Automated analysis platform and method integrating multiple immunoassay technologies

Abstract

The invention relates to the technical field of immune detection and analysis, and discloses an automated analysis platform and method integrating multiple immune detection technologies. The method comprises receiving raw data from multiple immune detection technologies such as ELISA, flow cytometry and immunofluorescence, and extracting features; constructing a mutual information matrix between feature sets; judging whether feature dimensions meet preset requirements, and if not, dynamically adjusting a binarization threshold and repeating binarization processing and redundant feature elimination until the requirements are met; calculating an immune score based on the processed feature set, grading the immune level of samples according to the immune score, and identifying samples with abnormal immune function; the method integrates data from multiple immune detection technologies, and through the steps of feature extraction, redundant elimination and dimension control, realizes efficient and accurate identification of samples with abnormal immune function, and improves the comprehensiveness and reliability of immune function evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of immunoassay analysis, and more specifically, to an automated analysis platform and method integrating multiple immunoassay technologies. Background Art

[0002] In recent years, immunoassay technology has played an increasingly important role in fields such as medical diagnosis, drug screening, and disease mechanism research. However, traditional immunoassay methods have problems such as cumbersome operation, low analysis efficiency, and low result accuracy. To overcome these limitations, researchers have begun to explore automated and integrated immunoassay systems and methods.

[0003] Chinese Patent Application with Publication No. CN104714009A discloses an immunoassay detection system and detection method. This system uses two-dimensional code technology to identify detection items by reading the two-dimensional code on the detection card, and automatically sets a corresponding countdown reminder according to the detection items to indicate the reaction completion time of the detection card. This method improves the detection efficiency and accuracy to a certain extent. However, this patent mainly focuses on the automation of the detection process and does not involve the integration of multiple immunoassay technologies and improvements in data analysis.

[0004] Chinese Patent Application with Publication No. CN112964637A discloses an automated detection method for an immunoassay comprehensive system. This method is applied to a comprehensive system integrating functions such as two-dimensional code scanning, incubation processing, and fluorescence detection. By identifying the two-dimensional code of the test card to determine the test item, the hardware device is automatically adjusted to complete the incubation and detection of the sample, realizing the automation of the entire detection process. This integrated design shortens the process transfer distance and improves the detection efficiency and accuracy. However, although this patent realizes the series connection of multiple detection links, it does not involve how to deeply mine and comprehensively analyze the data obtained by different detection technologies.

[0005] In summary, existing automated immunoassay methods still lack effective methods for better integrating multiple immunoassay technologies and extracting valuable information from a large amount of heterogeneous data. Summary of the Invention

[0006] To overcome the above-mentioned defects of the prior art, the present invention provides an automated analysis platform and method integrating multiple immunoassay technologies. By extracting features from raw data from different sources and using the mutual information matrix to achieve feature selection and redundancy elimination, finally, based on the optimized feature set, an immune scoring model is constructed to identify samples with abnormal immune function. Compared with the prior art, the core of the present invention lies in realizing the deep integration and analysis of multi-platform immunoassay data, greatly improving the comprehensiveness and reliability of immune function evaluation.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] Automated analysis methods that integrate multiple immunoassay technologies, including:

[0009] Receiving raw data from sample detection using a plurality of immunoassay technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data, and immunofluorescence data; extracting features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set, respectively, to obtain a first feature set, a second feature set, and a third feature set;

[0010] Constructing mutual information matrices between the first feature set, the second feature set, and the third feature set, determining a binarization threshold, binarizing the mutual information matrix and eliminating redundant features to obtain a fourth feature set, a fifth feature set, and a sixth feature set;

[0011] Determine whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements. If so, identify samples with abnormal immune function based on the fourth feature set, the fifth feature set, and the sixth feature set; if not, dynamically adjust the binarization threshold to obtain an updated binarization threshold, and re-binarize the mutual information matrix and eliminate redundant features according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements.

[0012] Further, extracting features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set includes:

[0013] For ELISA data, extract the slope, intercept and goodness of fit R of the ELISA standard curve 2 As the first quantitative feature, the ratio of the sample absorbance to the negative control absorbance and the ratio of the sample absorbance to the positive control absorbance are extracted as the first qualitative feature to obtain a first feature set;

[0014] For the flow cytometry data, the mean fluorescence intensity MFI and cell ratio of each cell population are extracted as the second quantitative feature, and the peak type and peak position of the fluorescence signal distribution are extracted as the second qualitative feature to obtain the second feature set;

[0015] For immunofluorescence data, the mean and standard deviation of fluorescence intensity were extracted as the third quantitative feature, and the positive rate and colocalization coefficient of cell staining were extracted as the third qualitative feature to obtain the third feature set.

[0016] Furthermore, the mutual information matrix includes a first mutual information matrix M 12, the second mutual information matrix M 13 And the third mutual information matrix M 23 ;

[0017] The constructing of the mutual information matrix between the first feature set, the second feature set and the third feature set comprises:

[0018] Pair the features in the first feature set and the second feature set in pairs to construct a first feature pair set; pair the features in the first feature set and the third feature set in pairs to construct a second feature pair set; pair the features in the second feature set and the third feature set in pairs to construct a third feature pair set;

[0019] Calculate the mutual information between each pair of features in the first feature pair set, the second feature pair set, and the third feature pair set respectively;

[0020] According to the calculated mutual information, the first mutual information matrix M between the first feature set and the second feature set is constructed respectively. 12 , the second mutual information matrix M between the first feature set and the third feature set 13 , the third mutual information matrix M between the second feature set and the third feature set 23 .

[0021] Furthermore, the binarization threshold includes a first binarization threshold t 12 , the second binarization threshold t 13 and the third binarization threshold t 23 ;

[0022] The binarization processing of the mutual information matrix and the elimination of redundant features include:

[0023] M 12 Medium greater than t 12 Set the elements of 1 and the rest of the elements to 0 to get the first binarized matrix B 12 ;

[0024] M 13 Medium greater than t 13 Set the elements of 1 and the rest of the elements to 0 to get the second binarized matrix B 13 ;

[0025] M 23 Medium greater than t 23 Set the elements of 1 and the rest of the elements to 0 to get the third binarized matrix B 23 ;

[0026] To B 12 and B 13 Perform a logical OR operation to obtain the first redundant feature matrix R 1 ;

[0027] To B 12 and B 23 Perform a logical OR operation to obtain the second redundant feature matrix R 2 ;

[0028] To B 13 and B 23 Perform a logical OR operation to obtain the third redundant feature matrix R 3 ;

[0029] For the first redundant feature matrix R 1 , the second redundant feature matrix R 2 and the third redundant feature matrix R 3 Perform a logical OR operation to obtain the final redundant feature matrix R;

[0030] Traverse each element of the redundant feature matrix R, collect the row and column numbers corresponding to the elements with values ​​of 1, and obtain a set of redundant feature numbers;

[0031] According to the redundant feature number set, the features with corresponding numbers in the first feature set are removed;

[0032] According to the redundant feature number set, the features with corresponding numbers in the second feature set are removed;

[0033] According to the redundant feature number set, the features with corresponding numbers in the third feature set are removed.

[0034] Further, the determining whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet preset dimension requirements includes:

[0035] Calculate the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set respectively, denoted as dim 4 、dim 5 and dim 6 ;

[0036] Determine dim 4 、dim 5 and dim 6 Are they all less than or equal to the preset dimension threshold d? If so, it is considered that the feature dimension meets the preset dimension requirement; if not, it is considered that the feature dimension does not meet the preset dimension requirement;

[0037] The dynamically adjusting the binarization threshold comprises: setting the binarization threshold t 12 ,t 13 and t 23 Multiply them by the attenuation factor γ, 0<γ<1 respectively.

[0038] Furthermore, the identifying samples with abnormal immune function based on the fourth feature set, the fifth feature set and the sixth feature set includes:

[0039] Perform principal component analysis on the fourth feature set and extract the first principal component PC E1 , the second principal component PC E2 and the third principal component PC E3 , as the first immune score indicator;

[0040] Perform principal component analysis on the fifth feature set and extract the first principal component PC C1 , the second principal component PC C2 and the third principal component PC C3 , as the second immune score indicator;

[0041] Perform principal component analysis on the sixth feature set and extract the first principal component PC F1 , the second principal component PC F2 and the third principal component PC F3 , as the third immune score indicator;

[0042] According to the first immune score index, the second immune score index and the third immune score index, the immune score IS of the sample is calculated; the immune score IS is normalized to obtain a normalized immune score IS'; IS'∈[0,1];

[0043] According to the normalized immune score IS', the samples were graded for immune grade and samples with abnormal immune function were identified.

