CN112082957A

CN112082957A - Method for detecting monoclonal antibody and application thereof

Info

Publication number: CN112082957A
Application number: CN202010710574.3A
Authority: CN
Inventors: 陶然; 张卫凯; 李勤; 吕蒙; 李伟; 魏泽文
Original assignee: Beijing Institute of Technology BIT
Current assignee: Beijing Institute of Technology BIT
Priority date: 2020-07-22
Filing date: 2020-07-22
Publication date: 2020-12-15
Anticipated expiration: 2040-07-22
Also published as: CN112082957B

Abstract

The invention relates to a method for detecting a monoclonal antibody and application thereof, wherein the method for detecting the monoclonal antibody comprises the following steps: and acquiring hyperspectral data of the cell to be detected through hyperspectral imaging detection, and inputting the hyperspectral data into a machine learning model to acquire the antibody secretion type of the cell to be detected. According to the invention, by introducing a hyperspectral imaging technology and a microfluidic technology, the separation of single cells is realized, a hyperspectral image of a single cell is rapidly acquired, required characteristic information is rapidly and accurately excavated from a large amount of hyperspectral data by using a machine learning algorithm, and the high-precision identification and classification of the single cells are automatically realized. The analysis process has the characteristics of high throughput and intellectualization, and the result has high accuracy and high sensitivity, thereby providing a rapid, non-contact and nondestructive identification method for antibody drug discovery and cell analysis.

Description

Method for detecting monoclonal antibody and application thereof

Technical Field

The invention relates to the field of cell analysis, in particular to a method for detecting a monoclonal antibody and application thereof.

Background

The monoclonal antibody medicine has specific targeting property, high sensitivity and high affinity, and is often applied to the fields of other clinical diseases such as tumor, autoimmunity, infection and the like and diagnosis and detection. In the research and development process of antibody drugs, obtaining antibodies with high specificity, high affinity, strong competitiveness and neutralizing activity is a key step and evaluation index in the development of antibody drugs.

At present, methods for preparing antibodies include hybridoma technology, single B cell technology, antibody library (phage library, ribosome library, artificially synthesized antibody library, etc.) technology, and immunology map technology; firstly, the hybridoma technology is the most classical technology, and the target antibody can be locked by separating B lymphocytes (bone marrow, peripheral blood, lymph nodes, spleen and other tissues and organs) from the bodies of mice, rabbits, camels and other species inoculated with antigens, cells or other immunogens, fusing the B lymphocytes with immortalized myeloma cells and then screening for several months; secondly, the single B cell technology is to separate B cells from human or animals infected with antigen, screen out the B cells of specific antibody by a fluorescence labeling method, and then amplify a gene sequence for encoding the antibody; then the antibody library technology is that antibody genes separated from B cells are inserted into expression vectors such as bacteriophage to form an antibody library, washing and screening are carried out for multiple times, artificial directed evolution is carried out to obtain high-affinity antibodies, and the complex process and the screening period restrict the speed and the efficiency of antibody discovery; in addition, the immunomapping technique, based on Next-generation Sequencing (NGS), enables complete transcriptome Sequencing analysis of B-cell populations. mRNA is extracted from the separated B cell, and the reverse transcription amplification antibody gene is reverse transcribed to establish an immune map, and the pairing antibody can be presumed according to the frequency of the heavy chain and the light chain, and the heavy chain and the light chain can also be freely matched, so that the diversity of the antibody is improved.

In the prior art, methods for detecting target antibodies include enzyme-linked immunosorbent assay (ELISA), flow cytometry, surface plasmon resonance technology, microfluidic technology, and the like. Among them, the traditional ELISA method has the disadvantages of low throughput, low efficiency, low accuracy, etc. The flow cytometry can analyze each cell in real time under flowing liquid, so that high-throughput detection is realized; however, the method depends on fluorescent signal detection, such as fluorescent substance labeled antigen, antibody and the like, and the fluorescent signal has the problems of quenching, signal to noise ratio and experimental environment interference, so that the result is misjudged; flow cytometry is based on the high-frequency electric field to realize the sorting of cells, and the strong electric field has influence on the activity of the cells.

The microfluidic technology realizes the detection and analysis of single cells and provides a new idea for screening the antibody. Micro/nano processing technology is adopted to manufacture micro/nano volume geometric structure chambers, uniform or different-sized physical barrier structures and the like, the size of the physical barrier structures conforms to the size of cells, and the physical barrier structures are used for capturing single cells. According to the literature report, the antibody secretion rate of each cell is 3.66 multiplied by 10⁵fM/min, in nanoliter/picoliter volume, the antibody secreted in 1 hour or more can reach the detection level, and no monoclonal is required to be formed, so that the detection period is shortened. The other method is a liquid drop-based method, wherein single cells and fluorescent probes are wrapped inside by oil phase, fiber materials and the like to form liquid drops, and specific cells are sorted out according to the fluorescent intensity by microspheres. Potential target antibodies are mined based on high-throughput single-cell analysis of microfluidic technology. The above techniques still rely on fluorescent signals to find specific, high affinity, functional antibodies from a large population of cells.

The Hyperspectral imaging technology (HSI) combines the imaging technology with the spectroscopic technology to detect two-dimensional geometric space and one-dimensional spectral information of a target, utilizes the difference of reflection and transmission spectra of an object in the range of electromagnetic waves (including visible light, Near Infrared (NIR), mid-infrared (MIR) and far infrared) to distinguish different material information, and has the characteristics of wide spectral range, high resolution and integration of spectra. In the HSI technology, a sample is scanned and an optical image of the sample at each wavelength is acquired through hundreds of nanoscale spectral bands (spectral bands), so that image information of the sample at each band is acquired, the spectral information of each pixel point at each band is read, and fine spectral differences among different substances are captured. The hyperspectral image data not only contain spectral information, but also contain image information, can reflect the physical morphology and biochemical component information of a detection target at the same time, and has high spatial resolution and high spectral resolution.

