papers one by one, also making sure there are no bubbles between each layer, put the last 2 sponges on the top, clamp the plywood, then place the sandwich transfer clamp in the transfer core, add SDS-PAGE transfer solution to the electrophoresis tank, and transfer the membrane at 300mA constant current for 1h. Since the transfer process will release a lot of heat, the electrophoresis tank should be placed in an ice bath throughout the process; Blocking and antibody incubation: After the transfer, the nitrocellulose membrane was placed in an incubation box and washed twice with Western blot cleaning solution for 5 minutes each time. The washed nitrocellulose membrane was then placed in a blocking solution containing 5% bovine serum albumin for 1 hour. The blocked nitrocellulose membrane was taken out and placed in an incubation solution containing the primary antibody. The membrane was incubated at room temperature on a shaker (110 rpm) for at least 1 hour. The primary antibody was collected and the nitrocellulose membrane was washed 3 times with TBST for 5 minutes each time. The washed nitrocellulose membrane was placed in an incubation solution containing the secondary antibody. The membrane was incubated at room temperature on a shaker (110 rpm) for at least 1 hour. The secondary antibody was then collected and the nitrocellulose membrane was washed 3 times with TBST and once with TBS for 5 minutes each time. Development and visualization: The target gel strips were developed using an ECL chemiluminescence kit and visualized using an Odyssey infrared imaging system.

[0009] As a further technical solution of the present invention, step S2 includes: 1) Image acquisition: A dataset was constructed based on the Western blot images of the marine mussel protein expression experiment in the laboratory. 867 images were selected to form the dataset for model training, of which 694 images were used for model training and 173 images were used for model validation. The dataset was used to train the model to identify the target bands in the Western blot images. The training set was used to update the model parameters, and the validation set was used to evaluate the performance of the model. 2) Data preprocessing: Convert the collected Western blot images into grayscale images, use the open cv-rolling-ball algorithm to subtract the image background, and then invert the image according to the pixel value. Perform the operation of subtracting the current pixel value from 255 for each pixel value to obtain a new pixel value to form a new image with a black background and a white foreground. 3) Data labeling: Use LabelImg software to label images, mark the bounding box of the white strip area, and construct the data set for model training from the batch annotated images; 4) Model training parameter setting: During the training process, online data enhancement strategies were adopted, including mosaic enhancement, hybrid enhancement, and random perspective conversion; these techniques are used to increase the diversity and quantity of training data, prevent model overfitting, and improve generalization ability. For optimization, the AdamW optimizer was selected. The maximum number of training epochs was set to 300, and the initial learning rate was 0.001. The batch size was set to 16. The pre-trained parameters of the YOLOv8m model were used as the initial parameters for training, and transfer learning was adopted; 5) Training configuration: The training process is carried out on a supercomputing platform, which uses 2Intel8358P32C / 2.6G|512GB|4NVIDIAA800-80GB configuration and CentOS7.9 operating system. In order to utilize the computing power of GPU, PyTorch deep learning framework and NVIDIACUDA are used for GPU acceleration; 6) Intensity calculation and counting: Post-process each target band in the Western blot gel image, calculate the pixel area of ​​the image according to the coordinates of the a and b points of the identification box, Area = (xb-xa) * (yb-ya), extract the maximum brightness and minimum brightness of the identification box, accumulate the intensity of each pixel value in the identification box to obtain the integrated density, and divide the integrated density by the area to obtain the Mean gray value. The unit of all data is pixels. Count each identification box in a single image and mark the order above the identification box; 7) Result output: The result output is the recognized image and the result txt file, each image corresponds to a .txt file.

[0010] As a further technical solution of the present invention, step S3 includes: 1) Preprocessing of the original input gel image: Convert the Western blot gel image into a grayscale image, use the open cv-rolling-ball algorithm to subtract the image background, and then invert the image; 2) Reading the band to be tested: using the target band recognition model constructed in step S2 to read the pixel value of the band to be tested in the pre-processed Western blot gel image; 3) Quantitative calculation of protein expression: After obtaining the band pixels of the target protein and the internal reference protein, the formula RP=(TT / TR) / (CT / CR)① was used to perform relevant calculations, where RP represents relative protein expression; TT represents the pixels of the target protein in the treatment group, TR represents the pixels of the internal reference protein in the treatment group, CT represents the pixels of the target protein in the control group, and CR represents the pixels of the internal reference protein in the control group.

