CN106548204A - The fast automatic grouping method of Flow cytometry data - Google Patents

Abstract

The present invention provides a method for fast and automatic grouping of flow cytometry data. The method includes the following steps: Step 1, using principal component analysis to process the lost cell data, including the following sub-steps: 1) Carrying out the sample matrix X Standardize to obtain the normalized matrix X ^* ; 2) Find the correlation coefficient matrix and perform eigendecomposition to obtain the eigenvalues (λ ₁ ≥λ ₂ ≥…≥λ _p ) and their corresponding eigenvectors a ₁ , a ₂ ,…, a _p ; 3) Determine the number k of principal components according to the variance contribution rate of the principal components; 4) According to the eigenvectors U=[λ ₁ ,λ ₂ ...λ _k ] corresponding to the first k principal components, obtain k pairs of sample data The eigenvector matrix W=X ^* U composed of principal component vectors; Step 2, use the improved K-means algorithm to cluster the flow cells to obtain the cluster labels; Step 3, set the principal component with the largest contribution rate as the coordinate axis to draw Scatter diagram; Step 4, realize automatic grouping.

Description

A fast and automatic clustering method for flow cytometry data

technical field

The invention relates to the field of biomedical detection, in particular to a method for quickly and automatically grouping flow cytometry data.

Background technique

Flow Cytometry has become the most important tool for biological research and clinical diagnosis. Flow Cytometry is a multi-parameter, rapid analysis or sorting of suspended cells or other particles. technology. Flow cytometry can detect a variety of physical and chemical properties of a single cell, and at the same time obtain scattered light signals (SC) representing cell volume and granularity and various fluorescent pulse signals (FL) representing the content of each antigen from the cell, and extract the signal Characteristic parameters such as peak value, pulse width and area. Scattered light and fluorescence signals induced by each cell are recorded as individual events, and all events are aggregated into complete flow data of the measured cell population.

Flow cytometry data analysis is one of the difficulties in flow cytometry, and its main purpose is to identify and divide subpopulation cells in the sample. When performing flow cytometry data analysis, the obtained data is usually visualized and analyzed using a two-dimensional scatter diagram that can display two measurement channel parameters, such as forward scattered light (SSC), side scattered light (FSC) ) or fluorescent signal. However, the two-dimensional scatter diagram can only analyze the parameters of two dimensions at a time. Since the multi-parameter streaming data has high dimensions and a large amount of data, if the number of streaming data parameters is n, two parameters are randomly selected as horizontal, The ordinate, the number of scatter plots that can be drawn is Usually, in a scatterplot drawn with randomly selected coordinate axis parameters, the distinction of cell subgroups is not obvious, requiring the operator to have a higher level of professional knowledge and select a specific combination of parameters for analysis to obtain a more ideal grouping As a result, the process is cumbersome and time-consuming.

Contents of the invention

In order to solve the above problems, the object of the present invention is to provide a method for fast and automatic grouping of flow cytometer data, said method comprising the following steps: Step 1, using principal component analysis to process the lost cell data, including the following sub-steps: 1) Standardize the sample matrix X to obtain the standardized matrix X ^* ; 2) Find the correlation coefficient matrix and perform eigendecomposition to obtain the eigenvalues (λ ₁ ≥ λ ₂ ≥...≥ λ _p ) and their corresponding eigenvectors a ₁ , a ₂ ,..., a _p ; 3) Determine the number k of principal components according to the variance contribution rate of the principal components; 4) According to the eigenvectors U=[λ ₁ , λ ₂ ...λ _k ] corresponding to the first k principal components , to obtain the eigenvector matrix W=X ^* U composed of sample data to k principal component vectors; Step 2, use the improved K-means algorithm to cluster the flow cells to obtain the group labels; Step 3, set the contribution rate to be the largest Draw a scatter diagram with the principal component as the coordinate axis; Step 4, realize automatic grouping.

Preferably, the step 2 specifically includes: determining a data point as the first initial cluster center, selecting the data point with the largest distance from the first cluster center as the second cluster center, and selecting the distance from the first two cluster centers The data point with the largest distance between the cluster centers is the third cluster center, and so on, and finally determine n initial cluster centers; finally, each data point is clustered in the initial cluster. The distance between centers is iteratively calculated to achieve clustering.

It is to be understood that both the foregoing general description and the following detailed description are exemplary illustrations and

explanation, and should not be used as a limitation to the claimed content of the present invention.

Description of drawings

With reference to the accompanying drawings, more objects, functions and advantages of the present invention will be clarified through the following description of the embodiments of the present invention, wherein:

Fig. 1 is the flowchart of the method for carrying out rapid automatic grouping to flow cytometry data of the present invention;

Fig. 2 is the result schematic diagram that utilizes traditional artificial grouping method to draw two-dimensional scatter diagram to obtain;

Fig. 3 is the contribution rate and the cumulative contribution rate of the principal component obtained after utilizing the PCA method of the present invention to process;

Fig. 4 is a schematic diagram of the grouping results obtained by using the method of the present invention.

detailed description

The objects and functions of the present invention and methods for achieving the objects and functions will be clarified by referring to the exemplary embodiments. However, the present invention is not limited to the exemplary embodiments disclosed below; it can be implemented in various forms. The essence of the description is only to help those skilled in the relevant art comprehensively understand the specific details of the present invention.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

The present invention proposes to apply Principal Component Analysis (PCA) to multi-parameter flow data analysis, by performing dimensionality reduction processing and sign extraction on flow data, using two principal component variables that can best reflect the differences between different subgroups of cells As the abscissa and ordinate axes of the two-dimensional scatter diagram, the samples were grouped and analyzed in the scatter diagram.