HSI is a new, non-contact optical diagnostic technique with dual functions of spectroscopy and imaging, and has been widely used in the biomedical field. The spectral characteristics of different tissues or organs depend on their own biochemical and tissue structure characteristics, which provide theoretical basis for distinguishing normal tissues from diseased tissues. At a particular wavelength, normal tissue and different diseased tissue differ in their chemical composition and physical characteristics, and therefore their reflectivity, absorption of electromagnetic energy, and thus characteristic spectral peaks. Qualitative or quantitative detection can be provided for different tissue pathological states through the spectral signals, and visualization of tissue pathology is realized; the spectral imaging provides an effective auxiliary diagnostic means for clinical medicine, and the method is already used for detecting cancers, such as gastric cancer, breast cancer, cervical cancer, skin cancer, prostatic cancer, colorectal cancer, ovarian cancer and the like, and detecting diabetes, retina and the like, and can assist in early diagnosis of diseases and monitoring of prognosis curative effect.

Disclosure of Invention

In order to solve at least one of the problems in the prior art, the invention provides a method for detecting a monoclonal antibody, which is used for finding a target antibody according to the spectral information difference among different cells and expanding the application of a hyperspectral imaging technology in the field of antibody drug development.

In a first aspect, the present invention provides a method for detecting a monoclonal antibody, comprising:

acquiring hyperspectral data of a cell to be detected through hyperspectral imaging, inputting the hyperspectral data into a machine learning model, and acquiring an antibody secretion type of the cell to be detected;

and the machine learning model is obtained by training based on sample cells which are homologous with the cells to be detected and corresponding cell antibody secretion types.

Further, the construction of the machine learning model comprises the following procedures:

acquiring hyperspectral data of a plurality of sample cells which are homologous with the cells to be detected to construct a training sample set, constructing a plurality of training sample subsets according to the antibody secretion types of the sample cells by using the training sample set, wherein each training sample subset corresponds to one antibody secretion type label;

and inputting all training sample subsets into a machine learning model for training.

Specifically, the training sample subset is constructed according to the antibody secretion type of the cells, for example, the labels of the training sample subset composed of all antibody-secreting cells are 1, and the labels of all antibody-non-secreting cells are 0; the cell labels secreting different antibodies are different natural numbers

Further, the machine learning model is constructed by any algorithm of a Bayesian classifier, a support vector machine and a deep learning network, and/or the machine learning model is preferably checked by a K-fold cross-validation method, wherein K is any integer of 1-10.

Further, the machine learning model is a convolutional neural network combined with Gabor filtering, the size of a convolutional kernel in each convolutional layer is assumed to be 3 × 3, four branches are provided, and each branch is provided with a modulated convolutional kernel, which is defined as:

wherein

A modulation convolution kernel representing the connection between the ith input and the jth output in the kth branch of the l layer;

then the convolution kernel with the ith input and the jth output connected in l layers, and in order to train fewer parameters and fit the data with more parameters, only one set of convolution kernels is generated in each layer of convolution layer, and then modulated by different Gabor kernelsObtaining different modulation convolution kernels to relieve network overfitting, thereby improving the generalization capability of the network;

representing a frequency f_kAnd the direction is theta_jThe Gabor kernel is generated by using parameters of 4 different frequencies (1, 1/2, 1/3, 1/4) and 8 different directions (0, pi/8, 2 pi/8, 3 pi/8, 4 pi/8, 5 pi/8, 6 pi/8 and 7 pi/8) in a network model of project design, and 8 Gabor kernels with different directions and the same frequency parameters in each branch are used for modulating convolution kernels.

The convolutional neural network combined with Gabor filtering comprises four convolutional layers, an average pooling layer and a full-connection layer, wherein 32 feature maps are set and output by each convolutional layer, the feature maps are converted into one-dimensional feature vectors by the last average pooling layer, dense calculation is carried out by the full-connection layer, and the prediction probability of each category is given through a Softmax activation function.

And further, a hyperspectral data preprocessing flow is further included before the hyperspectral data is input into the machine learning model, and the hyperspectral data preprocessing flow is one or more combinations of filtering processing, baseline removal and normalization.

Further, the cell to be detected is a single cell suspension or is prepared into the single cell suspension after being incubated with a detection substance; the detection substance is preferably any one or more of protein, nucleic acid, small molecule and fluorescent dye, or derivatives thereof; the derivative is preferably a product after modification or self-assembly.

Further, the cell to be tested is selected from one or more of human cells, animal cells, plant cells, microbial cells or phage;

preferably, the human cells are preferably living cells or modified cells derived from human organs, tissues and peripheral blood in vitro, and/or the animal cells are preferably living cells or modified cells derived from animal organs, tissues and peripheral blood in vitro, and/or the plant cells are plant pollen cells or living cells derived from root, stem and leaf in vitro, and/or the microbial cells are mononuclear microbial cells.

Further preferably, the animal cell may be a mammalian cell, and the microbial cell may be a yeast cell.

Further, before the hyperspectral imaging detection, transferring the cells to be detected into a microfluidic chip to form a single cell array; the material of the microfluidic chip is preferably one or more of silicon chip, quartz material, glass material, organic polymer or calcium fluoride material or a plurality of materials obtained by doping, and the doping mode is vapor deposition, ion beam sputtering, a chemical method and the like; the material has characteristics of photosensitivity, conductivity, biocompatibility, electromagnetism and the like.