[0011] The beneficial effects of the present invention are: The present invention determines an efficient protein extraction method for marine mussels through comparative experiments, and determines a Western blot detection system that is effectively applicable to marine mussels through comparative experiments. The combination of the efficient protein extraction method and the Western blot detection system increases the stability and credibility of Western blot quantitative detection of marine mussel proteins, and the Western blot gel image processing and analysis method based on deep learning greatly improves the analysis throughput and detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is the overall technical roadmap of the embodiment of the present invention.

[0013] Figure 2 Diagram of the marine mussel protein expression detection system.

[0014] Figure 3 To construct a Western blot band identification method.

[0015] Figure 4 For the examination of dyed glue diagram.

[0016] Figure 5 This figure shows the effects of different loading proteins and Western blot systems on gel images.

[0017] Figure 6 This is a diagram showing the effects of sulfamethoxazole stress on the responses of key proteins in marine mussels. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0019] See also Figure 1The embodiment of the present invention provides a method for quantitative detection of marine mussel protein Western blot based on deep learning, comprising the following steps: S1. Marine mussel protein expression detection system: marine mussel collection, protein extraction and Western blot system optimization and marine mussel Western blot gel image collection; S2. Classification and recognition model weight parameter training: Build a database based on existing Western blot images, distinguish the target bands in the Western blot images, remove background interference, and train the weight parameters of the target band recognition model; S3. Relative quantitative analysis of protein expression based on the training model: original input Western blot gel image preprocessing, recognition and quantitative calculation of protein expression.

[0020] See also Figure 2 , Figure 2 The invention relates to a marine mussel protein expression detection system. (1) tissue sample; (2) protein extraction; (3) protein concentration detection; (4) SDS denaturation; (5) protein loading; (6) protein concentration and separation; (7) membrane transfer; (8) blocking; (9) primary antibody incubation; (10) secondary antibody incubation; (11) visualization; (12) analysis. In the embodiment of the present invention, the marine mussel protein expression detection system: Experimental material preparation: marine mussels, ophthalmic scissors, tweezers, Western and IP cell lysate, PMSF, BCA protein concentration determination kit, protein loading buffer, SDS-PAGE gel rapid preparation kit, SDS-PAGE electrophoresis fluid, nitrocellulose membrane, Western transfer fluid, Western washing buffer, Western blocking buffer, Western antibody diluent, QuickBlock blocking buffer (TBS and TBST), bovine serum albumin, ECL chemiluminescence detection kit, p38MAPK antibody, p-p38MAPK antibody, GAPDH antibody, goat anti-mouse IgG, antibody incubation box, handheld homogenizer, pipette tips, centrifuge, vertical electrophoresis system, electrophoresis power supply, water bath, horizontal shaker, Odyssey infrared imaging system, etc.

[0021] Extraction of marine mussel tissue: Use scissors to pry open the mussel shell from the ventral side, drain the seawater inside, cut the posterior adductor muscle of the mussel, open the shell, take out the gill tissue and digestive gland tissue and place them on ice for later use.

