A cytoskeleton parameter testing method based on random forest algorithm

Abstract

The application discloses a cytoskeleton parameter testing method based on a random forest algorithm, and cells are cultured by using a modified culture dish; when the cells in the dish continuously proliferate and cover 60% of the bottom area of the culture dish, the positions of selected experimental cells are marked by using an AFM, and mechanical parameters and topographic parameters of the experimental cells are obtained; meanwhile, Tubulin-tackerRed and Action-TrackerGreen are used to dye the cytoskeleton of the selected experimental cells, and the microfilament fluorescence intensity and the microtubule fluorescence intensity of the cells are obtained, which are used as cytoskeleton parameters. The cytoskeleton parameters are regressed and predicted according to the mechanical parameters and the topographic parameters of the selected cells by using the random forest algorithm. The cytoskeleton parameters of the cells can be quickly obtained by the application, and the economic and time costs are saved; and the cells are non-toxic and non-injurious during the detection of the topographic parameters and the mechanical parameters by using the AFM. After the detection is completed, the cells still have activity and can be used for other detections.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of biotechnology, and relates to a cytoskeleton parameter testing method based on a random forest algorithm. BACKGROUND

[0002] The cytoskeleton is a complex network structure composed of proteins in the cytoplasm of eukaryotic cells, mainly including microfilaments, microtubules and intermediate fibers, which are highly coordinated and distributed, and on the one hand determine and maintain the morphology of the cell, and on the other hand participate in the movement, division and cytoplasmic transport of the cell, and have important significance for signal transmission. In recent years, research has further confirmed that the cytoskeleton is closely related to life phenomena such as carcinogenesis, and researchers can achieve the purpose of treating cancer by changing the content of microfilaments and microtubules in cancer cells through drugs to inhibit cancer cell metastasis or induce cancer cell apoptosis. Therefore, the testing technology of cytoskeleton parameters is a complex and significant work. At present, the cytoskeleton parameters of cells are mainly obtained by biological experiments. Chemical reagents such as Tubulin-tackerRed and Action-Tracker Green are used to perform immunofluorescence staining on fixed cells, and then a fluorescence microscope is used to obtain the fluorescence picture of the cytoskeleton, and then the microfilament fluorescence intensity and the microtubule fluorescence intensity of the cell, i.e. the cytoskeleton parameters, are obtained. The workload is very heavy, time-consuming and labor-intensive, and the obtained cytoskeleton parameters are affected by subjective factors such as experimental conditions, researcher experience and fatigue. Therefore, it is necessary to develop a new cytoskeleton parameter testing technology that can accurately obtain the cytoskeleton parameters of cells and reflect the current degree of canceration of the cells.

[0003] Studies have shown that with the occurrence of cell carcinogenesis, the cytoskeleton parameters, the mechanical parameters and the morphology parameters of the cells all change accordingly, and the three are strongly related. Therefore, it is feasible to select the mechanical parameters and the morphology parameters of the cells to predict the cytoskeleton parameters of the cells. Among various cell detection technologies, atomic force microscopy (AFM) has a unique advantage in detecting the mechanical parameters and the morphology parameters of the cells with nanoscale spatial resolution in a liquid environment. Therefore, it is of great significance to select AFM to detect the mechanical parameters and the morphology parameters of the cells, predict the cytoskeleton parameters of the cells, and evaluate the current degree of canceration of the cells. SUMMARY

[0004] The present application aims to provide a cytoskeleton parameter testing method based on a random forest algorithm, which solves the problem of large error and low efficiency of the existing method mentioned in the background. Specifically, it combines micro-nano technology detection, confocal imaging technology and statistical methods. It is mainly used for predicting cytoskeleton parameters and has potential applications in the field of single-cell analysis.

[0005] In order to achieve the above object, the present application provides the following technical scheme: a cytoskeleton parameter testing method based on a random forest algorithm, comprising the following steps:

[0006] Step 1: The used culture dish is modified, and in the clean bench, a ruler and a carving knife are used to draw three 1cm-long marks in the horizontal and vertical directions at the center of the bottom of the culture dish, the interval between adjacent marks in each direction is 0.25cm, and the 9 points formed by the intersection of the 6 marks are labeled and sorted as 1-9;

[0007] Step 2: The modified culture dish is used for cell culture, and then the 9 intersection points on the bottom of the culture dish are photographed in the order of the label according to the optical microscope of the AFM, and the two cells closest to the intersection points are selected on each picture as the cells to be tested near the intersection points, and the order of the cells to be tested is determined;