[0044] Furthermore, the immune grade of the samples is graded according to the normalized immune score IS', and the samples with abnormal immune function are identified, including:

[0045] Set the first immune threshold IMM 1 , the second immune threshold IMM 2 and the third immunity threshold IMM 3 Among them, IMM 3 >IMM 2 >IMM 1 ;

[0046] If 0≤IS'<IMM 1 , then the sample level is immunosuppression level;

[0047] If IMM 1 ≤IS'<IMM 2 , then the sample level is low immunity;

[0048] If IMM 2 ≤IS'<IMM 3, then the sample level is immune normal level;

[0049] If IMM 3 ≤IS'≤1, the sample level is immune enhancement level;

[0050] Samples at the immunosuppressed and immunocompromised levels were marked as immune dysfunction.

[0051] An automated analysis platform integrating multiple immunoassay technologies, which is used to implement the automated analysis method integrating multiple immunoassay technologies, comprises:

[0052] Feature extraction module: used to receive raw data from sample detection using multiple immune detection technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data and immunofluorescence data; extract features from the ELISA data, flow cytometry data and immunofluorescence data in the first data set respectively to obtain a first feature set, a second feature set and a third feature set;

[0053] Redundancy elimination module: used to construct mutual information matrices between the first feature set, the second feature set and the third feature set, determine the binarization threshold, perform binarization processing on the mutual information matrix and eliminate redundant features to obtain the fourth feature set, the fifth feature set and the sixth feature set;

[0054] Dimension judgment module: used to judge whether the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements. If so, samples with abnormal immune function are identified based on the fourth feature set, the fifth feature set and the sixth feature set; if not, the binarization threshold is dynamically adjusted to obtain an updated binarization threshold, and the mutual information matrix is ​​re-binarized and redundant features are eliminated according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements.

[0055] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned automated analysis method integrating multiple immunoassay technologies is implemented.

[0056] A computer-readable storage medium stores a computer program, which, when executed, implements the above-mentioned automated analysis method integrating multiple immunoassay technologies.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] The present invention integrates data from multiple immune detection technologies such as ELISA, flow cytometry and immunofluorescence, providing a more comprehensive perspective on immune function assessment. This multi-dimensional data integration can capture immune system features that may be overlooked by a single technology, greatly improving the accuracy and reliability of immune function assessment; targeted feature extraction is performed on different types of immune detection data, including quantitative features and qualitative features. This intelligent feature extraction method can maximize the retention of useful information in the original data, laying a solid foundation for subsequent analysis; by constructing a mutual information matrix and performing binarization processing, efficient redundant feature elimination is achieved, and this step significantly reduces the data dimensionality, reducing computational complexity while retaining the most recognizable feature information; introducing a mechanism for dynamically adjusting the binarization threshold to ensure that the dimension of the final feature set meets the preset requirements. This adaptive dimensionality control method not only ensures data streamlining, but also avoids information loss caused by excessive dimensionality reduction; a multi-level principal component analysis is performed on the processed feature set to extract key immune scoring indicators. This method effectively reduces the complexity of the data while retaining the most representative immune features; the entire analysis process is automated, which greatly improves analysis efficiency and reduces human errors. At the same time, the standardized analysis process ensures the consistency and repeatability of the results. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 The principle flow chart of the automated analysis method integrating multiple immunoassay technologies in the present invention;

[0061] Figure 2 A flow chart of a method for obtaining a first data set based on raw data in an automated analysis method integrating multiple immunoassay technologies of the present invention;

[0062] Figure 3 A flow chart of a method for extracting features from ELISA data, flow cytometry data and immunofluorescence data in a first data set in an automated analysis method integrating multiple immune detection technologies of the present invention;

[0063] Figure 4 A flow chart of a method for constructing a mutual information matrix between a first feature set, a second feature set and a third feature set in an automated analysis method integrating multiple immunoassay technologies of the present invention;

[0064] Figure 5 A flow chart of a method for binarizing a mutual information matrix in an automated analysis method integrating multiple immune detection technologies of the present invention;

[0065] Figure 6 A flow chart of a method for performing logical OR operations on multiple binary matrices in the automated analysis method integrating multiple immunoassay technologies of the present invention;

[0066] Figure 7 A flow chart of a method for eliminating redundant features in a first feature set, a second feature set, and a third feature set in an automated analysis method integrating multiple immunoassay technologies of the present invention;

[0067] Figure 8 A flow chart of a method for performing principal component analysis on the fourth feature set, the fifth feature set and the sixth feature set respectively in the automated analysis method integrating multiple immune detection technologies of the present invention, and extracting principal components as immune scoring indicators;

[0068] Fig. 9 This is a functional module diagram of the automated analysis platform that integrates multiple immunoassay technologies in the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] Example 1

[0071] See also Figure 1 As shown, this embodiment provides an automated analysis method integrating multiple immunoassay techniques, including:

[0072] Step S1000, receiving raw data from sample detection using multiple immunoassay technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data, and immunofluorescence data; extracting features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set, respectively, to obtain a first feature set, a second feature set, and a third feature set;

[0073] Furthermore, step S1000 includes:

[0074] Step S1100, receiving raw data from a sample detected by a plurality of immunoassay techniques, and obtaining a first data set according to the raw data; the first data set includes ELISA data, flow cytometry data and immunofluorescence data;

[0075] Furthermore, if Figure 2 As shown, step S1100 includes:

[0076] Step S1110, receiving raw data from sample detection using various immunoassay techniques such as ELISA, flow cytometry, and immunofluorescence to form a raw data set;

[0077] Step S1120, calculating the signal-to-noise ratio of each data in the original data set, marking the data with a signal-to-noise ratio lower than a preset signal-to-noise ratio threshold as noise data; using a local outlier factor (LOF) algorithm to perform outlier detection on the original data, marking the data with a LOF value exceeding a preset LOF threshold as an outlier;

[0078] Step S1130, remove noise data and outliers to obtain a first data set; and perform data normalization on the first data set to eliminate the influence of different data sources and dimensions.

[0079] Specifically, step S1100 improves the data quality by preprocessing the raw data, laying a solid foundation for subsequent analysis. A variety of immunoassay techniques are used to detect samples, and the samples are biological materials containing immune molecules or immune cells such as blood, serum, cells, and tissue sections. In the actual immunoassay process, due to the influence of factors such as the experimental environment, operators, and instruments and equipment, noise interference and outliers are inevitably present in the raw data. Noise data usually manifests itself as weak signal intensity and low signal-to-noise ratio, which will interfere with the accuracy of data analysis. Outliers refer to individual extreme values ​​that are significantly deviated from the center of the data set, which may be caused by sample contamination, detection errors, and other reasons. In order to ensure the reliability of data analysis, it is necessary to identify and eliminate noise and outliers before analysis. The signal-to-noise ratio is an important indicator for measuring data quality. By calculating the ratio of the signal intensity of the data to the noise level, the degree of noise interference of the data can be determined. It is generally believed that the higher the signal-to-noise ratio, the better the reliability of the data. In step S1120, by setting a signal-to-noise ratio threshold, data with a signal-to-noise ratio lower than the signal-to-noise ratio threshold is marked as noise data. The selection of the threshold needs to be empirically judged based on the characteristics of the detection technology and the data distribution. A too high signal-to-noise ratio threshold may mistakenly eliminate valid signals, while a too low signal-to-noise ratio threshold may retain too much noise. It is necessary to balance the noise reduction effect and information retention. Outlier detection can be judged based on the distribution of data in the feature space. The LOF algorithm calculates the degree of abnormality by measuring the local density of each data point relative to the surrounding neighborhood. Outliers usually show that the local density is significantly lower than the surrounding data points, so they have a higher LOF value. By setting the LOF threshold, abnormal cases that deviate from the normal data distribution can be identified. The selection of the LOF threshold also needs to weigh the detection rate and the false positive rate, and can be judged according to the significance level with the help of methods such as statistical hypothesis testing. After eliminating noise data and outliers, a first data set with higher quality is obtained. In order to eliminate the systematic deviation between different data sources, the first data set needs to be normalized and converted to a unified numerical scale. Commonly used data normalization methods include maximum and minimum value normalization, zero mean normalization, decimal calibration normalization, etc., which can be selected according to the distribution characteristics of the data and subsequent analysis requirements. Normalized data facilitates horizontal comparison and statistical modeling of different sources and batches, which helps to improve the generalization performance of the model.

[0080] Step S1100 obtains the first data set with reliable quality and unified format from the original massive heterogeneous data through systematic data preprocessing. This is not only a necessary prerequisite for ensuring the reliability of subsequent analysis, but also greatly reduces data redundancy and reduces the consumption of computing resources. At the same time, the data quality is characterized by quantitative indicators, which provides a basis for data management and traceability, and improves the controllability and repeatability of the detection process. The automatic identification and elimination of noise and outliers avoids the subjectivity and inefficiency of manual identification, and ensures the consistency and efficiency of judgment. The data normalization processing achieves the fusion integration of multi-source heterogeneous data, integrating the detection data of different fluxes and different batches into a unified analysis framework, which is convenient for unified modeling and comprehensive analysis, so as to dig out deeper biological mechanisms and laws.