[0022] Total protein extraction and concentration detection: Weigh 30 mg of mussel tissue into a 1.5 mL centrifuge tube, chop the tissue with scissors, add 150 μL of tissue lysis solution (Western and IP cell lysis solution: PMSF = 100:1, and pre-cool to 4 ° C for standby use) to the chopped tissue, use a handheld homogenizer to homogenize until there is no obvious large piece of tissue, place the homogenate on ice for extraction for 30 minutes, and vortex mix every 10 minutes to fully extract the mussel tissue protein, then place the mixture in a centrifuge for 25 minutes (12000 rpm, 4 ° C), carefully aspirate the supernatant with a pipette, and place it in a pre-marked 1.5 mL centrifuge tubes, use the BCA protein concentration determination kit to detect the concentration of the extracted protein, dilute each tube of protein extract 10 times with ultrapure water, and dilute the BSA (2μg / μL) protein standard by half times into 6 concentration gradients (2, 1, 0.5, 0.25, 0.125, 0.0625, 0) to avoid multiple freeze-thaw cycles of the protein sample to affect the activity of the protein. Pipette 5μL of the protein to be tested and the protein standard into a 96-well plate, then add 200μL of the BCA working solution, incubate at room temperature for 5min, use an enzyme reader to measure the absorbance at a wavelength of 562nm, and draw a standard curve with the protein standard. Calculate the concentration of the protein to be tested based on the standard curve; Preparation of protein samples: According to the calculation results, each group of proteins was diluted to 3 μg / μL to ensure the consistency of the initial protein loading amount for Western blot (8 μL, 24 μg). After the concentration was unified, the protein samples were mixed with SDS loading buffer at a ratio of 5:1 (SDS loading buffer was sourced from Shanghai Bio-Tech Biotechnology Co., Ltd.), and boiled in a 100°C water bath for 3-5 minutes to destroy the quaternary structure of the protein and neutralize the surface charge, ensuring that the migration speed of the protein to be tested in the SDS-PAGE gel was only related to the molecular weight of the protein to be tested, and had nothing to do with its spatial structure and other factors. The obtained protein samples were stored in a -80°C refrigerator for later use; SDS-PAGE gel protein separation: After adding concentrated gel premix to the gel interlayer, quickly add overlying water on top to isolate the air. When a clear boundary appears between the overlying water and the separation gel (at least 30 minutes), discard the overlying water and dry it with filter paper strips to prevent the separation gel and concentrated gel from being loosely combined. Continue to add concentrated gel premix to the separation gel that has been drained of water, and insert a gel comb. After the concentrated gel is completely solidified, carefully remove the gel comb, fix the gel plate in the electrophoresis tank, and add SDS-PAGE electrophoresis fluid to the electrophoresis tank. Add the protein to be tested and the prestained protein molecular weight marker to the spotting wells in the established order, perform electrophoresis at a constant voltage of 80V for 30 minutes, concentrate the protein, ensure that the starting point of protein migration in the separation gel is consistent, and then perform electrophoresis at a constant voltage of 120V for 70 minutes to allow the protein to migrate and separate in the separation gel. Transfer: Add SDS-PAGE transfer solution to the tray, put 4 sponges, 6 transfer filter papers, and 1 nitrocellulose membrane in it, take 2 sponges and place them on the black plywood of the sandwich transfer clamp, and place three filter papers soaked in transfer solution on them, stack them one by one, and use a scraper to remove bubbles, place the SDS-PAGE gel on the filter paper, place the nitrocellulose membrane on the gel, and then stack 3 transfer filter papers one by one, also making sure there are no bubbles between each layer, put the last 2 sponges on the top, clamp the plywood, then place the sandwich transfer clamp in the transfer core, add SDS-PAGE transfer solution to the electrophoresis tank, and transfer the membrane at 300mA constant current for 1h. Since the transfer process will release a lot of heat, the electrophoresis tank should be placed in an ice bath throughout the process; Blocking and antibody incubation: After the transfer, the nitrocellulose membrane was placed in an incubation box and washed twice with Western blot cleaning solution for 5 minutes each time. The washed nitrocellulose membrane was then placed in a blocking solution containing 5% bovine serum albumin for 1 hour. The blocked nitrocellulose membrane was taken out and placed in an incubation solution containing the primary antibody. The membrane was incubated at room temperature on a shaker (110 rpm) for at least 1 hour. The primary antibody was collected and the nitrocellulose membrane was washed 3 times with TBST for 5 minutes each time. The washed nitrocellulose membrane was placed in an incubation solution containing the secondary antibody. The membrane was incubated at room temperature on a shaker (110 rpm) for at least 1 hour. The secondary antibody was then collected and the nitrocellulose membrane was washed 3 times with TBST and once with TBS for 5 minutes each time. Development and visualization: The target gel strips were developed using an ECL chemiluminescence kit and visualized using an Odyssey infrared imaging system.