[0008] Step 3: Extract the morphology parameters and mechanical parameters of the cells: compare the labeled cell pictures, and use the AFM to scan the selected cells near each intersection point in the order of the label, select the features that can effectively predict the cytoskeleton parameters according to the scanned data, including cell viscoelasticity, cell height, cell length, cell stiffness, cell surface roughness, and cell volume;

[0009] Step 4: The cells in the culture dish are fixed using an immunostaining fixative, and then the cytoskeleton is dyed using Tubulin-tackerRed and Action-TrackerGreen, and the optical lens of the fluorescence microscope is used to compare the labeled cell pictures, and the cytoskeleton images of each cell after dyeing are obtained in the order of the labeled cells near each intersection point, and the obtained pictures are analyzed by ImageJ software to obtain the microfilament fluorescence intensity and microtubule fluorescence intensity of the cell, i.e. the cytoskeleton parameters;

[0010] Step 5: Since the cytoskeleton parameters of the experimental cells will be affected by the first AFM experiment, in order to further ensure the accuracy of the experimental data, the fluorescence of the other cells in the dish after the dyeing experiment is photographed, and the cytoskeleton fluorescence images of the experimental cells and the other cells are processed, and the cytoskeleton parameters of the cells similar in size to the experimental cells and in normal growth state are calculated, and the cytoskeleton parameters of the experimental cells and the corresponding cells without AFM experiment are subjected to t test, P<0.05 proves that there is a significant difference between the two groups of data, and the difference between the cytoskeleton parameters of the experimental cells and the corresponding cells without AFM experiment is calculated, which represents the influence of the AFM experiment on the cytoskeleton parameters of the cells;

[0011] Step 6: the difference between the skeleton parameters of the experimental cell and the corresponding similar cell is calculated, and the mean value is taken, which represents the impact value of the skeleton parameters of the experimental cell affected by the AFM experiment;

[0012] Step 7: the impact value is used to modify the skeleton parameters of the experimental cell, and then the t test is performed on the skeleton parameters of the corresponding similar cell, P>0.05, proving that there is no significant difference between the two groups of data, and the impact value is effective;

[0013] Step 8: the cell staining causes the cell death, the cell skeleton parameters, the mechanical parameters of the cell and the morphology parameters of the cell to change, the first intersection near the No. 1 cell after the fluorescence staining experiment is selected for the second AFM scanning, and the height, length and volume of the cell at this time are obtained;

[0014] Step 9: according to the difference between the cell morphology parameters obtained by the two AFM experiments of the first intersection near the No. 1 cell, the change of the cell morphology parameters after the fluorescence staining experiment is represented;

[0015] Step 10: the box plot is used for abnormal value identification on the data of all cells measured by the AFM, and the abnormal value is discarded, the cell morphology parameters of the first intersection near the No. 1 cell obtained by the two AFM experiments in each experiment are counted, and the difference is calculated, and finally the mean value is taken as the correction value;

[0016] Step 11: in order to ensure the effectiveness of the correction value, the t test is used to test the correction value, and finally the P value>0.05, proving that the correction value is effective and is reserved;

[0017] Step 12: the morphology parameters of all cells measured by the first AFM experiment are further modified by using the correction value, and then the mechanical parameters and the cell skeleton parameters of each experimental cell are corresponded to obtain a complete data set;

[0018] Step 13: the cell skeleton parameters of the experimental cell are predicted by using the random forest regressor:

[0019] The modified morphology parameters and mechanical parameters of each cell are used as the input of the random forest regressor, the parameters of the random forest regressor are set, and the skeleton parameters of the cell are predicted by regression.

[0020] The beneficial effects of the present application are:

[0021] (1) the cell skeleton parameters are predicted by scanning the cell by the AFM, the purpose of quickly obtaining the cell skeleton parameters is achieved, the economic and time cost is saved, and the cancer degree of the cell is evaluated according to the prediction result.

[0022] (2) The present application is non-toxic and non-invasive to cells during the detection of cell morphology parameters and mechanical parameters using AFM. After the detection, the cells remain active and can be used for other detection. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The present application is a whole block diagram.

[0024] Figure 2 The present application is a schematic diagram of the bottom of the culture dish. DETAILED DESCRIPTION

[0025] The present application will be described in detail below in conjunction with the specific embodiments.