[0081] Step S1200: extract features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set, respectively, to obtain a first feature set, a second feature set, and a third feature set.

[0082] Furthermore, if Figure 3 As shown, step S1200 includes:

[0083] Step S1210, extracting features from the ELISA data in the first data set, including: extracting the slope, intercept and goodness of fit R of the ELISA standard curve 2 As the first quantitative feature, the ratio of the sample absorbance to the negative control absorbance and the ratio of the sample absorbance to the positive control absorbance are extracted as the first qualitative feature to obtain a first feature set;

[0084] Step S1220, performing feature extraction on the flow cytometry data in the first data set, including: extracting the mean fluorescence intensity MFI and cell ratio of each cell population as the second quantitative feature, extracting the peak type and peak position of the fluorescence signal distribution as the second qualitative feature, and obtaining a second feature set;

[0085] Step S1230, performing feature extraction on the immunofluorescence data in the first data set, including: extracting the mean and standard deviation of the fluorescence intensity as the third quantitative feature, extracting the positive rate and colocalization coefficient of cell staining as the third qualitative feature, and obtaining a third feature set.

[0086] Specifically, step S1200 adopts a targeted feature extraction strategy for the data characteristics of different immune detection technologies to extract key information with biological significance and diagnostic value from the data. As a classic antibody quantitative detection method, the core of ELISA data analysis is to fit the standard curve of antigen-antibody reaction. By extracting parameters such as the slope, intercept, and goodness of fit of the standard curve, the corresponding relationship between the antibody concentration and the absorbance signal can be quantitatively reflected. In addition, indicators such as the ratio of sample absorbance to negative control absorbance and the ratio of sample absorbance to positive control absorbance can be used to qualitatively evaluate the degree of positivity of the sample and provide a basis for positive judgment. Flow cytometry can perform multi-parameter quantitative analysis of cell populations to obtain information such as the size, proportion, and phenotypic characteristics of each subpopulation. By extracting the mean fluorescence intensity (MFI) of each cell subpopulation, the difference in antigen expression in different populations can be compared; and the proportion of cells occupied by each subpopulation reflects the population size and immune cell composition. At the same time, the qualitative characteristics of flow cytometry data, such as the peak shape and peak position of the fluorescence signal distribution, can be used to determine the heterogeneity and activation state of the cell population, and provide a reference for cell function analysis. Immunofluorescence technology is used to locate and analyze the expression and distribution of antigens in tissue cells, and intuitively presents the spatial distribution pattern through fluorescent labeling. Statistics such as the mean and standard deviation of fluorescence intensity quantitatively describe the overall level and discreteness of the fluorescence signal, reflecting the level and uniformity of antigen expression. The positive cell rate qualitatively reflects the proportion of antigen positive expression and can be used for initial screening of positive samples. For the co-localization analysis of multiple antigen markers, the overlap coefficient of fluorescence signals in different channels can be calculated to determine the spatial interaction and co-localization relationship of antigen molecules, providing a basis for studying pathway regulation mechanisms.

[0087] Step S1200 extracts biologically significant qualitative and quantitative features from multi-source immune detection data, characterizes key attributes such as the intensity, specificity, and dynamic distribution of antibody-antigen reactions, and achieves a comprehensive characterization of the immune response process and state. Through feature engineering of technical data such as ELISA, flow cytometry, and immunofluorescence, it can not only reduce data redundancy and reduce analysis complexity, but also build a characteristic basis for subsequent diagnostic typing, prognosis prediction, and efficacy evaluation. At the same time, these immune features have important indicative significance for the body's immune function, help study the mechanism of immune system disorders, and have important guiding value for the clinical diagnosis and treatment of immune-related diseases and drug development. In addition, the feature integration of multi-platform detection data also lays a data foundation for subsequent comprehensive analysis and joint modeling. By integrating immune information of different dimensions and levels, it is expected to obtain a more accurate and comprehensive understanding of the immune status, providing new ideas for the development of precision medicine.

[0088] For example,

[0089] Suppose there is a topic to study autoimmune diseases, and it is necessary to analyze peripheral blood samples from 30 patients and 30 healthy controls. Each sample was tested by ELISA, flow cytometry and immunofluorescence.

[0090] Data preprocessing:

[0091] The ELISA experiment detected the concentrations of four cytokines, IL-2, IL-4, IL-6, and IFN-γ. Eight gradient standards were set for each factor, three replicate wells were set for each sample, and the experiment was repeated three times independently, so 4×8×3×3=288 raw data points were obtained for each sample.

[0092] Flow cytometry was used to analyze five lymphocyte subsets, including CD3+T cells, CD4+T cells, CD8+T cells, B cells, and NK cells. Two parameters, cell proportion and mean fluorescence intensity, were detected for each subset. Two replicate wells were set for each sample, and the experiment was repeated three times independently, so 5×2×2×3=60 raw data points were obtained for each sample.

[0093] The biopsy tissue sections were stained with immunofluorescence, and the two effector cytokines IL-17 and IFN-γ were detected. Three areas were randomly selected from each field of view, and the number of positive cells in each area was randomly counted. Three consecutive sections were taken for each sample, and the experiment was repeated three times independently. Therefore, 2×3×3×3=54 original data points were obtained for each sample.

[0094] Data points with a signal-to-noise ratio lower than 3 were removed as noise, and data points with a LOF value greater than 2 were removed as outliers using the LOF algorithm, and the maximum and minimum values ​​were normalized.

[0095] After preprocessing, a high-quality first data set was obtained, including 60 samples, each of which contained an average of 265 ELISA data, 56 flow cytometry data, and 50 immunofluorescence data.

[0096] Feature extraction:

[0097] For ELISA data, the slope, intercept and R of the standard curves of IL-2, IL-4, IL-6 and IFN-γ were extracted. 2 The values ​​were taken as 12 quantitative features, and the means and standard deviations of the concentration ratios of 60 samples to the negative and positive control concentrations were extracted as 8 qualitative features to obtain the first feature set, which contained 20 features.

[0098] For the flow cytometry data, the mean cell proportion and MFI of five subpopulations, including CD3+T cells, CD4+T cells, CD8+T cells, B cells, and NK cells, were extracted as 10 quantitative features, and the CV value, kurtosis, and skewness of the five subpopulations were extracted as 15 qualitative features to obtain the second feature set, which contained 25 features.

[0099] For immunofluorescence data, the mean and standard deviation of IL-17 and IFN-γ positive cell density were extracted as four quantitative features, and the mean of IL-17 / IFN-γ co-staining positive rate was extracted as one qualitative feature to obtain the third feature set, which contained five features.

[0100] Step S2000, constructing mutual information matrices between the first feature set, the second feature set, and the third feature set, determining a binarization threshold, binarizing the mutual information matrix and eliminating redundant features to obtain a fourth feature set, a fifth feature set, and a sixth feature set; judging whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements, and if so, identifying samples with abnormal immune function; if not, dynamically adjusting the binarization threshold to obtain an updated binarization threshold, and re-binarizing the mutual information matrix and eliminating redundant features according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements;

[0101] Furthermore, step S2000 includes:

[0102] Step S2100, constructing a mutual information matrix between the first feature set, the second feature set and the third feature set; the mutual information matrix includes a first mutual information matrix M 12 , the second mutual information matrix M 13 And the third mutual information matrix M 23 ;

[0103] Furthermore, if Figure 4 As shown, step S2100 includes:

[0104] Step S2110, pairing the features in the first feature set and the second feature set in pairs to construct a first feature pair set; pairing the features in the first feature set and the third feature set in pairs to construct a second feature pair set; pairing the features in the second feature set and the third feature set in pairs to construct a third feature pair set;

[0105] Step S2120, respectively calculating the mutual information between each pair of features in the first feature pair set, the second feature pair set, and the third feature pair set;

[0106] Step S2130: construct a first mutual information matrix M between the first feature set and the second feature set according to the calculated mutual information. 12 , the second mutual information matrix M between the first feature set and the third feature set 13 , the third mutual information matrix M between the second feature set and the third feature set 23 .

[0107] The method for calculating the mutual information between each pair of features includes:

[0108]

[0109] in:

[0110] I g (X; Y): represents the mutual information, X and Y represent two features in a pair of features respectively;

[0111] x i : represents the i-th value of feature X, which has a total of n values;

[0112] y j : represents the j-th value of feature Y, with a total of m values;

[0113] p(x i ): represents the marginal probability of feature X taking the i-th value;

[0114] p(y j ): represents the marginal probability of feature Y taking the jth value;

[0115] p(x i ,y j ): represents the joint probability that feature X takes the i-th value and feature Y takes the j-th value;

[0116] α: information entropy coefficient of feature X; reflects the discrete degree of feature value, 0<α≤1;

[0117] β: Information entropy coefficient of feature Y; reflects the discrete degree of feature value, 0<β≤1.