[0023] See also Figure 3 , Figure 3 A represents the image preprocessing process, Figure 3 B is an example of data annotation. Figure 3 C is the optimized output image sample, Figure 3 D represents the calculation area of ​​the Western blot band identification box a and b points. Figure 3 E represents the output result after processing the image. In the embodiment of the present invention, the classification and recognition model weight parameter training includes: 1) Image acquisition: A dataset was constructed based on the Western blot images of the marine mussel protein expression experiment in the laboratory. 867 images were selected to form the dataset for model training, of which 694 images were used for model training and 173 images were used for model validation. The dataset was used to train the model to identify the target bands in the Western blot images. The training set was used to update the model parameters, and the validation set was used to evaluate the performance of the model. 2) Data preprocessing: Convert the collected Western blot images into grayscale images, use the open cv-rolling-ball algorithm to subtract the image background, and then invert the image according to the pixel value. Perform the operation of subtracting the current pixel value from 255 for each pixel value to obtain a new pixel value to form a new image with a black background and a white foreground. 3) Data labeling: Use LabelImg software to label images, mark the bounding box of the white strip area, and construct the data set for model training from the batch annotated images; 4) Model training parameter setting: During the training process, online data enhancement strategies were adopted, including mosaic enhancement, hybrid enhancement, and random perspective conversion; these techniques are used to increase the diversity and quantity of training data, prevent model overfitting, and improve generalization ability. For optimization, the AdamW optimizer was selected. The maximum number of training epochs was set to 300, and the initial learning rate was 0.001. The batch size was set to 16. The pre-trained parameters of the YOLOv8m model were used as the initial parameters for training, and transfer learning was adopted; 5) Training configuration: The training process is carried out on a supercomputing platform, which uses 2Intel8358P32C / 2.6G|512GB|4NVIDIAA800-80GB configuration and CentOS7.9 operating system. In order to utilize the computing power of GPU, PyTorch deep learning framework and NVIDIACUDA are used for GPU acceleration; 6) Intensity calculation and counting: Post-process each target band in the Western blot gel image, calculate the pixel area of ​​the image according to the coordinates of the a and b points of the identification box, Area = (xb-xa) * (yb-ya), extract the maximum brightness and minimum brightness of the identification box, accumulate the intensity of each pixel value in the identification box to obtain the integrated density, and divide the integrated density by the area to obtain the Mean gray value. The unit of all data is pixels. Count each identification box in a single image and mark the order above the identification box; 7) Result output: The result output is the recognized image and the result txt file, each image corresponds to a .txt file.

[0024] In an embodiment of the present invention, the relative quantitative analysis of protein expression based on the training model includes: 1) Preprocessing of the original input gel image: Convert the Western blot gel image into a grayscale image, use the open cv-rolling-ball algorithm to subtract the image background, and then invert the image; 2) Reading the band to be tested: using the target band recognition model constructed in step S2 to read the pixel value of the band to be tested in the pre-processed Western blot gel image; 3) Quantitative calculation of protein expression: After obtaining the band pixels of the target protein and the internal reference protein, the formula RP=(TT / TR) / (CT / CR)① was used to perform relevant calculations, where RP represents relative protein expression; TT represents the pixels of the target protein in the treatment group, TR represents the pixels of the internal reference protein in the treatment group, CT represents the pixels of the target protein in the control group, and CR represents the pixels of the internal reference protein in the control group.

[0025] In order to facilitate those skilled in the art to better understand the technical solution of the present invention, specific embodiments of the present invention are given as follows: Example 1 (protein extraction effect): Test organisms and reagents: Adult Mediterranean mussels (54-63 mm) were purchased from Donghe Aquatic Products Market in Zhoushan City, Zhejiang Province in June 2022. Western and IP cell lysis buffers, PMSF, BCA protein concentration determination kits, and protein molecular weight standards were purchased from Bio-Tech Biotechnology Co., Ltd. 1× ​​PBS and Coomassie Brilliant Blue Rapid Staining Solution were purchased from Beijing Solebold Technology Co., Ltd. All other reagents were domestic analytical grade.