[0026] Example 1:

[0027] Due to the complexity of cell mechanical parameters, morphology parameters and cytoskeleton parameters, the regression prediction of cytoskeleton parameters is divided into the following 5 steps: 1. The culture dish is modified. 2. The esophageal cancer cells (TE-1) are cultured using the modified culture dish, and the selected experimental cells are labeled and sorted using AFM, and the mechanical parameters and morphology parameters of each cell are extracted in order according to the order. 3. The cells are fluorescently stained with Tubulin-tackerRed and Action-TrackerGreen, and the fluorescence intensity of microfilaments and microtubules of each cell is obtained using a fluorescence microscope, i.e. cytoskeleton parameters, and the influence of AFM experiment on the cytoskeleton parameters of experimental cells is calculated, and the cytoskeleton parameters of experimental cells are modified. 4. The first intersection near the first cell is scanned again using AFM, and the morphology parameters of the cell are obtained, all data are processed for outliers, and the correction value is obtained according to the morphology parameters of the first cell near the first intersection in each experiment, and the morphology parameters of the experimental cells are corrected, and a complete data set is established. 5. The random forest algorithm is used to realize the regression prediction of cytoskeleton parameters.

[0028] Step 1: Modification of culture dish

[0029] Since this experiment requires the use of two instruments and equipment, and the mechanical parameters, morphology parameters and cytoskeleton parameters of the same cell need to be corresponded, the culture dish where the experimental cells are located needs to be modified to determine the location of the tested cells. In the clean bench, three 1cm long marks are drawn in the horizontal and vertical directions at the center of the culture dish bottom using a ruler and a carving knife, the distance between adjacent marks in the same direction is 0.25cm, and the 9 intersections formed by the 6 marks are labeled and sorted in order from 1 to 9.

[0030] Step two: parameter extraction of selected cells using AFM:

[0031] Take the frozen esophageal cancer cells (TE-1) for cell recovery, and then use the improved culture dish to culture the cells, replace the new culture medium every day, until the cells proliferate to 60% of the area of the bottom of the culture dish, then according to the optical microscope of AFM, take pictures of the 9 intersection points on the bottom of the culture dish in order, and mark and sort the 2 cells closest to the intersection point on each picture to determine the order of the cells near each intersection point for detection. Using the QI mode in AFM, comparing the marked cell pictures, selecting 64x64 resolution to scan the cells according to the marked order, and obtaining the viscoelasticity, height, length, stiffness, surface roughness and volume of the cells.

[0032] Step three: fluorescent staining of cells to obtain cell skeleton parameters:

[0033] 1. After the AFM experiment, the cells need to be fluorescently stained. First, use the optical microscope module on the fluorescence microscope to compare the marked cell pictures to determine the location of each experimental cell near each intersection point after the AFM experiment, and then use 1x PBS (PH=7.4) preheated at 37° to wash the cells twice.

[0034] 2. Add Biaolai Bio's immunostaining fixative (YT086) to the culture dish to fully cover the cells, with a covering time of 10 minutes to fix the experimental cells.

[0035] 3. After fixing the cells, add Biaolai Bio's immunostaining wash solution (YT089) to the culture dish to completely cover the sample, and then use a shaker to gently shake and wash for 5 minutes. After removing the wash solution, add new wash solution, a total of 3 times.

[0036] 4. After the wash solution is removed, add Biaolai Bio's immunofluorescence staining secondary antibody diluent (YT090) to the culture dish, dilute Action-Tracker Green obtained by diluting Action-Tracker Green with Biaolai Bio's immunofluorescence staining secondary antibody diluent (YT090) at a ratio of 1:100, and dilute Tubulin-tackerRed obtained by diluting Tubulin-tackerRed with Biaolai Bio's immunofluorescence staining secondary antibody diluent (YT090) at a ratio of 1:75 to completely cover the experimental cells, and incubate at room temperature for 45 minutes.

[0037] 5. Add Biaolai Bio's immunostaining wash solution (YT089) to the culture dish to completely cover the sample, and then use a shaker to gently shake and wash for 5 minutes. After removing the wash solution, add new wash solution, a total of 4 times.

[0038] 6. Using fluorescent microscope to contrast the pictures of labeled cells, get the skeleton fluorescent images of experimental cells near each intersection in order.

[0039] 7. Using ImageJ software to process the obtained images, get the microfilament fluorescent intensity and microtubule fluorescent intensity of each cell, i.e. the cytoskeleton parameters.

[0040] Step four: Take fluorescent pictures of other cells in the dish to get the influence of AFM experiment on the cytoskeleton parameters of experimental cells:

[0041] 1. After getting the cytoskeleton fluorescent images of experimental cells in order, take cytoskeleton fluorescent pictures of other cells in the dish.