[0118] Parameter acquisition and calculation process:

[0119] 1. Calculate p(x i )、p(y j )、p(x i ,y j ); First, collect data points about feature pairs, build a frequency table for each pair of features, record the number of times each feature combination occurs, and the joint probability p(x i ,y j) is the frequency of feature X taking the i-th value and feature Y taking the j-th value occurring simultaneously divided by the total number of data points; the marginal probability p(x i ) is the frequency of occurrence of feature X taking the i-th value divided by the total number of data points; the marginal probability p(y j ) is calculated in the same way; then normalized to ensure that the sum of all probabilities is 1;

[0120] 2. Set the initial values ​​of α and β, for example, α = β = 1, which degenerates into the classic mutual information formula;

[0121] 3. The calculated p(x i )、p(y j )、p(x i ,y j ), as well as α and β, are substituted into the above mutual information formula to calculate I g (X; Y);

[0122] 4. The values ​​of α and β are set according to the discreteness of the features, and can be optimized through methods such as cross-validation; try different combinations of α and β, repeat step 3, and select α and β when the mutual information is the largest as the optimal parameters.

[0123] Analysis of the changing trend of mutual information:

[0124] When features X and Y are independent, the joint distribution p(x, y) = p(x)p(y) and the mutual information is 0.

[0125] When features X and Y are completely correlated, the joint distribution p(x, y) = p(x) = p(y) and the mutual information reaches its maximum value.

[0126] As the correlation between features X and Y increases, the mutual information increases monotonically.

[0127] The larger the values ​​of α and β, the more sensitive the mutual information is to the degree of feature discreteness. By properly selecting α and β, the mutual information can be accurately evaluated when the features are highly discrete or highly continuous.

[0128] The formula increases the flexibility of the formula by introducing information entropy coefficients α and β. When α=β=1, the formula degenerates into the classical mutual information, which is the generalized mutual information I g (X; Y). By adjusting α and β, we can adapt to the probability distribution of different types of features and improve the calculation accuracy of mutual information. When the feature discreteness is high, increasing the values ​​of α and β can suppress the noise effect caused by discrete values. When the feature discreteness is low, reducing the values ​​of α and β can enhance the difference between continuous values. Generalized mutual information I g(X; Y) can be regarded as the generalized correlation coefficient between features X and Y, reflecting the degree of their nonlinear correlation.

[0129] Step S2100 quantitatively evaluates the correlation and redundancy between different features by calculating the mutual information between the features of different immune detection technologies. Mutual information originates from information theory and is an indicator to measure the degree of mutual dependence between two random variables. In this step, the features of different detection data are paired in pairs to form multiple feature pair sets.

[0130] Exemplarily, the features are paired two by two:

[0131] Let the first feature set be denoted by A = {a 1 , a 2 , a 3}, the second feature set is recorded as B = {b 1 , b 2 , b 3 , b 4}, the third feature set is recorded as C = {c 1 , c 2}.

[0132] Then the feature pairing result of A and B is:

[0133]

[0134] Similarly, the feature pairing results of A and C, and B and C are:

[0135] {(a 1 , c 1 ), (a 1 , c 2 ), (a 2 , c 1 ), (a 2 , c 2 ), (a 3 , c 1 ), (a 3 , c 2 )};

[0136] {(b 1 , c 1 ), (b 1 , c 2 ), (b 2 , c 1 ), (b 2 , c 2 ), (b 3 , c 1 ), (b 3 , c 2 ), (b4 , c 1 ), (b 4 , c 2 )};

[0137] In this way, three feature pair sets are obtained, each of which reflects the feature combination between two detection techniques.

[0138] Mutual Information I g (X; Y) describes the correlation between features X and Y from the perspective of probability distribution. When features X and Y are independent, the mutual information is 0; when X and Y are completely correlated, the mutual information reaches its maximum value. Therefore, the larger the mutual information, the higher the redundancy between the two features. Based on the mutual information, the mutual information matrix between different feature sets can be constructed. The mutual information matrix M 12 、M 13 、M 23 The redundant relationship between the features of the three detection data of ELISA, flow cytometry, and immunofluorescence is characterized respectively. Through the mutual information matrix, the correlation between features from different sources can be fully characterized, providing a quantitative basis for identifying redundant features. 12 The rows of M correspond to the features of the first feature set, and the columns correspond to the features of the second feature set. 12 (i*, j*) represents the mutual information between the i*th feature in the first feature set and the j*th feature in the second feature set.

[0139] Construct the mutual information matrix based on the mutual information:

[0140] Assume that the mutual information between features has been calculated as follows:

[0141] I(a 1 ; b 1 )=0.5,I(a 1 ; b 2 )=0.3,I(a 1 ; b 3 )=0.6,I(a 1 ; b 4 )=0.1;

[0142] I(a 2 ; b 1 )=0.2,I(a 2 ; b 2 )=0.7,I(a 2 ; b 3 )=0.4,I(a 2 ; b 4 )=0.4;

[0143] I(a 3 ; b 1)=0.8,I(a 3 ; b 2 )=0.1,I(a 3 ; b 3 )=0.9,I(a 3 ; b 4 )=0.2;

[0144] I(a 1 ;c 1 )=0.4,I(a 1 ;c 2 )=0.6;

[0145] I(a 2 ;c 1 )=0.3,I(a 2 ;c 2 )=0.8;

[0146] I(a 3 ;c 1 )=0.5,I(a 3 ;c 2 )=0.7;

[0147] I(b 1 ;c 1 )=0.6,I(b 1 ;c 2 )=0.3;

[0148] I(b 2 ;c 1 )=0.2,I(b 2 ;c 2 )=0.9;

[0149] I(b 3 ;c 1 )=0.7,I(b 3 ;c 2 )=0.4;

[0150] I(b 4 ;c 1 )=0.1,I(b 4 ;c 2 )=0.5;

[0151] According to the calculated mutual information, the mutual information matrix is ​​constructed as follows:

[0152] The first mutual information matrix M 12 :

[0153]

[0154] The second mutual information matrix M 13 :

[0155]

[0156] The third mutual information matrix M 23 :

[0157]

[0158] By constructing these three mutual information matrices, the correlation between the features generated by different immunoassay techniques can be quantitatively evaluated.

[0159] Step S2100 uses mutual information, an information theory tool, to establish a quantitative framework for evaluating the redundancy of multi-source heterogeneous features. The traditional Pearson correlation coefficient is often used to characterize linear correlation, which is difficult to accurately reflect the relevant characteristics of nonlinear data. Mutual information can capture the nonlinear correlation between features and provide a more universal method for feature association analysis. By constructing feature pairs in pairs, the redundant relationship of features of different dimensions and levels can be systematically evaluated, and different data types such as discrete and continuous types can be flexibly processed. The mutual information matrix provides an intuitive method for visualizing redundant features. The redundancy of features can be judged by the value of matrix elements, and redundant features with less impact on classification and prediction can be identified. At the same time, the calculation of mutual information can also be used for feature selection. By selecting a feature subset with a small amount of mutual information with other features, a feature combination with minimum redundancy and maximum correlation can be constructed, which improves the robustness and generalization performance of subsequent modeling. In addition, the construction of the mutual information matrix also provides a new idea for the association analysis of multi-source data. By comparing the mutual information of features from different sources, the complementary and exclusive relationship between technologies can be revealed, the integration strategy of multi-omics data can be guided, and multi-level evidence can be provided for disease diagnosis. Therefore, mutual information and mutual information matrix have important application value in feature engineering and data fusion.

[0160] Step S2200, binarize the mutual information matrix, determine the binarization threshold, set the elements in the mutual information matrix that are greater than the binarization threshold to 1, and set the rest to 0, to obtain multiple binarized matrices; the binarization threshold includes a first binarization threshold t 12 , the second binarization threshold t 13 and the third binarization threshold t 23 ; The binarization matrix includes a first binarization matrix B 12 , the second binarization matrix B 13 And the third binarization matrix B 23 ;

[0161] Furthermore, if Figure 5 As shown, step S2200 includes:

[0162] Step S2210: 12 Medium greater than t12 Set the elements of 1 and the rest of the elements to 0 to get the first binarized matrix B 12 ;

[0163] Step S2220: M 13 Medium greater than t 13 Set the elements of 1 and the rest of the elements to 0 to get the second binarized matrix B 13 ;

[0164] Step S2230: M 23 Medium greater than t 23 Set the elements of 1 and the rest of the elements to 0 to get the third binarized matrix B 23 .