[0026] Instruments and equipment: ophthalmic scissors, tweezers, antibody incubation box, handheld homogenizer, pipette tips, centrifuge, water bath, horizontal shaker, Odyssey infrared imaging system, etc.

[0027] Total protein extraction and concentration detection: Weigh 30 mg of mussel tissue into a 1.5 mL centrifuge tube, chop the tissue with scissors, and add 150 μL of different tissue lysate working solutions to the chopped tissue (PBS lysate and Western and IP cell lysate, respectively, and pre-cool to 4 ° C for standby use). Use a handheld homogenizer to homogenize until there is no obvious large piece of tissue. The homogenate is placed on ice for extraction for 30 minutes, and vortexed every 10 minutes to achieve the purpose of fully extracting mussel tissue protein. Then place the mixed solution in a centrifuge for 25 minutes (12000rpm, 4 ° C). Carefully aspirate the supernatant with a pipette and place it in a pre-marked 1.5 mL centrifuge tube. Calculate the protein concentration of each sample to be tested according to the standard curve (Table 1), and dilute the protein to be tested to 3 μg / μL according to their respective concentrations.

[0028] Table 1 Protein concentration

[0029] Protein separation and staining: After SDS denaturation, use SDS-PAGE gel to separate the protein to be tested. Finally, use Coomassie Brilliant Blue to stain the SDS-PAGE gel containing the protein to be tested and scan it for imaging.

[0030] Result analysis: Figure 4 M represents the protein molecular weight marker. 1 and 2 represent the protein extraction effect of PBS lysis method. 3 and 4 represent the protein extraction effect of this method. Figure 4 It can be seen that under the same protein loading amount (24 μg), the protein extraction effect of this method is better (see Appendix Figure 4 , columns 3 and 4), the resulting protein abundance and integrity are better.

[0031] Example 2 (Western blot system stability): Samples and reagents: Sample protein (protein obtained by two different extraction methods in Comparative Example 1). SDS-PAGE gel rapid preparation kit, SDS-PAGE electrophoresis solution, nitrocellulose membrane, Western transfer solution, Western washing solution, Western blocking solution, Western antibody diluent, QuickBlock blocking solution (TBS and TBST), bovine serum albumin, and ECL chemiluminescence detection kit were all purchased from Biyuntian Biotechnology Co., Ltd. pp38MAPK antibody was purchased from SantaCruzBiotechnology. Goat anti-mouse IgG was purchased from Hangzhou Xianzhi Biotechnology Co., Ltd. The remaining reagents (acrylamide, ammonium persulfate, tris(hydroxymethyl)aminomethane, glycine, sodium dodecyl sulfate, β-mercaptoethanol, methanol, ethanol, etc.) were all domestic analytical grade. The rest is the same as Comparative Example 1.

[0032] Total protein extraction and concentration detection: same as Example 1.

[0033] SDS-PAGE gel protein separation: Add concentrated gel premix to the gel plate interlayer, cover with overlying water to isolate the air, wait for 30 minutes to form a clear boundary, and then remove the overlying water. Use filter paper to absorb the moisture to prevent the gel from being loosely bound. Then add the concentrated gel premix to the separation gel again, insert the gel comb, and remove the comb after the concentrated gel solidifies. Fix the gel plate in the electrophoresis tank and inject SDS-PAGE electrophoresis fluid. Add the test protein and the labeled protein to the spotting wells in sequence, and concentrate the protein by electrophoresis at 80V for 30 minutes. Then electrophoresis at 120V for 70 minutes to separate and migrate the proteins. Two separation systems are used, namely the self-prepared system and the commercial system.

[0034] Transfer: Add SDS-PAGE transfer solution, 4 sponges, 6 transfer filter papers, and 1 nitrocellulose membrane to the tray. Take 2 sponges and place them on the black plywood of the sandwich transfer clamp, put three soaked filter papers on top, stack them one by one, and use a scraper to remove bubbles. Place the SDS-PAGE gel on the filter paper, the nitrocellulose membrane on the gel, and stack 3 transfer filter papers one by one to ensure that there are no bubbles in each layer. Finally, place the last 2 sponges on the top, clamp the plywood, place it in the transfer core, and add it to the electrophoresis tank. Transfer the membrane for 1 hour under 300mA constant current conditions, and keep the electrophoresis tank in an ice bath throughout the process. Two separation systems are used, namely the self-prepared system and the commercial system.