According to the invention, single cells are captured in the structure, the cavity, the liquid drop or the microsphere through the microfluidic array chip, and meanwhile, a group of single cells are subjected to imaging analysis, so that the rapid identification of the single cells can be completed. Meanwhile, the microarray chip, the structural shape, the number and the size can be adjusted according to different types of cells, the existing processing technology can meet the requirements, and the microarray chip is easy to realize and low in cost for technicians in the field.

The invention adopts a microfluidic technology and a micro-processing method, a hyperspectral imaging technology and a computer algorithm to image single cells, designs a brand new single cell detection method and a solution aiming at antibody discovery, and can screen out cells secreting specific antibodies from a large number of cells; and provides a corresponding operation method and an implementation scheme to achieve the aim of rapidly screening cells.

The invention further provides the use of the method in screening for cells secreting monoclonal antibodies.

The cell is preferably one or more of a B cell, a hybridoma cell, a CHO cell, a yeast cell, a bacterial or a plant cell.

The invention further provides the use of the method in spectral imaging of single cells.

The invention further provides the use of the method in the analysis of the composition, morphology and other taxonomic identification of single cells.

The invention further provides for the use of the method for the quantitative and qualitative detection of cells secreting various cytokines or other cellular secretions.

The method can also be used for cell analysis related to drug development and scientific research, and is not limited to single cell level analysis.

As a preferred embodiment, the present invention provides a method of constructing a machine learning model, comprising:

s1, obtaining a single cell suspension, or incubating the single cell suspension with a detection substance, and then preparing the single cell suspension;

s2, introducing the S1 single-cell suspension into a microfluidic chip to prepare a single-cell array;

s3, placing the single cell array under a microscope for hyperspectral imaging and spectral data acquisition;

s4, preprocessing the hyperspectral data, establishing an optimal machine learning model, and identifying and analyzing the preprocessed hyperspectral data by using the optimal model.

Further, S1 is that the cell to be detected is obtained to be preprocessed into single cell suspension, washed for 2 times, suspended by phosphate buffer solution, added with proper amount of detection substance, reacted for 1 hour, washed for 2 times, and prepared into single cell suspension by using normal saline or cell and other penetrating solution for standby; the detection substance is protein, nucleic acid, small molecule, fluorescent dye and their derivatives, modified products and self-assembly products.

Further, S2 is to introduce the single cell suspension obtained in step S1 into a microfluidic chip to separate the single cell array, and introduce a cell culture solution to wash away the excess cells.

Further, in step S3, the microfluidic chip containing the single cell array is placed under a microscope to adjust the focal length and light intensity, and parameters such as the acquisition time are set, so as to perform hyperspectral imaging to obtain hyperspectral data of each single cell.

Further, step S4 is to preprocess the hyperspectral data obtained in step S3, input the established machine learning model, and perform identification, analysis and prediction on the sample. The processing method of the hyperspectral data preprocessing comprises one or any combination of filtering processing, baseline removing and normalization.

The established optimal model is divided into a training set (the training sample set establishes a plurality of training sample subsets according to the antibody secretion types of sample cells, and each training sample subset corresponds to an antibody secretion type label), a verification set and an inspection set by establishing a corresponding cell model database and adopting a uniform random sampling mode after the hyperspectral information of each cell is preprocessed; the method is used for training, verifying and checking the model respectively, and finally the optimal model is established, the model is verified by adopting a K-fold cross verification method, and K is any integer from 1 to 10.

The invention utilizes the microfluidic chip technology to complete the high-flux separation and the rapid hyperspectral imaging of single cells, utilizes a machine learning model to extract the spectral characteristics of the single cells and carries out the identification of comparison, classification and statistical analysis on the spectral characteristics, and has the following beneficial effects:

1. according to the invention, a new method for discovering the single-cell antibody is developed by introducing a hyperspectral imaging technology and a microfluidic technology, so that the capture, imaging and identification of single cells are realized, an idea is provided for screening the single cells, and antibody secretory cells can be obtained rapidly without damage.

2. The invention realizes the capture of single cells by a non-fluorescence labeling method, and can simultaneously screen a large number of single cells of cell groups in the fields of single cell analysis, especially antibody discovery. Compared with the existing method, firstly, the method provided by the invention is beneficial to reducing the reaction process by detecting the co-incubation one-step reaction of the antigen and the cell or directly performing hyperspectral imaging detection; secondly, the method can detect the absorption of cells in a bright field based on a hyperspectral imaging technology, reduce the interference and quenching of fluorescence and enable the identification result to be more accurate and reliable; thirdly, external forces such as an electric field, a magnetic field, sound waves, laser and the like are not needed, so that the damage to cells is effectively reduced; finally, the invention can realize the simultaneous detection of a plurality of antigens and realize the detection of any wave band in the detection range.

3. The method has high identification accuracy and high sensitivity, solves the problem of fluorescence quenching interference of fluorescence analysis, reduces complicated experimental steps, can simultaneously detect various detection antigens, and provides a technical scheme for screening broad-spectrum neutralizing antibodies and screening parallel multi-target antibodies. Therefore, the method has potential application value in antibody discovery, antibody drug development and cell analysis.

Drawings

FIG. 1 is a flow chart of a technical solution provided in embodiment 1 of the present invention;

fig. 2 is an overall structure diagram of a classification model provided in embodiment 1 of the present invention;

FIG. 3 is a graph of spectral information of anti-CD 45 protein antibody-secreting hybridoma cells as provided in example 2 of the present invention;

FIG. 4 is a graph of spectral information for the detection of mouse B lymphocytes provided in example 3 of the present invention;

FIG. 5 is a diagram of the experimental protocol for antibodies specific for SARS-CoV19-2 provided in example 4 of the present invention;

FIG. 6 is a diagram of the experimental protocol for the competitive antibody against SARS-CoV19-2 provided in example 5 of the present invention;

FIG. 7 is a diagram of the experimental protocol for neutralizing antibodies against SARS-CoV19-2 provided in example 6 of the present invention;

FIG. 8 is a diagram showing an experimental protocol of a method for detecting a broadly neutralizing monoclonal antibody according to example 7 of the present invention;

FIG. 9 is a diagram of an experimental scheme of a parallel multi-target monoclonal antibody detection method provided in example 8 of the present invention.