[0165] Specifically, step S2200 binarizes the mutual information matrix, converts quantitative analysis to qualitative analysis, and converts the correlation between features into two categories of strong correlation and weak correlation by setting a binarization threshold. Binarization limits the matrix elements to two values ​​of 0 and 1, simplifies the data structure of the matrix, and facilitates subsequent logical operations and association analysis. The selection of the binarization threshold is the key to binarization. If the binarization threshold is too high, some feature pairs with strong correlation will be missed, while if the threshold is too low, too many redundant features will be retained. The determination of the binarization threshold needs to comprehensively consider the distribution characteristics of the data, the number of selected features, the requirements of the correlation strength, and other factors, and is determined by empirical judgment or data-driven methods. In practical applications, different binarization thresholds can be set for different mutual information matrices according to specific analysis tasks and domain knowledge. For example, ELISA and flow cytometry are more widely used in quantitative analysis, and the binarization thresholds of their features can be set higher to extract strongly correlated features; while immunofluorescence focuses on qualitative analysis, and its binarization threshold can be appropriately relaxed. Flexible setting of thresholds can effectively balance the removal of redundant features and the retention of relevant features. 12 , B 13 and B 23 It intuitively depicts the strong and weak correlations between features from different sources, and provides a simplified data structure and judgment rules for subsequent redundant feature identification.

[0166] Step S2200 uses threshold binarization to convert the quantitative mutual information matrix into a qualitative binary matrix, thereby achieving dimensionality reduction and classification representation of feature correlation. By setting the binarization threshold, the continuous mutual information is mapped to two discrete categories of strong and weak correlation, highlighting the primary and secondary relationship of the correlation between features and reducing the complexity of quantitative analysis. Unifying the association analysis of multi-source data into the framework of the binary matrix can conveniently perform logical operations and judgment reasoning, simplifying the subsequent redundant feature identification process. At the same time, the binary matrix also provides convenience for the mining of association rules. Through Boolean operations, frequent co-occurring feature combinations can be quickly discovered, and the causal relationship and branch-leaf pattern between features can be mined, which is of great significance to the study of disease occurrence mechanisms. In addition, discretizing continuous variables also helps to deal with noise and outliers in the data and improve the robustness of data analysis. The binary matrix is ​​intuitive and easy to visualize, which is convenient for conveying feature correlation information to non-professionals. Experts can adjust the thresholds at different stages based on prior knowledge, balance the removal of redundant features and the retention of relevant features, and flexibly optimize the process of feature selection. Therefore, the binarization processing of mutual information matrix has broad application prospects in feature screening, association analysis and knowledge mining.

[0167] The method for determining the binarization threshold comprises:

[0168]

[0169] in:

[0170] t i′i′ : Mutual information matrix M between the i′th feature set and the j′th feature set i′j′ The binarization threshold, for example, t 12 represents the first binarization threshold;

[0171] μ i′j′ : Mutual information matrix M i′j′ The mean of all elements in ;

[0172] σ i′j′ : Mutual information matrix M ij The standard deviation of all elements in ;

[0173] k: Proportional coefficient of standard deviation, used to control the deviation of the binarization threshold from the mean, and is adjusted based on experience or cross-validation methods; k>0;

[0174] λ′: shape parameter, used to adjust the steepness of the binarization threshold. The larger the λ′, the steeper the t i′j′ In r i′j′ The smaller the value, the faster the change; λ′≥0; determined by technicians in this field through a large number of experiments;

[0175] r i′j′: The canonical correlation coefficient between the i′th feature set and the j′th feature set can be calculated by canonical correlation analysis (CCA), which is implemented using the CCA function in mathematical toolkits such as MATLAB and Python;

[0176] m i′ : The number of features in the i′th feature set;

[0177] m j′ : The number of features in the j′th feature set;

[0178] Feature number ratio;

[0179] α′: The proportional coefficient of the feature number ratio, which controls the influence of the feature number ratio on the binarization threshold. The larger α′ is, the greater the change in the threshold caused by a small change in the feature number ratio; the smaller α′ is, the smaller the influence of the feature number ratio on the threshold. By adjusting α′, the influence of the two factors of mutual information and feature number on the threshold can be weighed to adapt to different task requirements, which is determined by technicians in this field through a large number of experiments.

[0180] Mean μ i′j′ reflects the overall level of mutual information, and the standard deviation σ i′j′ It reflects the degree of discreteness of the mutual information. When the mutual information is generally large, the threshold is increased accordingly to screen out stronger correlations; when the mutual information varies greatly, the adjustment range of the threshold is expanded accordingly to adapt to correlations of different strengths.

[0181] Canonical correlation coefficient r i′j′ It measures the linear correlation between two feature sets and reflects their similarity as a whole. i′j′ The larger the value is, the stronger the correlation between the two feature sets is. At this time, the deviation of the threshold from the mean should be greater to obtain a more distinct combination of related features. It is a random i′j′ A monotonically increasing function, r i′j′ The value of is mapped to the interval (0, 1). i′j′ When it is small, the function is close to 0, and the threshold is mainly determined by the mean μ i′j′ decision; with r i′j′ As increases, the function gradually approaches 1, and the deviation of the threshold from the mean increases continuously to obtain stronger feature correlation.

[0182] The proportional coefficient k controls the degree of deviation of the threshold from the mean. A larger k value makes the threshold deviate more from the mean, and the selected feature combinations are more correlated; a smaller k value makes the threshold close to the mean, and more feature combinations are obtained.

[0183] The shape parameter λ′ affects the steepness of the threshold function. A larger λ′ value makes the threshold function i′j′ When it is small, the change is drastic and approaches the mean more quickly, while when it is small, the change is drastic and approaches the mean more quickly. i′j′ When it is larger, the change is gentle and the deviation from the mean is greater. This helps to distinguish strong correlation from weak correlation characteristics.

[0184] The ratio of the number of features Measures the degree of balance between two feature sets in terms of the number of features. When the number of features of the two sets is close, their ratio is close to 1, indicating that they have similar coverage in the feature space; when the number of features of the two sets is very different, their ratio is close to 0.

[0185] The proportional coefficient α′ controls the strength of the influence of the feature number ratio on the threshold. The larger α′ is, the greater the change in the threshold caused by a small change in the feature number ratio; the smaller α′ is, the smaller the influence of the feature number ratio on the threshold. By adjusting α′, the influence of the two factors of mutual information and feature number on the threshold can be weighed to meet different task requirements.

[0186] In summary, this formula makes full use of the statistical characteristics of the mutual information matrix and the overall correlation of the feature set, and can adaptively determine the binarization threshold. By adjusting k and λ', the screening strength of the threshold can be flexibly controlled to obtain the required feature correlation combination. At the same time, this formula avoids the subjectivity of artificial empirical threshold selection, and provides a data-driven threshold determination method, which can improve the accuracy and efficiency of feature selection.

[0187] The adaptive threshold obtained by this formula can be dynamically determined according to the distribution characteristics and redundancy of the feature set. When the feature set is more redundant or has a large correlation, this method can automatically increase the threshold, thereby eliminating a large number of redundant features and obtaining the most critical feature combination; when the feature set is less correlated or has a small number of features, this method can lower the threshold to obtain more meaningful feature combinations and avoid information loss due to over-screening. Compared with the fixed threshold, this adaptive threshold enables the algorithm to have better adaptability and robustness to different data. In addition, for immunogenomic data with higher dimensions, this method can effectively reduce the dimension, alleviate the problem of dimensionality disaster, reduce computational overhead, and facilitate subsequent modeling and analysis. In summary, this adaptive threshold formula combined with mutual information and canonical correlation analysis can efficiently screen out the correlation characteristics between different immune detection technologies, optimize the fusion representation of multi-source data, and play a positive role in improving the subsequent predictive analysis performance.

[0188] Step S2300, performing a logical OR operation on multiple binary matrices to obtain a redundant feature matrix R;

[0189] Furthermore, if Figure 6 As shown, step S2300 includes:

[0190] Step S2310: Binarize the matrix B 12 and B 13 Perform a logical OR operation to obtain the first redundant feature matrix R 1 ;

[0191] Step S2320, binarize the matrix B 12 and B 23 Perform a logical OR operation to obtain the second redundant feature matrix R 2 ;

[0192] Step S2330: Binarize the matrix B 13 and B 23 Perform a logical OR operation to obtain the third redundant feature matrix R 3 ;

[0193] Step S2340: the first redundant feature matrix R 1 , the second redundant feature matrix R 2 and the third redundant feature matrix R 3 Perform a logical OR operation to obtain the final redundant feature matrix R.