PCA is a commonly used multivariate statistical analysis technique. Based on the principle of variance maximization, PCA selects fewer important variables to replace the original multiple variables through linear transformation, reduces the data dimension and maximizes the effective amount of information saved in the data. The PCA algorithm first standardizes the sample matrix X to obtain the standardized matrix X ^* ; then finds its correlation coefficient matrix and performs eigendecomposition to obtain the eigenvalues (λ ₁ ≥ λ ₂ ≥…≥ λ _p ) and their corresponding eigenvectors a ₁ , a ₂ ,..., a _p ; Next, determine the number k of principal components according to the variance contribution rate of the principal components; finally, according to the eigenvectors U=[λ ₁ , λ ₂ ...λ _k ] corresponding to the first k principal components , to obtain the eigenvector matrix W=X ^* U composed of sample data pairs of k principal component vectors. Multi-parameter flow cytometry data has the characteristics of large data volume and high dimensionality. The PCA method can reduce the dimensionality and redundant information of flow cytometry data, select principal component variables as new feature variables, automatically set coordinate axes, and draw scatter diagrams. Realize automatic grouping.

K-means algorithm is a typical distance-based clustering algorithm, which is fast, simple and efficient. This method uses the improved K-means algorithm to realize the automatic gating of cells. The improvement of the algorithm is mainly reflected in the determination of the position of the initial clustering center. The traditional K-means clustering algorithm often randomly selects n values as the initial clustering center, resulting in unstable clustering results. This method is: first determine a data point as the first initial cluster center, then select the data point with the largest distance from the first cluster center as the second cluster center, and then select the distance from the first two cluster centers The data point with the largest distance is the third cluster center, and so on, and finally determine n initial cluster centers; finally, iteratively calculate the distance from each data point to the initial cluster center to achieve clustering.

The method provided by the present invention can realize the automatic grouping of flow cytometry data, without manually setting the coordinate axes of the scatter diagram, by automatically setting the first two or three principal components with the largest contribution rates obtained after processing as the coordinate axes, Automatic clustering of flow cytometry data can be realized. In addition, by using the improved Kmeans clustering algorithm to cluster and analyze the processed flow cytometry data, the classification labels of each event in the flow cytometry data are obtained, and the gates of different subgroups of cells are realized. Fig. 1 is a flow chart of the method for fast and automatic clustering of flow cytometry data according to the present invention. The automatic clustering results of this method are consistent with the traditional manual clustering results, and the analysis time is much lower than the manual analysis time, which improves the efficiency of cell clustering and improves the reliability of the clustering results. This method is used in multi-parameter flow cytometry data analysis. It has a good application prospect and can be applied to other biomedical data analysis fields. Figure 2 is a schematic diagram of the results obtained by drawing a two-dimensional scatter plot using the traditional manual clustering method. Fig. 3 is the contribution rate and cumulative contribution rate of the principal components obtained after processing by the PCA method of the present invention. Fig. 4 is a schematic diagram of the grouping results obtained by using the method of the present invention. From the comparison of Fig. 2 and Fig. 4, it can be seen that the grouping effect of the present invention is due to the manual grouping method of the transmission.

The flow cytometry data of human peripheral blood lymphocytes were used as the processing object, and the samples contained 4811 cells and 3 surface differentiation antigens (CD3+, CD19+ and CD56+) of lymphocytes. The flow data of each cell includes 11 parameters, namely pulse height (FITC-H, PE-H, APC-H), pulse area (FSC-A, SSC-A, FITC-A, PE-A, APC -A) and pulse width (FITC-W, PE-W, APC-W).

Table 1 The eigenvalues and eigenvectors of the principal components PC0 and PC1 with the largest contribution rate

Tab.1 Characteristic value and characteristic vector of PC1 and PC2

Table 2: Accuracy rate of PCA clustering results

Other embodiments of the invention will be apparent to and understood by those skilled in the art from consideration of the specification and practice of the invention disclosed herein. The description and examples are considered exemplary only, with the true scope and spirit of the invention defined by the claims.

Claims (2)

1. A method for fast automatic grouping of flow cytometry data, said method comprising the following steps: Step 1, using the principal component analysis method to process the lost cell data, including the following sub-steps: 1) Standardize the sample matrix X to obtain a standardized matrix X ^* ; 2) Calculate its correlation coefficient matrix and perform eigendecomposition to obtain eigenvalues (λ ₁ ≥λ ₂ ≥…≥λ _p ) and their corresponding eigenvectors a ₁ , a ₂ ,…,a _p ; 3) Determine the number k of principal components according to the variance contribution rate of the principal components; 4) According to the eigenvectors U=[λ ₁ , λ ₂ ... λ _k ] corresponding to the first k principal components, the eigenvector matrix W=X ^* U formed by the sample data pair k principal component vectors is obtained; Step 2, using the improved K-means algorithm to cluster the flow cells to obtain group labels; Step 3, set the principal component with the largest contribution rate as the coordinate axis to draw a scatter diagram; Step 4, realizing automatic grouping. 2. The method according to claim 1, said step 2 specifically comprising: determining a data point as the first initial cluster center, selecting the data point with the largest distance from the first cluster center as the second cluster Center, select the data point with the largest distance from the first two cluster centers as the third cluster center, and so on, finally determine n initial cluster centers; finally iterate the distance from each data point to the initial cluster center The operation implements clustering.