Detailed Description

The following examples apply the method for discovering monoclonal antibodies based on hyperspectral imaging technology to the screening of cells secreting antibodies. The specific method is to help those skilled in the art understand the function and application method of the present invention, and is not to limit the application scope of the device of the present invention.

Example 1

This example provides a method for detecting monoclonal antibodies, the flow chart is shown in fig. 1, and the specific steps include:

(1) preparation of cells

Training sample cells: live cells and/or known cell lines are obtained and prepared into single cell suspensions with buffer. The buffer solution is 0.85% NaCl physiological saline, phosphate buffer solution or other buffer solution suitable for the physiological concentration of the cells. Two processing methods are as follows: 1. adding one or more detection antigens, and incubating at 4 deg.C, 37 deg.C or room temperature for 1 hr; 2. single cell suspensions were prepared directly without addition of antigen. Then washing with buffer solution for 2 times, and re-suspending the cells with the cell culture solution to prepare a single cell suspension for later use. The cell culture solution is nutrient solution RPMI1640 suitable for the growth of the cells, DMEM or other nutrient solution suitable for the growth of the cells.

Detecting the sample cells: obtaining blood, tissue organ and the like of other animals or human such as immune mice, rabbits and the like, and separating B cells by using flow cytometry, magnetic bead separation, microfluidic technology, sound wave, electric field and other modes. Prepare single cell suspension with buffer. The buffer solution is 0.85% NaCl physiological saline, phosphate buffer solution or other buffer solution suitable for the physiological concentration of the cells. Two processing methods are as follows: 1. adding one or more detection antigens, and incubating at 4 deg.C, 37 deg.C or room temperature for 1 hr; 2. single cell suspensions were prepared directly without addition of antigen. Then washing with buffer solution for 2 times, and re-suspending the cells with the cell culture solution to prepare a single cell suspension for later use. The cell culture solution is nutrient solution RPMI1640 suitable for the growth of the cells, DMEM or other nutrient solution suitable for the growth of the cells.

The individual such as the immunized mouse can be immunized by one antigen or multiple antigens; or multiple mice immunized with different antigens.

The cell includes B cell of human and animal, mammal cell expressing antibody in vitro or host cell expressing antibody derivative after modification, bacteria expressing antibody in vitro, plant cell expressing antibody in vitro and virus, etc. and its number is at least greater than 1 cell.

The antibody derivative refers to one or more of antibody fragment, single-chain antibody, single-domain antibody and the like, and modified products thereof.

The human refers to normal humans, patients, vaccinators, autoimmune disease patients and other populations producing antibodies.

The animal refers to a mouse, a rat, a rabbit, a monkey, etc. and corresponding transgenes and/or gene editors; and/or animals immunized with the immunizing antigen.

The immune antigen refers to protein, polypeptide, nucleic acid, small molecule or corresponding conjugate, group modification and/or alteration substance.

The immunization refers to the delivery of an antigen into an animal by subcutaneous, intraperitoneal, intramuscular, lymph node, caudal vein, nasal or oral routes; the delivery refers to injection needles, electroporation and viral delivery means.

The detection antigen refers to the same protein, polypeptide and derivative, natural and recombinant expression product and the like as the immunogen and/or modified immunogen, and other substances used for researching substances combined with the antibody and/or substances not combined with the antibody; the number of detection antibodies is not limited, and is 1 or more.

The detectable agent or the agent different from the immunogen refers to a ligand, receptor, partial polypeptide sequence of antigen, competitive agent or other related agent which interacts with the antigen.

The modification refers to methods such as gene mutation, gene editing, chemical modification coupling, self-assembly and the like.

The gene mutation refers to the formation of mutants and derivatives by changing the gene information of proteins and the like.

The gene editing means a method of accessing gene information such as an altered protein, for example, splicing, insertion, substitution, etc.

(2) Capture of single cells

Adjusting the single cell suspension obtained in the step (1) to a proper concentration, then introducing the single cell suspension onto a microfluidic chip, fixing single cells at a specific position by using the structure of the chip or temperature-sensitive gel, and then cleaning redundant cells by using a cell culture solution to obtain the single cell array.

The material of the micro-fluidic chip is one or more of silicon-based material, glass material, organic polymer (PDMS, PMMA) and other nontoxic and harmless compatible materials.

The structure of the chip is a triangular capture structure consisting of a positive direction, a long direction, a circle, a V shape and a cylindrical array and other figures capable of capturing single cells.

The gel is temperature sensitive, the state of the gel can be changed at different temperatures, and the gel is changed from liquid to solid, so that the fixation of cells is realized; the release of the cells is achieved by changing from a solid state to a liquid.

(3) Hyperspectral imaging

Placing the chip containing the single cell array in the step (2) on image acquisition equipment, adjusting the intensity and the focal length of a light source, setting spectrum acquisition time and grating parameters, and carrying out optical imaging on cells on the chip to obtain the spectrum data of each single cell; the acquisition equipment for single cell hyperspectral imaging can be any hyperspectral equipment.

The image acquisition equipment is a common optical microscope, a fluorescence microscope or other imaging devices and is provided with an interface compatible with the hyperspectral imaging equipment; the optical imaging is an operation at any magnification that images at least one entire cell capture building block.

The hyperspectral equipment is any equipment with resolution below micron level.

The wavelength of the hyperspectral equipment is 400 nm-1000 nm or the wavelength in other ranges.