[0042] 2. Grayscale and binarize all cytoskeleton fluorescent images.

[0043] 3. Erode the obtained binarized images to reduce image noise and reduce the influence of residual dye and cell debris on experimental results.

[0044] 4. Dilate the eroded images to reduce the influence of erosion on image information.

[0045] 5. Use region growing method to calculate the number of pixel points in the cell area of each image to represent the size of the cell in the image.

[0046] 6. Select 5 images of cells with similar size and normal growth state to the experimental cells, and calculate the cytoskeleton parameters of these cells through ImageJ software.

[0047] 7. Compare the cytoskeleton parameters of these cells with the corresponding cytoskeleton parameters of experimental cells, and after t-test, finally P<0.05, proving that there is a significant difference between the cytoskeleton parameters of experimental cells and similar cells. Calculate the difference between the cytoskeleton parameters of experimental cells and similar cells in each experiment and take the mean value, which represents the influence of AFM experiment on the cytoskeleton parameters of experimental cells, i.e. the influence value.

[0048] 8. In order to ensure the effectiveness of the influence value, modify the cytoskeleton parameters of experimental cells using the influence value, and then perform t-test with the corresponding cytoskeleton parameters of similar cells, finally P>0.05, proving that there is no significant difference between the two data, and the influence value is effective.

[0049] Step five: Perform AFM experiment on cell No. 1 near intersection No. 1 again to get the correction value:

[0050] 1. After the fluorescence staining experiment, remove all the reagents from the culture dish. Using AFM and the marked cell images mentioned above, scan cell 1 near intersection 1 again with AFM. Select the tapping mode to obtain the current morphological parameters of the cell, namely height, length, and volume.

[0051] 2. For each experiment, box plots were used to identify outliers in the cell mechanical and morphological parameters, and outliers were discarded.

[0052] 3. Calculate the difference between the morphological parameters of cell 1 near intersection 1 in each experiment and the two AFM experiments, and sum them up and take the average as the final correction value.

[0053] 4. The morphological parameters of cells near the intersection of point 1 in all experiments were corrected using the correction value. Then, the correction value was compared with the actual morphological parameters of cells near the intersection of point 1 in all experiments using the second AFM scan. After t-test, the final P value was greater than 0.05, which proved that there was no significant difference between the two and the correction value was effective.

[0054] 5. Use correction values ​​to further correct the morphological parameters of cells measured in the first AFM scan in all experiments.

[0055] 6. Correspond the corrected cell morphology and mechanical parameters to the cytoskeleton parameters to establish a complete dataset.

[0056] Step 5: Use the random forest algorithm to perform regression prediction on the cellular skeleton parameters:

[0057] 1. The corrected morphological parameters, mechanical parameters, and cytoskeleton parameters of each cell are used as a complete dataset for training and prediction of the random forest regressor.

[0058] 2. Adjust the parameters of the random forest regressor to perform regression prediction of the cell's skeletal parameters. The random forest size was set to 100, and the dimension of the splitting attribute selected for each split node was the integer part of the square root of the total dimension of the original features. The experiment was repeated 100 times. In each regression prediction experiment, 50% of the samples in the dataset were selected for training the random forest regressor, and the remaining samples were used for regression prediction testing. The overall experimental result was the average of the 50 regression prediction experiments.

[0059] Matters not covered in this invention are common knowledge.

[0060] The above examples are only for illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and to implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the spirit and essence of the present application shall be covered within the protection scope of the present application.