[0194] Specifically, step S2300 comprehensively determines the redundant features in the multi-source data by performing a logical OR operation on the binary matrices. In this step, firstly, a pairwise logical OR operation is performed on the binary matrices of ELISA and flow cytometry, ELISA and immunofluorescence, and flow cytometry and immunofluorescence to obtain three redundant feature matrices R 1 , R 2 and R 3 The logical OR operation rule is: as long as B 12 , B 13 , B 23 If one of the feature pairs in is 1, the corresponding redundant feature matrix element is 1. This means that no matter which two features of the detection data are, as long as one pair of features is highly correlated, they can be determined as redundant features. 1 , R 2 and R 3 Perform a logical OR operation to obtain the final redundant feature matrix R. If a feature is in the three redundant matrices R 1 , R 2 , R 3If the features are highly correlated with other features, it can be determined that they are redundant in multi-platform data and have little contribution to the discrimination task. The elements with a value of 1 in the matrix R represent redundant features that appear repeatedly in multi-source data and have little influence. Through layer-by-layer logical OR operations, the correlation of features is comprehensively evaluated in multiple dimensions, and different types of redundant features are fully identified, which can maximize the removal of information overlap between data.

[0195] Step S2300 uses logical OR operation to aggregate and judge the information of binary matrices from different sources, overcomes the limitations of single matrix judgment, and identifies redundant features from a global perspective. Traditional feature selection usually targets a single data source and is difficult to reveal the intrinsic connection between features at different levels. Feature screening of multi-source data requires comprehensive consideration of the complementarity and consistency of different test results. By comparing matrices from different sources in pairs, the differences in feature correlations between different technologies can be revealed, and redundant rules with universal significance can be found. By summarizing these local correlation information into a unified redundant matrix, redundant features with indicative significance for different technologies can be obtained. At the same time, the design idea of ​​multi-level logical OR operation draws on the basic strategy of integrated learning, and obtains more robust and accurate results by combining the judgments of multiple judges, reducing the risk of misjudgment and missed judgment. In addition, logical OR operation is simple and effective, with high computational efficiency, and is suitable for processing large-scale feature combination problems. The redundant feature representation in matrix form also provides convenience for parallel computing and distributed storage, and has good scalability. Therefore, the logical OR operation of binary matrix is ​​an effective multi-source feature redundancy identification method, which has broad application prospects in multi-omics data mining.

[0196] Step S2400, according to the redundant feature matrix R, redundant features in the first feature set, the second feature set and the third feature set are eliminated to obtain a fourth feature set, a fifth feature set and a sixth feature set;

[0197] Furthermore, if Figure 7 As shown, step S2400 includes:

[0198] Step S2410, traverse each element of the redundant feature matrix R, collect the row and column numbers corresponding to the elements with values ​​of 1, and obtain a redundant feature number set;

[0199] Step S2420, according to the redundant feature number set, remove the features with corresponding numbers in the first feature set to obtain a fourth feature set;

[0200] Step S2430, according to the redundant feature number set, remove the features with corresponding numbers in the second feature set to obtain a fifth feature set;

[0201] Step S2440: According to the redundant feature number set, the features with corresponding numbers in the third feature set are removed to obtain a sixth feature set.

[0202] Specifically, step S2400 removes redundancy from the original multi-source feature set according to the information of the redundant feature matrix, and obtains an optimized feature set with the largest amount of information and the lowest redundancy. First, by traversing the elements of the redundant feature matrix, the row and column numbers where the value is 1 are collected to form a set, and the number of the redundant features in the original feature set is recorded. Since the row and column order of the mutual information matrix corresponds to the original feature set one by one, the redundant features that need to be removed can be accurately located by using the corresponding relationship of the number. Then, the three feature sets are respectively subjected to redundancy removal. For the first feature set, the features in the redundant number set are removed from the set to obtain the fourth feature set after removing the redundancy. Similarly, the same removal operation is performed on the second feature set and the third feature set to obtain the fifth feature set and the sixth feature set. Through this series of redundant cleaning operations, a feature set with strong complementarity and high discrimination is finally obtained, the information overlap between the data is removed, the scale of the feature set is reduced, and computing resources are saved for subsequent modeling analysis. At the same time, redundant elimination of feature sets can also help improve the interpretability of analysis results, highlight key biological features, and reduce interference from irrelevant information.

[0203] Step S2400 reduces the dimension of the original feature space by removing the redundant feature matrix, and obtains the optimal feature combination across platforms. Different immune detection technologies reflect the immune response state of the body from different levels, but there is inevitably information overlap. The purpose of feature optimization is to minimize the redundancy of the feature set while maintaining the discriminant performance. Eliminating redundancy is a combinatorial optimization process that requires weighing the relevance and complementarity of features. Through the regular mapping of the redundant matrix, the corresponding relationship between different sets is established at the level of feature numbering. Through the feature deletion operation at the corresponding position, the collaborative optimization of multi-source data is achieved, avoiding the inefficiency of multiple traversals and comparisons. At the same time, the strategy of retaining non-redundant features maintains the discriminant information of the original data to the maximum extent, and overcomes the information loss problem of traditional dimensionality reduction methods. In addition, feature selection through set operations also provides a concise logical framework for the search of feature subspaces, facilitates the design of efficient pruning strategies and evaluation functions, and flexibly weighs multiple goals such as the number of features and mutual information. It is a feature selection paradigm that takes into account both optimization efficiency and generalization performance. Therefore, the feature elimination strategy based on redundant matrix has wide application value in multi-source data fusion.

[0204] Step S2500, determine whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements; if not, dynamically adjust the binarization threshold to obtain an updated binarization threshold, and re-perform binarization processing, logical OR operation, and redundant feature elimination according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements; if so, execute step S3000.

[0205] Further, step S2500 includes:

[0206] Step S2510, respectively calculate the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set, denoted as dim 4 、dim 5 and dim 6 ;

[0207] Step S2520, determine dim 4 、dim 5 and dim 6 Are they all less than or equal to the preset dimension threshold d? If so, it is considered that the feature dimension meets the preset dimension requirement, and step S3000 is executed; if not, it is considered that the feature dimension does not meet the preset dimension requirement, and step S2530 is executed;

[0208] Step S2530: transform the mutual information matrix M 12 、M 13 and M 23 The binarization threshold t 12 ,t 13 and t 23 Multiply them by the attenuation factor γ to get the updated binarization threshold t 12 '、t 13 ' and t 23 ';0<γ<1;

[0209] Step S2540, using the updated binarization threshold t 12 '、t 13 ' and t 23 ', repeat steps S2200-S2400 until dim 4 、dim 5 and dim 6 are all less than or equal to d, and finally the fourth feature set, the fifth feature set and the sixth feature set that meet the dimensionality requirements are obtained.

[0210] Specifically, step S2500 performs dimensional evaluation on the feature set obtained in the previous step by setting the expected dimensionality threshold d to determine whether the effect of redundant feature elimination is as expected. The value of the dimensionality threshold d needs to be weighed according to factors such as the complexity of the actual problem, the sample size, and the interpretability. Usually, the value of d is between 10% and 50% of the original feature dimension. If the dimension of the feature set is still higher than the threshold d after binary feature selection, it means that the redundant features are not completely eliminated and the binary threshold needs to be further tightened. Step S2530 introduces the attenuation factor γ to proportionally reduce the binary threshold, where 0<γ<1. The value of γ determines the speed of threshold attenuation. The smaller γ is, the faster the threshold decays and the more aggressive the redundant feature elimination is; the closer γ is to 1, the slower the threshold decays and the more cautious the redundant feature elimination is. γ can be empirically selected based on data distribution and domain knowledge, usually taking γ=0.8~0.95. Using the updated binarization threshold, repeat the mutual information calculation, threshold comparison and binarization processing until the dimensions of all feature subsets meet the preset requirements. This adaptive threshold adjustment strategy can maximize the elimination of redundant features while avoiding information loss, ensuring the simplicity of feature combinations while taking into account the accuracy of classification predictions. At the same time, by setting the stop condition, the infinite loop of threshold adjustment is avoided and the computational efficiency is improved. It is worth noting that in practical applications, the dimension threshold d and the attenuation factor γ can be adaptively optimized by introducing penalty terms and adding prior knowledge, so that they can be automatically adjusted according to different data characteristics, thereby further improving the intelligent level of feature selection.

[0211] Step S3000: Identify samples with abnormal immune function based on the fourth feature set, the fifth feature set and the sixth feature set.

[0212] Furthermore, step S3000 includes:

[0213] Step S3100, performing principal component analysis on the fourth feature set, the fifth feature set and the sixth feature set respectively, and extracting principal components as immune scoring indicators; the immune scoring indicators include a first immune scoring indicator, a second immune scoring indicator and a third immune scoring indicator;

[0214] Furthermore, if Figure 8 As shown, step S3100 includes:

[0215] Step S3110, perform principal component analysis on the fourth feature set to extract the first principal component PC E1 , the second principal component PC E2 and the third principal component PC E3 , as the first immune score indicator;

[0216] Step S3120, perform principal component analysis on the fifth feature set to extract the first principal component PC C1 , the second principal component PC C2 and the third principal component PC C3 , as the second immune score indicator;

[0217] Step S3130, perform principal component analysis on the sixth feature set to extract the first principal component PC F1 , the second principal component PC F2 and the third principal component PC F3 , as the third immune score indicator.