(4) Modeling sample analysis

And (3) constructing a sample database by using different types of cell sap such as known cell lines homologous with the cells to be detected, mouse B cells, human B cells and the like through the steps (1) to (3).

Grouping the constructed sample database by adopting a uniform random sampling method, and dividing the sample database into a training set and a verification set, wherein the training set is used for establishing a plurality of training sample subsets according to the antibody secretion types of cells, each training sample subset corresponds to a corresponding antibody secretion type label, and the verification set is uniformly and randomly sampled to form an inspection set; the training set, the verification set and the test set are respectively used for training, verifying and testing the model.

And (3) performing model training, verification and inspection by using the spectral data of the sample database as a training set, and establishing a machine learning optimization model.

The model comprises machine learning models such as a Bayesian classifier, a clustering algorithm, a Support Vector Machine (SVM), a deep learning network and the like, a K-fold cross verification method is adopted for verification of the model, and K is any integer of 1-10.

And (4) preprocessing the spectral data obtained in the step (3). And adopting one or any combination of smoothing, baseline removal and normalization. The smoothing method adopts a convolution translation method, a moving average method, Gaussian filtering, bilateral filtering or mean filtering and the like; the baseline removing method comprises polynomial fitting, a wavelet algorithm, a BEADS algorithm, empirical mode decomposition and the like. The normalization method adopts a maximum-minimum method, area normalization, vector normalization and other methods. And after pretreatment, carrying out sample identification prediction by using the established machine learning optimization model.

(4) Sample detection

And (3) preprocessing the detection sample cells in the step (1) to obtain hyperspectral information, inputting the hyperspectral information into the machine learning model established in the step (4), outputting antibody secretion type labels corresponding to the detection sample cells, and judging the antibody secretion types of the sample cells through the labels.

The machine learning model may be various types of models, and the identification of the sample cell can be achieved by selecting the optimal model, and the following model is exemplified as a model related to the present application:

the machine learning model is a convolutional neural network combined with Gabor filtering, the input of the network is the characteristics obtained after preprocessing, and the overall structure diagram of the classification model is shown in FIG. 2.

The size of the convolution kernel in each convolution layer is supposed to be 3 × 3, four branches are provided, each branch is provided with a modulated convolution kernel, and the definition is as follows:

wherein

in order to train fewer parameters and fit data by more parameters, only one group of convolution kernels is generated in each convolution layer, and then different modulation convolution kernels are obtained by modulating different Gabor kernels so as to relieve overfitting of the network, thereby improving the generalization capability of the network;

The model inputs the preprocessed features and outputs type labels which can be preset numbers, colors, figures, characters, letters and the like.

Example 2

In this example, the method of example 1 is applied to a hybridoma cell secreting anti-CD 45 antibody and a cell not secreting Sp2/0 antibody, and the specific application method is illustrated as follows:

all experimental steps are carried out in the cell-cell of the sterile environment, consumables and instruments are disinfected in advance, and reagents are sterile.

Collecting hybridoma cells and Sp2/0 cells, centrifuging at 1000rpm for 3min, resuspending the cells in RPMI1640 culture medium, respectively, to adjust the cell concentration to 10 per ml⁶And (4) cells. 100ul of each cell suspension was added to a 96-well plate and allowed to stand for 10 minutes. Spectral information of cells is captured under a microscope using hyperspectral imaging. 10 pictures were taken for each cell. Each cell is divided into a training set, a verification set and a check set, the cell secreting the anti-CD 45 antibody is marked as 1 in the training set, the cell not secreting the anti-CD 45 antibody is marked as 0, an optimal model is obtained after the model is trained and optimized, and the check set is analyzed. The results of the spectra for both cell types are shown in FIG. 3. The spectral information of the two cells is different, and the peak value of the hybridoma cell secreting the antibody is higher than that of the Sp2/0 cell not secreting the antibody. In this embodiment, the spectral information of the cell is further input into the optimal model for comparison and analysis, and the result is output: 1, representing the cell secreting CD45 antibody; and (4) outputting a result: 0, indicating that the cell does not secrete CD45 antibody. Spectral information of 9 cells was analyzed per group.

The results are as follows: 9 cells labeled 1; the 9 cell tags are 0. 5 cells of the two labels are respectively selected for culturing, and the antibody in the supernatant fluid is detected. The result of the antibody secreted by the cell is consistent with the predicted result, which shows that the accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100%; the overall accuracy of the machine learning model was 98.22%.

Example 3

In this example, the method of example 1 was used to screen the mouse lymphocytes for cells secreting PDL1, and the specific application method is illustrated as follows:

Two mice were taken, one normal and one immunized with the programmed death protein ligand 1(PD-L1) protein. Collecting spleen cells of mouse, aseptically grinding, cleaning with serum-free culture medium for 2 times, centrifuging at 1000rpm for 3min, and removing fat and other componentsThe adhesive of (3). B cells were isolated from spleen suspension using mouse B lymphocyte isolate. RPMI1640 medium resuspend B cells, adjust cells to 10 per ml⁶And (4) cells. 100ul of each cell suspension was added to a 96-well plate and allowed to stand for 10 minutes. The cell spectral information was captured under a microscope using hyperspectral imaging, 10 pictures were taken for each cell. Each cell is divided into a training set, a verification set and an inspection set, models are trained and optimized, an optimal model is selected, and the inspection set is analyzed. The results of the spectra of the two cells are shown in FIG. 4. The spectral information of the lymphocytes of the two mice is different, and the peak value of the lymphocytes of the mice secreting PDL1 is higher than that of normal mice. In this embodiment, the spectral information of 18 cells is further input into a machine learning model for comparison and analysis, and the result is output: 1, representing secretion of PDL1 antibody by the cell; and (4) outputting a result: 0, indicating that the cell did not secrete PDL1 antibody.