[0035] Blocking and antibody incubation: After transfer, the nitrocellulose membrane was placed in an incubation box and washed twice with Western blot washing solution, each for 5 minutes. It was placed in a blocking solution containing 5% bovine serum albumin for 1 hour. The nitrocellulose membrane was placed in the incubation solution containing the primary antibody and incubated at room temperature on a shaker (110 rpm) for at least 1 hour. The primary antibody was collected and the nitrocellulose membrane was washed 3 times with TBST, each for 5 minutes. The nitrocellulose membrane was placed in the incubation solution containing the secondary antibody and incubated at room temperature on a shaker (110 rpm) for at least 1 hour. The secondary antibody was collected and the nitrocellulose membrane was washed 3 times with TBST and once with TBS, respectively, for 5 minutes each time.

[0036] Development and visualization: The strips were developed using an ECL chemiluminescence kit and the proteins were visualized using the Odyssey infrared imaging system.

[0037] Result analysis: Figure 5 A represents the Western blot gel image. 1 and 2 represent the gel images of protein extracted by PBS lysis method and Western and IP cell lysate obtained under the self-prepared Western blot system. 3 and 4 represent the gel images of protein extracted by PBS lysis method and Western and IP cell lysate obtained under the commercial Western blot system. Figure 5 B represents the strip grayscale analysis result. Figure 5 It can be seen that when the amount of protein loaded (24 μg) is the same, the Western and IP cell lysate protein extraction method is better than the PBS lysis method. The gel image obtained under the commercial Western blot system is due to the self-made Western blot system. The combination of the protein extraction method (Western and IP cell lysate protein extraction method) and the Western blot system (commercial product) proposed in this method can stably output clear gel images to carry out relevant mechanism mining research.

[0038] Example 3 (Effect of sulfamethoxazole stress on the response of key proteins in marine mussels): Stress experiment setup: Marine mussels were used for stress experiments, with two culture systems: control group (only artificial seawater added) and sulfamethoxazole stress group (50 μg / L). The stress experiment lasted for 6 days, with artificial seawater replaced and sulfamethoxazole working solution added every day. After the stress experiment, the mussels were dissected to obtain gill tissue for subsequent analysis.

[0039] Reagents and instruments: p38 MAPK antibody was purchased from Santa Cruz Biotechnology. GAPDH antibody was purchased from Hangzhou Xianzhi Biotechnology Co., Ltd. Sulfamethoxazole was purchased from Beijing Bailingwei Technology Co., Ltd. The remaining reagents and instruments were the same as those in Examples 1 and 2.

[0040] Total protein extraction and concentration detection, SDS-PAGE gel protein separation, membrane transfer, blocking and antibody incubation, development and visualization: all the same as Example 2 (the present invention establishes a stable Western blot system).

[0041] Relative quantitative analysis of protein expression: The gel image was analyzed using the deep learning-based image processing method established by the present invention, and relative quantitative calculations were performed. Figure 6 A represents Western blot. 1, 2, and 3 represent three biological replicates. p38 represents p38MAPK protein, pp38 represents phosphorylated p38MAPK protein, and GAPDH represents glyceraldehyde-3-phosphate dehydrogenase (as an internal reference protein). T represents the treatment group, and C represents the control group. Figure 6 The relative quantitative analysis results of B protein expression. Figure 6 As can be seen from A, the target proteins (p38 and pp38) and the internal reference protein (GAPDH) have clear and single bands. Figure 6 B shows that after sulfamethoxazole stress, the expression of key immune regulatory proteins in marine mussels changed, with p38 upregulated by 1.523 times and pp38 upregulated by 2.369 times. It can be seen that the protein extraction method of the present invention combined with the Western blot system can be effectively used in the quantitative experiment of marine mussel protein.