[0218] Specifically, step S3100 uses the principal component analysis (PCA) method to reduce the dimension of multidimensional feature data and extract the main variation information of the data. PCA linearly transforms the original feature space to a new feature space through orthogonal transformation, so that the features of the new space (i.e., the principal components) are statistically independent of each other and are sorted according to the size of the variance. The first few principal components can often explain most of the data variation, so they can be used as low-dimensional representations of the original high-dimensional data. In this step, PCA is performed on the feature data of the three detection technologies respectively, and the first three principal components of each are extracted as immune scoring indicators. These three principal components occupy the largest variance ratio in their respective data sets and contain the most critical immune feature information. Through principal component extraction, not only the computational complexity of subsequent modeling is reduced, but also it helps to eliminate multicollinearity between features and improve the stability and generalization ability of the model. In addition, since the principal component has good interpretability, it is easier to use and promote it as an immune scoring indicator in clinical practice. Medical staff can intuitively understand and monitor the changes in the principal component scores, evaluate the dynamic evolution of the patient's immune function, and provide a basis for the optimization of immunotherapy regimens. The introduction of the principal component immune score indicator provides a new idea for the integrated application of multi-platform detection data. By mining the common characteristics of data from different sources, a comprehensive evaluation of multi-dimensional immune information is achieved, which is of great significance for a comprehensive understanding of the body's immune status.

[0219] Step S3200, calculating the immune score IS of the sample according to the first immune score indicator, the second immune score indicator and the third immune score indicator; normalizing the immune score IS to obtain a normalized immune score IS'; IS'∈[0,1];

[0220] IS=λ 1 ×(ω 11 ×PC E1 +ω 12 ×PC E2 +ω 13 ×PC E3 )+λ 2 ×(ω 21×PC C1 +ω 22 ×PC C2 +ω 23 ×PC C3 )+λ 3 ×(ω 31 ×PC F1 +ω 32 ×PC F2 +ω 33 ×PC F3 );

[0221] Among them, IS represents the immune score of the sample, and the larger the value, the stronger the immune function; λ 1 ,λ 2 ,λ 3 are the weight coefficients of the first immune score indicator, the second immune score indicator, and the third immune score indicator, respectively, reflecting the contribution of different detection technologies to the immune score. These three weight coefficients can be learned from historical data through expert experience or machine learning algorithms and satisfy λ 1 +λ 2 +λ 3 =1.

[0222] ω 11 ,ω 12 ,ω 13 ,ω 21 ,ω 22 ,ω 23 ,ω 31 ,ω 32 ,ω 33 is the weight parameter of each principal component, which controls the influence of the principal component on the immune score. It can be determined by statistical methods such as variance contribution rate, or the optimal weight combination can be obtained through optimization algorithm training. These 9 weight parameters also meet the normalization condition: ω 11 +ω 12 +ω 13 =1,ω 21 +ω 22 +ω 23 =1,ω 31 +ω 32 +ω 33 =1.

[0223] When the immune level of a sample changes, it will be reflected in the test data of ELISA, flow cytometry, and immunofluorescence, causing the extracted principal components to increase or decrease accordingly. According to the changing trend of the principal component value and the principal component weight, the immune score IS will eventually present a numerical performance consistent with the strength of the sample's immune function, objectively quantifying the overall immune status reflected by the multi-source test data. Directly using the original high-dimensional test data for immune assessment is not only computationally complex, but may also be interfered by data redundancy and noise. The principal component analysis method can reduce the data dimension while retaining data information to the greatest extent, effectively reducing the redundancy of the original data, and using a few principal components to reflect the main characteristics and main change trends of the immune function. A variety of immune detection technologies can reflect the immune function of samples from different aspects, but a single technology often has problems such as insufficient information and large errors, and it is necessary to integrate multi-source data to comprehensively and accurately evaluate the immune status. This formula integrates the information of multiple immune detection methods, comprehensively evaluates the immune function of samples, and makes up for the limitations of a single technology.

[0224] The introduction of weight coefficients not only takes into account the differences in the importance of different detection technologies, but also takes into account the contribution of each principal component to the immune status, making the score more accurate and reliable. Different detection technologies and extracted principal components have different responsiveness to the immune function of samples. The weighted average strategy can reasonably regulate the importance of each part in the immune assessment. The weight is consistent with the corresponding test data and the strength of the immune indication of the principal component. When the immune function of the sample changes, the test data will inevitably change accordingly, which will lead to changes in the extracted principal component value, and the weight ratio may also be adjusted accordingly, ultimately making the immune score and the immune level show a consistent trend.

[0225] Integrating multi-source data in the form of weighted summation is simple and easy to calculate, and is convenient for promotion in practical applications. The main components of the three techniques of ELISA, flow cytometry, and immunofluorescence are multiplied by the weight coefficients respectively, and then added up. The obtained immune score summarizes the immune function information provided by multi-source data, and can objectively and quantitatively reflect the overall immune status of the sample. Quantifying complex immune functions as a scoring indicator is intuitive and clear, which is conducive to the rapid judgment and longitudinal comparison of immune status. The immune score is positively correlated with the immune level of the sample. When the immune function improves or deteriorates, the score will respond promptly and sensitively. Using the immune score as a quantitative indicator of the sample's immune function can intuitively judge the quality of the immune status and monitor the dynamic changes of the immune level. It has broad application prospects in disease diagnosis, efficacy evaluation, and drug screening.

[0226] Normalizing the immune scores of samples can eliminate the dimensional effect of the scores, making the immune scores of different batches and scales comparable. The normalized immune score is monotonically increasing with the relative immune level. The larger the value, the better the comprehensive immune status of the sample, and vice versa.

[0227] In summary, this formula integrates the data information provided by various immune detection technologies through a calculation method that combines principal component analysis and weighted average, overcomes the shortcomings of a single detection method, maximizes the immune function characteristics contained in the data, and can evaluate the immune status of samples more comprehensively, accurately and quantitatively, providing a powerful tool for immune-related research and applications.

[0228] Step S3300: According to the normalized immune score IS', the samples are graded according to their immune levels to identify samples with abnormal immune function.

[0229] Further, step S3300 includes:

[0230] Step S3310, setting the first immune threshold IMM 1 , the second immune threshold IMM 2 and the third immunity threshold IMM 3 Among them, IMM 3 >IMM 2 >IMM 1 ;

[0231] Step S3320, if 0≤IS'<IMM 1 , then the sample level is immunosuppression level;

[0232] Step S3330, if IMM 1 ≤IS'<IMM 2 , then the sample level is low immunity;

[0233] Step S3340, if IMM 2 ≤IS'<IMM 3 , then the sample level is immune normal level;

[0234] Step S3350, if IMM 3 ≤IS'≤1, the sample level is immune enhancement level;

[0235] Step S3360: For samples at the immunosuppression level and the immunodeficiency level, mark them as abnormal immune function.

[0236] Exemplary, immunosuppressive level (IS'∈[0,0.25)), immunocompromised level (IS'∈[0.25,0.5)), immune normal level (IS'∈[0.5,0.75)), immune enhanced level (IS'∈[0.75,1]).

[0237] Specifically, step S3300 divides the sample into different immune levels according to the immune score level of the sample, and performs a hierarchical diagnosis on the immune status of the sample. By setting the immune threshold, the sample can be classified into different levels such as immunosuppression, immunodeficiency, normal immunity, and immune enhancement. The normalized immune score is positively correlated with the immune level. The higher the score, the stronger the comprehensive immune function of the sample. This grading method fully considers the distribution characteristics of the immune score, and divides the immune status equally with quartiles as the boundary. Through statistical analysis of the number and proportion of samples at different immune levels, the overall immune status of the sample can be fully grasped, and the high-risk population with abnormal immune function can be identified. The statistical results of immune grading can be visualized in the form of histograms, pie charts, etc., which is convenient for medical staff to quickly understand and interpret the test results. For samples with low immune scores, especially individuals with immunosuppression and immunodeficiency, it is suggested that their immune function is impaired, the body's ability to resist external pathogens is reduced, and the risk of infection is significantly increased. Clinicians need to focus on these patients with abnormal immune function, and comprehensively analyze other disease indicators and medical history data to clarify the cause of immunosuppression. For patients with impaired immune function, intervention measures such as immunomodulatory therapy and key protection can be taken to maintain the body's immune homeostasis and reduce the occurrence of infectious complications. By setting multiple immune thresholds, a refined grading of the immune function of samples is achieved. This grading method can not only identify samples with abnormal immune function, but also distinguish different degrees of immune function status, providing more accurate guidance for clinical diagnosis and treatment.