The results are as follows: 9 cells labeled 1; the 9 cell tags are 0. And respectively selecting 5 expression antibodies from the cells of the two labels and detecting the specificity, wherein the detection result is consistent with the prediction result output by the machine learning model. The accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100 percent; the overall accuracy of the machine learning model was 92.47%.

Example 4

This example uses the method of example 1 for the hybridoma cells secreting anti-SARS-CoV-2 specific antibody, and illustrates the specific application method as follows (see FIG. 5 in the flow chart):

The method for collecting the mouse B cells and establishing the model is as follows: the first step is as follows: isolating the lymphocytes; 1. whole blood from normal mice was collected and added to a centrifuge tube containing an anticoagulant, diluted 5-fold with RPMI1640 or other isotonic buffer, slowly added to an equal amount of lymphocyte separation solution, and subjected to density gradient centrifugation. 2. Collecting spleen, bone marrow and/or lymph node tissue of mouse, grinding with grinder to obtain cell suspension, adding erythrocyte lysate, reacting in ice bath for 15 min until erythrocyte is completely lysed, and centrifuging to collect lymphocyte. The second step is that: b cells are enriched; 1. incubating lymphocytes separated in the first step with anti-mouse IgG antibody, one or more B cell marker antibodies such as CD19, CD20, CD27 and the like, and separating B cell population by flow cytometry; 2. and (3) incubating magnetic beads marked by anti-mouse IgG antibody, CD19, CD20 and CD27 one or more B cell typing antibodies with the lymphocytes separated in the first step, and separating B cell populations by the action of a magnetic field. The B cells were incubated with SRAS-CoV19-2 RBD protein. Removing unbound RBD protein, and collecting cells. The cells were resuspended by adding the appropriate amount of buffer.

The third step: resuspend the cells in RPMI1640 medium, adjust the cell concentration to 10/ml⁶And (4) cells. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. Spectral information of cells is captured under a microscope using hyperspectral imaging. 10 pictures of each cell are taken as training samples and are divided into a training set, a verification set and an inspection set. And selecting the optimal model through training and optimizing the model, and then analyzing the inspection set.

Mice were immunized with SRAS-CoV19-2 RBD protein immunogen (RBD) and the serum antibody titers of the mice were measured. The SARS-CoV19-2 immunized mouse B cells were collected as above. RBD protein (Alexa Fluor647-RBD) modified with probe marker was incubated with B cells. Removing unbound RBD protein, and collecting cells. The cells were resuspended by adding the appropriate amount of buffer. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. Spectral information of cells is captured under a microscope using hyperspectral imaging. And inputting the spectral information into the established model, comparing and analyzing the spectral information, and outputting a classification result.

The method of this example detects 6400 cells, wherein the detection output of 152 cells is label 1 (secretion of RBD antibody); the detection output results of the other cells are label 0 (no antibody is secreted); selecting 10 cells with output labels of 1 to express the antibody in vitro and detect the specificity, wherein the detection result is consistent with the output result of the model, and the 10 antibodies are all specifically combined with the RBD protein, which indicates that the accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100 percent, and the overall accuracy of the machine learning model is 98.27 percent.

Example 5

This example uses the method of example 1 for hybridoma cells secreting anti-SARS-CoV 19-2 competitive antibody, and illustrates the specific application method as follows (see FIG. 6 in the flow chart):

The method for collecting the mouse B cells and establishing the model is as follows: the first step is as follows: isolating the lymphocytes; 1. whole blood from normal mice was collected and added to a centrifuge tube containing an anticoagulant, diluted 5-fold with RPMI1640 or other isotonic buffer, slowly added to an equal amount of lymphocyte separation solution, and subjected to density gradient centrifugation. 2. Collecting spleen, bone marrow and/or lymph node tissue of mouse, grinding with grinder to obtain cell suspension, adding erythrocyte lysate, reacting in ice bath for 15 min until erythrocyte is completely lysed, and centrifuging to collect lymphocyte. The second step is that: b cells are enriched; 1. incubating lymphocytes separated in the first step with anti-mouse IgG antibody, one or more B cell marker antibodies such as CD19, CD20, CD27 and the like, and separating B cell population by flow cytometry; 2. and (3) incubating magnetic beads marked by anti-mouse IgG antibody, CD19, CD20 and CD27 one or more B cell typing antibodies with the lymphocytes separated in the first step, and separating B cell populations by the action of a magnetic field. Re-suspending the RPMI1640 culture solution in B cell suspension, adding 2 or more than 2 substances or their derivatives in anti-SARS-CoV-2 RBD antigen, SARS-CoV-2 competitive product, etc., incubating, and removing unbound substances; resuspend into single cell suspension and adjust to appropriate concentration.

Thirdly, resuspending the cells in RPMI1640 medium to adjust the cell concentration to 10/ml⁶And (4) cells. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. Modifying immunogen, detecting source or substance to be detected at the bottom of the single cell array or in the device. After the single cells have reacted for a certain time in the microwells, competing substances, i.e. the above-mentioned antigens or derivatives, are added.The buffer removes unbound competitive material from the cells. Spectral information of cells is captured under a microscope using hyperspectral imaging. 10 pictures were taken for each cell. Each cell is divided into a training set, a verification set and an inspection set, models are trained and optimized, an optimal model is selected, and the inspection set is analyzed.

The SARS-CoV-2 immunized mouse B cells were collected as above. Single cell suspensions were prepared and adjusted to appropriate concentrations. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. The structure of the single cell micro array chip or other single cell structure device is modified with immunogen, detecting source or the detected matter. After the single cells have reacted for a certain time in the microwells, competing substances, i.e. the above-mentioned antigens or derivatives, are added. The buffer removes unbound competitive material from the cells. And comparing and analyzing the spectral information and outputting a classified prediction result based on the established model.