[0042] Comparison between deep learning image analysis method and common method (ImageJ): This method and ImageJ method were used to analyze the same Western blot gel map, and the results are shown in Table 2. The relative protein expression level trends obtained by this method and ImageJ method were consistent, and there was no significant difference between the two. In addition, this method greatly improved the analysis efficiency while ensuring the accuracy of the analysis results.

[0043] Table 2 Comparison of relative quantitative analysis methods for protein expression

[0044] In summary, the present invention provides a stable and effective protein expression analysis system for marine mussels. The deep learning image processing method increases the analysis throughput while improving the efficiency. The present invention provides an efficient and stable implementation plan for research related to marine mussel protein expression and environmental pollution environmental effect evaluation.

[0045] It should be noted that, in this article, the terms "include", "comprises" or any other variations 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 sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0046] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for quantitative detection of marine mussel protein by Western blot based on deep learning, characterized in that: The following steps are involved: S1. Marine mussel protein expression detection system: marine mussel collection, protein extraction and Western blot system optimization and marine mussel Western blot gel image collection; S2. Classification and recognition model weight parameter training: Build a database based on existing Western blot images, distinguish the target bands in the Western blot images, remove background interference, and train the weight parameters of the target band recognition model; S3. Relative quantitative analysis of protein expression based on the training model: original input Western blot gel image preprocessing, recognition and quantitative calculation of protein expression.

2. A method for quantitative detection of marine mussel protein by Western blot based on deep learning according to claim 1, characterized in that: Step S1 includes: Total protein extraction and concentration detection: weigh the mussel tissue in a centrifuge tube, chop the tissue with scissors, add tissue lysis solution to the chopped tissue, use a handheld homogenizer to homogenize until there is no obvious large piece of tissue, place the homogenate on ice for extraction, and vortex mix it several times to fully extract the mussel tissue protein, then centrifuge the mixture, carefully aspirate the supernatant with a pipette, and place it in a pre-marked centrifuge tube, use the BCA protein concentration determination kit to detect the concentration of the extracted protein, dilute each tube of protein extract 10 times with ultrapure water, dilute the BSA protein standard by half times in 6 concentration gradients, avoid multiple freezing and thawing of the protein sample, which affects the activity of the protein, respectively aspirate the protein to be tested and the protein standard into a 96-well plate, then add the BCA working solution, incubate at room temperature for 5 minutes, use an enzyme reader to measure the absorbance at a wavelength of 562nm, and draw a standard curve with the protein standard, and calculate the concentration of the protein to be tested according to the standard curve; Preparation of protein samples: According to the calculation results, each group of proteins was diluted to 3 μg / μL to ensure the consistency of the initial protein loading amount for Western blot. After the concentration was unified, the protein sample was mixed with SDS loading buffer at a ratio of 5:1 and boiled in a 100°C water bath for 3-5 minutes. The obtained protein sample was stored in a -80°C refrigerator for later use; SDS-PAGE gel protein separation: After adding concentrated gel premix to the gel interlayer, quickly add overlying water on top to isolate the air. When a clear boundary appears between the overlying water and the separation gel, discard the overlying water and dry it with a filter paper strip to prevent the separation gel and concentrated gel from being loosely combined. Continue to add concentrated gel premix to the separation gel that has been drained of water, and insert a gel comb. After the concentrated gel is completely solidified, carefully remove the gel comb, fix the gel plate in the electrophoresis tank, and add SDS-PAGE electrophoresis fluid to the electrophoresis tank. Add the protein to be tested and the prestained protein molecular weight marker to the spotting wells in the established order, perform electrophoresis at a constant voltage of 80V for 30 minutes, concentrate the protein, ensure that the starting point of protein migration in the separation gel is consistent, and then perform electrophoresis at a constant voltage of 120V for 70 minutes to allow the protein to migrate and separate in the separation gel. Transfer: Add SDS-PAGE transfer solution to the tray, put 4 sponges, 6 transfer filter papers, and 1 nitrocellulose membrane in it, take 2 sponges and place them on the black plywood of the sandwich transfer clamp, and place three filter papers soaked in transfer solution on them, stack them one by one, and use a scraper to remove bubbles, place the SDS-PAGE gel on the filter paper, place the nitrocellulose membrane on the gel, and then stack 3 transfer filter papers one by one, also making sure there are no bubbles between each layer, put the last 2 sponges on the top, clamp the plywood, then place the sandwich transfer clamp in the transfer core, add SDS-PAGE transfer solution to the electrophoresis tank, and transfer the membrane at 300mA constant current for 1h. Since the transfer process will release a lot of heat, the electrophoresis tank should be placed in an ice bath throughout the process; Blocking and antibody incubation: After the transfer, the nitrocellulose membrane was placed in an incubation box and washed twice with Western blot cleaning solution for 5 minutes each time. The washed nitrocellulose membrane was then placed in a blocking solution containing 5% bovine serum albumin for 1 hour. The blocked nitrocellulose membrane was taken out and placed in an incubation solution containing the primary antibody. The membrane was incubated at room temperature on a shaker for at least 1 hour. The primary antibody was collected. The nitrocellulose membrane was washed 3 times with TBST for 5 minutes each time. The washed nitrocellulose membrane was placed in an incubation solution containing the secondary antibody. The membrane was incubated at room temperature on a shaker for at least 1 hour. The secondary antibody was then collected. The nitrocellulose membrane was washed 3 times with TBST and once with TBS for 5 minutes each time. Development and visualization: The target gel strips were developed using an ECL chemiluminescence kit and visualized using an Odyssey infrared imaging system.