Claims

1. A cytoskeletal parameter testing method based on a random forest algorithm, characterized by According to the following steps: Step 1: modify the culture dish used, in the clean bench, using a ruler and a carving knife to draw three 1cm long marks in the center of the bottom of the culture dish in both horizontal and vertical directions, the distance between the adjacent two marks in each direction is 0.25cm, and the 9 intersection points of the 6 marks are numbered 1-9; Step 2: use the modified culture dish to culture cells, then use the optical lens of the AFM to take pictures of the 9 intersection points on the bottom of the culture dish in the order of the numbers, and label and order the 2 cells closest to each intersection point in each picture to determine the order of cell detection near each point; Step 3: extract the morphology parameters and mechanical parameters of the cells: compare the labeled cell pictures, use the AFM to scan the selected cells near each intersection point in the order of the numbers, and select the characteristic data that can effectively predict the cytoskeleton parameters according to the scanned data; Step 4: fix the cells in the culture dish using immunostaining fixative, then stain the cytoskeleton using Tubulin-tackerRed and Action-TrackerGreen, and use the optical lens module of the fluorescence microscope to compare the labeled cell pictures to determine the location of the cells to be tested, and obtain the cytoskeleton fluorescence image of each experimental cell near each intersection point after staining, analyze the obtained image, and obtain the microfilament fluorescence intensity and microtubule fluorescence intensity of the cell as the cytoskeleton parameters; Step 5: Since the cytoskeleton parameters of the experimental cells will be affected by the first AFM experiment, to further ensure the accuracy of the experimental data, take fluorescence pictures of other cells in the dish after staining, process the cytoskeleton fluorescence images of the experimental cells and other cells, select cells similar in size to the experimental cells and in normal growth state to calculate their cytoskeleton parameters, and perform t-test on the cytoskeleton parameters of the selected cells and the corresponding experimental cells, P<0.05, proving that there is a significant difference between them, and calculating the difference between the cytoskeleton parameters of the experimental cells and the corresponding cells that have not undergone AFM experiment, which represents the impact of the AFM experiment on the cytoskeleton parameters of the experimental cells; Step 6: calculate the difference between the cytoskeleton parameters of the experimental cells and their corresponding similar cells in all experiments, and take the average value, which represents the impact value of the AFM experiment on the cytoskeleton parameters of the experimental cells; Step 7: use the impact value to modify the cytoskeleton parameters of the experimental cells, then perform t-test on the cytoskeleton parameters of the modified experimental cells and the corresponding similar cells, P>0.05, proving that there is no significant difference between them, and the impact value is effective; Step 8: Because the cell staining causes cell death, the cytoskeleton parameters, mechanical parameters and morphology parameters of the cells change, select the No.1 cell at the first intersection point after the fluorescence staining experiment to perform the second AFM scanning, and obtain the morphology parameters of the cell at this time; Step 9: According to the difference between the cell morphology parameters obtained by two AFM experiments of the first intersection cell No. 1, the change of the cell morphology parameters after the fluorescence staining experiment is represented. Step 10: Use box plot to identify outliers of all cell data obtained by AFM experiment, discard outliers, and count the morphology parameters of the first intersection cell No. 1 obtained by two AFM experiments in each experiment, and calculate the difference value respectively, and finally take the mean value as the correction value. Step 11: In order to ensure the effectiveness of the correction value, t-test is used to test the correction value, and finally P>0.05, which proves that the correction value is effective, and it is reserved. Step 12: Further correct the morphology parameters of all cells measured by the first AFM experiment, and correspond to the mechanical parameters and cytoskeleton parameters of each cell to obtain a complete data set. Step 13: Use the regression algorithm of random forest to obtain the prediction model of the cytoskeleton parameters of the cell.

2. The cytoskeletal parameter testing method based on random forest algorithm according to claim 1, characterized in that, By using atomic force microscope to detect the cell, the non-toxic and non-destructive detection of the cell is realized.

3. The cytoskeletal parameter testing method based on random forest algorithm according to claim 1, characterized in that, In step 3, the features that can effectively predict the cytoskeleton parameters of the cell are selected, which include the morphology parameters and mechanical parameters of the experimental cell, and the mechanical parameters of the experimental cell include the cell viscoelasticity, cell stiffness and cell surface roughness; the morphology parameters of the experimental cell include the cell volume, cell height and cell length.

4. The cytoskeleton parameter testing method based on random forest algorithm according to claim 1, characterized in that, In step 4, the cytoskeleton fluorescence image of the experimental cell is analyzed to obtain the microfilament fluorescence intensity and microtubule fluorescence intensity of the cell, i.e. the cytoskeleton parameters, which are realized by ImageJ software.

5. The cytoskeletal parameter testing method based on random forest algorithm according to claim 1, characterized in that, In step 5, non-experimental cells similar in size to experimental cells are selected, and the method is to extract the number of fluorescent pixels of experimental cells and non-experimental cells by region growing method, and the number of fluorescent pixels of two cells is similar, then the size of two cells is similar.

6. The cytoskeletal parameter testing method based on random forest algorithm according to claim 1, characterized in that, In step 8 or 9 or 12, the cell morphology parameters include: cell length, cell height and cell volume.

7. The cytoskeletal parameter testing method based on random forest algorithm according to claim 1, characterized in that, In step 13, the corrected morphology parameters and mechanical parameters of each cell are used as the input of the random forest regressor, the parameters of the random forest regressor are set, and the prediction model of the cytoskeleton parameters of the cell is obtained by cycle test, and the cytoskeleton parameters of the cell are regressed and predicted.

Images