The method of this example performed detection on 6400 cells (96 well plates), where 12 cells output a label 1 (secreting competitive RBD antibody); the detection output of the rest cells is label 0 (secreting non-competitive RBD antibody); the 12 cells express the antibody in vitro and detect the specificity, the detection result is consistent with the output result of the model, and the 11 antibodies have competitiveness, which shows that the accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100 percent, and the overall precision of the machine learning model is 97.46 percent.

Example 6

This example uses the method of example 1 for hybridoma cells secreting neutralizing antibody against SARS-CoV-2 as an example, and illustrates the specific application method as follows (see FIG. 7 in the flow chart):

the method for collecting the mouse B cells and establishing the model is as follows: the first step is as follows: isolating the lymphocytes; 1. whole blood from normal mice was collected and added to a centrifuge tube containing an anticoagulant, diluted 5-fold with RPMI1640 or other isotonic buffer, slowly added to an equal amount of lymphocyte separation solution, and subjected to density gradient centrifugation. 2. Collecting spleen, bone marrow and/or lymph node tissue of mouse, grinding with grinder to obtain cell suspension, adding erythrocyte lysate, reacting in ice bath for 15 min until erythrocyte is completely lysed, and centrifuging to collect lymphocyte. The second step is that: b cells are enriched; 1. incubating lymphocytes separated in the first step with anti-mouse IgG antibody, one or more B cell marker antibodies such as CD19, CD20, CD27 and the like, and separating B cell population by flow cytometry; 2. and (3) incubating magnetic beads marked by anti-mouse IgG antibody, CD19, CD20 and CD27 one or more B cell typing antibodies with the lymphocytes separated in the first step, and separating B cell populations by the action of a magnetic field. 3. Cell suspensions secreting known antibodies were incubated with different probe-modified receptors, antigens, and then centrifuged to collect cells.

The third step: resuspend the cells in RPMI1640 medium, adjust the cell concentration to 10/ml⁶And (4) cells. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. The bottom of the single cell array or the inside of the device is modified with immunogen, detecting source, receptor or substance to be detected. After the single cells have reacted for a certain time in the microwells, the receptor ACE-I/antigen or other interacting substances are added. The buffer removes unbound material from the cells. Spectral information of cells is captured under a microscope using hyperspectral imaging. 10 pictures were taken for each cell. Each cell is divided into a training set, a verification set and an inspection set, models are trained and optimized, an optimal model is selected, and the inspection set is analyzed.

The SARS-CoV-2 immunized mouse B cells were collected as above. Single cell suspensions were prepared and adjusted to appropriate concentrations. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. The structure of the single cell micro array chip or other single cell structure device is modified with immunogen, detecting source or the detected matter. After the single cells have reacted for a certain time in the microwells, the receptor ACE-I/antigen or other interacting substances are added. The buffer removes unbound material from the cells. And comparing and analyzing the spectral information and outputting a classified prediction result based on the established model.

The method of the embodiment detects 3800 cells, wherein the detection output of 6 cells is label 1 (neutralizing RBD antibody); the detection output of the remaining cells is label 0 (no neutralizing RBD antibody). The result of the antibody verification of the in vitro expression of 6 cells is consistent with the result of the model output, and 6 antibodies can prevent the RBD from being combined with the receptor, which shows that the accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100 percent, and the overall precision of the machine learning model is 97.87 percent.

Example 7

In this example, the method of example 1 is applied to hybridoma cells secreting anti-SARS-CoV 19-2 mutant R1, R2, R3 broad-spectrum antibodies, and the specific application method is illustrated as follows (see the flow chart in fig. 8):

the method for collecting the mouse B cells and establishing the model is as follows: the first step is as follows: isolating the lymphocytes; 1. whole blood from normal mice was collected and added to a centrifuge tube containing an anticoagulant, diluted 5-fold with RPMI1640 or other isotonic buffer, slowly added to an equal amount of lymphocyte separation solution, and subjected to density gradient centrifugation. 2. Collecting spleen, bone marrow and/or lymph node tissue of mouse, grinding with grinder to obtain cell suspension, adding erythrocyte lysate, reacting in ice bath for 15 min until erythrocyte is completely lysed, and centrifuging to collect lymphocyte. The second step is that: b cells are enriched; 1. incubating lymphocytes separated in the first step with anti-mouse IgG antibody, one or more B cell marker antibodies such as CD19, CD20, CD27 and the like, and separating B cell population by flow cytometry; 2. and (3) incubating magnetic beads marked by anti-mouse IgG antibody, one or more B cell typing antibodies such as CD19, CD20, CD27 and the like with the lymphocytes separated in the first step, and separating B cell populations through the action of a magnetic field. 3. Incubating cell suspension secreting known antibody with receptor and antigen modified by different probes, adding 2 or more than 2 substances or derivatives thereof of RBD or R1, R2, R3 mutant, etc., incubating, centrifuging to remove unbound substances, and collecting cells; resuspend into single cell suspension and adjust to appropriate concentration.

The third step: resuspend the cells in RPMI1640 medium, adjust the cell concentration to 10/ml⁶And (4) cells. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. Spectral information of cells is captured under a microscope using hyperspectral imaging. 10 pictures were taken for each cell. Dividing each cell into training setThe method comprises the steps of verifying and checking a set, training and optimizing a model, selecting an optimal model, and analyzing the checking set.

The SARS-CoV-2 immunized mouse B cells were collected as above. Single cell suspensions were prepared and adjusted to appropriate concentrations. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. The structure of the single cell micro array chip or other single cell structure device is modified with immunogen, detecting source or the detected matter. Adding 2 or more than 2 of RBD protein or R1, R2, R3 mutant, etc., or their derivatives, and incubating. The buffer removes unbound material from the cells. And comparing and analyzing the spectral information and outputting a classified prediction result based on the established model.