3. A method for quantitative detection of marine mussel protein Western blot based on deep learning according to claim 1, characterized in that: Step S2 includes: 1) Image acquisition: A dataset is constructed based on existing Western blot images. The dataset is used to train the model to identify the target bands in Western blot images. 2) Data preprocessing: Convert the collected Western blot images into grayscale images, use the open cv-rolling-ball algorithm to subtract the image background, and then invert the image according to the pixel value. Perform the operation of subtracting the current pixel value from 255 for each pixel value to obtain a new pixel value to form a new image with a black background and a white foreground. 3) Data annotation: Use LabelImg software to annotate images, mark the white stripe area with a bounding box, and construct a dataset for model training from the batch annotated images; 4) Model training parameter setting: During the training process, online data enhancement strategies were adopted, including mosaic enhancement, hybrid enhancement and random perspective conversion; 5) Training configuration: The training process is carried out on a supercomputing platform, which uses 2Intel8358P32C / 2.6G|512GB|4NVIDIAA800-80GB configuration and CentOS7.9 operating system. In order to utilize the computing power of GPU, PyTorch deep learning framework and NVIDIACUDA are used for GPU acceleration; 6) Intensity calculation and counting: Post-process each target band in the Western blot gel image, calculate the pixel area of ​​the image according to the coordinates of the a and b points of the identification box, Area = (xb-xa) * (yb-ya), extract the maximum brightness and minimum brightness of the identification box, accumulate the intensity of each pixel value in the identification box to obtain the integrated density, and divide the integrated density by the area to obtain the Mean gray value. The unit of all data is pixels. Count each identification box in a single image and mark the order above the identification box; 7) Result output: The result output is the recognized image and the result txt file.

4. A method for quantitative detection of marine mussel protein Western blot based on deep learning according to claim 1, characterized in that, Step S3 includes: 1) Preprocessing of the original input gel image: Convert the Western blot gel image into a grayscale image, use the open cv-rolling-ball algorithm to subtract the image background, and then invert the image; 2) Reading the band to be tested: using the target band recognition model constructed in step S2 to read the pixel value of the band to be tested in the pre-processed Western blot gel image; 3) Quantitative calculation of protein expression: After obtaining the band pixels of the target protein and the internal reference protein, the formula RP=(TT / TR) / (CT / CR) was used to perform relevant calculations, where RP represents relative protein expression; TT represents the pixels of the target protein in the treatment group, TR represents the pixels of the internal reference protein in the treatment group, CT represents the pixels of the target protein in the control group, and CR represents the pixels of the internal reference protein in the control group.

Images