The method of this example detects 2800 cells, wherein the detection output of 4 cells is tag 1; the detection output result of 4 cells is label 2; 3, the detection output result of the cell is a label 3; the detection output results of the other cells are label 0; the detection result after the antibody expression is consistent with the output result of the model, the label 0 is no antibody, and the antibody labeled 1 can neutralize R1; an antibody labeled 2 neutralizes R1, R2; an antibody labeled 3 neutralizes R1, R2, R3; the accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100%, and the overall accuracy of the machine learning model is 97.87%.

Example 8

This example uses the method of example 1 to target hybridoma secreting anti-SARS-CoV 19-2, H1N1, SARS parallel multi-target antibody, and illustrates the specific application method as follows (the flow chart is shown in FIG. 9):

the method for collecting the mouse B cells and establishing the model is as follows: the first step is as follows: isolating the lymphocytes; 1. whole blood from normal mice was collected and added to a centrifuge tube containing an anticoagulant, diluted 5-fold with RPMI1640 or other isotonic buffer, slowly added to an equal amount of lymphocyte separation solution, and subjected to density gradient centrifugation. 2. Collecting spleen, bone marrow and/or lymph node tissue of mouse, grinding with grinder to obtain cell suspension, adding erythrocyte lysate, reacting in ice bath for 15 min until erythrocyte is completely lysed, and centrifuging to collect lymphocyte. The second step is that: b fine enrichmentA cell; 1. incubating lymphocytes separated in the first step with anti-mouse IgG antibody, one or more B cell marker antibodies such as CD19, CD20, CD27 and the like, and separating B cell population by flow cytometry; 2. and (3) incubating magnetic beads marked by anti-mouse IgG antibody, one or more B cell typing antibodies such as CD19, CD20, CD27 and the like with the lymphocytes separated in the first step, and separating B cell populations through the action of a magnetic field. 3. Incubating cell suspension secreting known antibody with receptor and antigen modified by different probes to remove unbound substances; resuspend into single cell suspension and adjust to appropriate concentration. The third step: resuspend the cells in RPMI1640 medium, adjust the cell concentration to 10/ml⁶And (4) cells. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. SARS-CoV19-2, H1N1, SARS antigen was added to the microwell and bound to the secreted antibody. After washing the unbound detection substance, the spectral information of the cells is captured under a microscope using hyperspectral imaging. 10 pictures were taken for each cell. Each cell is divided into a training set, a verification set and an inspection set, models are trained and optimized, an optimal model is selected, and the inspection set is analyzed.

SARS-CoV19-2, H1N1, SARS-immunized mouse B cells were collected as above. The three cells are mixed together to prepare a single cell suspension, and the single cell suspension is adjusted to a proper concentration. The separation into single cell arrays/single cell droplets/or each cell was dispensed into 96-well plates and after resting for 10 minutes. Adding three or more substances for detecting antigen and receptor, reacting at proper temperature, and removing unbound substances with buffer solution. And comparing and analyzing the spectral information and outputting a classified prediction result based on the established model.

The method of the embodiment detects 8000 cells, wherein the detection output result of 182 cells is label 1; the detection output of 262 cells is tag 2; the detection output result of 218 cells is label 3; the detection output results of the other cells are label 0; cells of each tag type were selected in 5, expressed antibodies and tested for specificity. The detection result of the antibody is consistent with the output result of the model, the antibodies secreted by the cells labeled 1, 2 and 3 are respectively and specifically bound with CoV19, H1N1 and SARS protein, and the cell labeled 0 does not secrete the antibody. The accuracy of the method for detecting the monoclonal antibody provided by the invention reaches 100%, and the overall accuracy of the machine learning model is 98.35%.

Although the invention has been described in detail hereinabove with respect to a general description and specific embodiments thereof, it will be apparent to those skilled in the art that modifications or improvements may be made thereto based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.

Claims

1. A method for detecting a monoclonal antibody, comprising:

2. The method of claim 1, wherein the constructing of the machine learning model comprises the following steps:

3. The method according to claim 1 or 2, wherein the machine learning model is constructed by any one algorithm of a Bayesian classifier, a support vector machine and a deep learning network, and/or the machine learning model is preferably verified by a K-fold cross-validation method, wherein K is any integer from 1 to 10.

4. The method according to any one of claims 1 to 3, further comprising a hyperspectral data preprocessing process before the hyperspectral data is input into the machine learning model, wherein the hyperspectral data preprocessing process is one or more of filtering, baselining and normalization.

5. The method of any one of claims 1 to 4, wherein the test cells are single cell suspensions or are prepared as single cell suspensions after co-incubation with a detection substance.

6. The method according to claim 5, wherein the detection substance is any one or more of a protein, a nucleic acid, a small molecule, a fluorescent dye, or a derivative thereof; the derivative is preferably a product after modification or self-assembly.

7. The method according to any one of claims 1 to 6, wherein the test cells are selected from one or more of human cells, animal cells, plant cells, microbial cells or phages;

8. The method according to any one of claims 1 to 7, wherein the cells to be detected are transferred to a microfluidic chip to form a single cell array before the hyperspectral imaging detection; the material of the microfluidic chip is preferably one or more of a silicon wafer, a quartz material, a glass material, an organic polymer or a calcium fluoride material.

9. Use of the method of any one of claims 1-8 for screening cells for secretion of monoclonal antibodies; the cell is preferably one or more of a B cell, a hybridoma cell, a CHO cell, a yeast cell, a bacterial or a plant cell.

10. Use of the method of any one of claims 1 to 8 for the quantitative or qualitative detection of the composition and morphology of cells secreting any cytokines or other cellular secretions.