[0238] Example 2

[0239] This embodiment provides an automated analysis platform integrating multiple immunoassay technologies on the basis of embodiment 1, such as Fig. 9 As shown, including:

[0240] Feature extraction module: used to receive raw data from sample detection using multiple immune detection technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data and immunofluorescence data; extract features from the ELISA data, flow cytometry data and immunofluorescence data in the first data set respectively to obtain a first feature set, a second feature set and a third feature set;

[0241] Redundancy elimination module: used to construct mutual information matrices between the first feature set, the second feature set and the third feature set, determine the binarization threshold, perform binarization processing on the mutual information matrix and eliminate redundant features to obtain the fourth feature set, the fifth feature set and the sixth feature set;

[0242] Dimension judgment module: used to judge whether the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements. If so, samples with abnormal immune function are identified based on the fourth feature set, the fifth feature set and the sixth feature set; if not, the binarization threshold is dynamically adjusted to obtain an updated binarization threshold, and the mutual information matrix is ​​re-binarized and redundant features are eliminated according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements.

[0243] In the feature extraction module, the raw data from the sample detection by multiple immune detection technologies is received to obtain the first data set, including:

[0244] Step S1110, receiving raw data from sample detection using various immunoassay techniques such as ELISA, flow cytometry, and immunofluorescence to form a raw data set;

[0245] Step S1120, calculating the signal-to-noise ratio of each data in the original data set, marking the data with a signal-to-noise ratio lower than a preset signal-to-noise ratio threshold as noise data; using a local outlier factor (LOF) algorithm to perform outlier detection on the original data, marking the data with a LOF value exceeding a preset LOF threshold as an outlier;

[0246] Step S1130, remove noise data and outliers to obtain a first data set; and perform data normalization on the first data set to eliminate the influence of different data sources and dimensions.

[0247] In the feature extraction module, extracting features from the ELISA data, the flow cytometry data, and the immunofluorescence data in the first data set includes:

[0248] Step S1210, extracting features from the ELISA data in the first data set, including: extracting the slope, intercept and goodness of fit R of the ELISA standard curve 2 As the first quantitative feature, the ratio of the sample absorbance to the negative control absorbance and the ratio of the sample absorbance to the positive control absorbance are extracted as the first qualitative feature to obtain a first feature set;

[0249] Step S1220, performing feature extraction on the flow cytometry data in the first data set, including: extracting the mean fluorescence intensity MFI and cell ratio of each cell population as the second quantitative feature, extracting the peak type and peak position of the fluorescence signal distribution as the second qualitative feature, and obtaining a second feature set;

[0250] Step S1230, performing feature extraction on the immunofluorescence data in the first data set, including: extracting the mean and standard deviation of the fluorescence intensity as the third quantitative feature, extracting the positive rate and colocalization coefficient of cell staining as the third qualitative feature, and obtaining a third feature set.

[0251] The redundant elimination module comprises:

[0252] Mutual information matrix construction unit: used to construct mutual information matrices between the first feature set, the second feature set and the third feature set; the mutual information matrix includes a first mutual information matrix M 12 , the second mutual information matrix M 13 And the third mutual information matrix M 23 ;

[0253] Binarization unit: used to perform binarization processing on the mutual information matrix, determine the binarization threshold, set the elements in the mutual information matrix that are greater than the binarization threshold to 1, and the rest to 0, to obtain multiple binarization matrices; the binarization threshold includes a first binarization threshold t 12 , the second binarization threshold t 13 and the third binarization threshold t 23 ; The binarization matrix includes a first binarization matrix B 12 , the second binarization matrix B 13 And the third binarization matrix B 23 ;

[0254] Redundant matrix acquisition unit: used to perform logical OR operation on multiple binary matrices to obtain a redundant feature matrix R;

[0255] Feature elimination unit: used to eliminate redundant features in the first feature set, the second feature set and the third feature set according to the redundant feature matrix R, to obtain the fourth feature set, the fifth feature set and the sixth feature set.

[0256] In the mutual information matrix construction unit, constructing the mutual information matrix between the first feature set, the second feature set, and the third feature set comprises:

[0257] Step S2110, pairing the features in the first feature set and the second feature set in pairs to construct a first feature pair set; pairing the features in the first feature set and the third feature set in pairs to construct a second feature pair set; pairing the features in the second feature set and the third feature set in pairs to construct a third feature pair set;

[0258] Step S2120, respectively calculating the mutual information between each pair of features in the first feature pair set, the second feature pair set, and the third feature pair set;

[0259] Step S2130: construct a first mutual information matrix M between the first feature set and the second feature set according to the calculated mutual information. 12 , the second mutual information matrix M between the first feature set and the third feature set 13 , the third mutual information matrix M between the second feature set and the third feature set 23 .

[0260] In the binarization unit, the binarization processing of the mutual information matrix includes:

[0261] Step S2210: 12 Medium greater than t 12 Set the elements of 1 and the rest of the elements to 0 to get the first binarized matrix B 12 ;

[0262] Step S2220: M 13 Medium greater than t 13 Set the elements of 1 and the rest of the elements to 0 to get the second binarized matrix B 13 ;

[0263] Step S2230: M 23 Medium greater than t 23 Set the elements of 1 and the rest of the elements to 0 to get the third binarized matrix B 23 .

[0264] In the redundant matrix acquisition unit, performing a logical OR operation on a plurality of binary matrices to obtain a redundant feature matrix R includes:

[0265] Step S2310: Binarize the matrix B 12 and B 13 Perform a logical OR operation to obtain the first redundant feature matrix R 1 ;

[0266] Step S2320, binarize the matrix B 12 and B 23 Perform a logical OR operation to obtain the second redundant feature matrix R 2 ;

[0267] Step S2330: Binarize the matrix B 13 and B 23 Perform a logical OR operation to obtain the third redundant feature matrix R 3 ;

[0268] Step S2340: the first redundant feature matrix R 1 , the second redundant feature matrix R 2 and the third redundant feature matrix R 3 Perform a logical OR operation to obtain the final redundant feature matrix R.

[0269] In the feature elimination unit, eliminating redundant features in the first feature set, the second feature set, and the third feature set includes:

[0270] Step S2410, traverse each element of the redundant feature matrix R, collect the row and column numbers corresponding to the elements with values ​​of 1, and obtain a redundant feature number set;

[0271] Step S2420, according to the redundant feature number set, remove the features with corresponding numbers in the first feature set to obtain a fourth feature set;

[0272] Step S2430, according to the redundant feature number set, remove the features with corresponding numbers in the second feature set to obtain a fifth feature set;

[0273] Step S2440: According to the redundant feature number set, the features with corresponding numbers in the third feature set are removed to obtain a sixth feature set.

[0274] In the dimension judgment module, the identifying of samples with abnormal immune function based on the fourth feature set, the fifth feature set and the sixth feature set includes:

[0275] Step S3100, performing principal component analysis on the fourth feature set, the fifth feature set and the sixth feature set respectively, and extracting principal components as immune scoring indicators; the immune scoring indicators include a first immune scoring indicator, a second immune scoring indicator and a third immune scoring indicator;

[0276] Step S3200, calculating the immune score IS of the sample according to the first immune score indicator, the second immune score indicator and the third immune score indicator; normalizing the immune score IS to obtain a normalized immune score IS'; IS'∈[0,1];

[0277] Step S3300: According to the normalized immune score IS', the samples are graded according to their immune levels to identify samples with abnormal immune function.

[0278] The step S3100 includes:

[0279] Step S3110, perform principal component analysis on the fourth feature set to extract the first principal component PC E1 , the second principal component PC E2 and the third principal component PC E3 , as the first immune score indicator;

[0280] Step S3120, perform principal component analysis on the fifth feature set to extract the first principal component PC C1 , the second principal component PC C2 and the third principal component PC C3 , as the second immune score indicator;

[0281] Step S3130, perform principal component analysis on the sixth feature set to extract the first principal component PC F1 , the second principal component PC F2 and the third principal component PC F3 , as the third immune score indicator.

[0282] The step S3300 includes:

[0283] Step S3310, setting the first immune threshold IMM 1 , the second immune threshold IMM 2 and the third immunity threshold IMM 3 Among them, IMM 3 >IMM 2 >IMM 1 ;

[0284] Step S3320, if 0≤IS'<IMM 1 , then the sample level is immunosuppression level;

[0285] Step S3330, if IMM 1 ≤IS'<IMM 2 , then the sample level is low immunity;

[0286] Step S3340, if IMM 2 ≤IS'<IMM 3 , then the sample level is immune normal level;

[0287] Step S3350, if IMM 3 ≤IS'≤1, the sample level is immune enhancement level;

[0288] Step S3360: For samples at the immunosuppression level and the immunodeficiency level, mark them as abnormal immune function.

[0289] Example 3

[0290] This embodiment discloses an electronic device, which may include one or more processors and one or more memories, wherein the memories store computer-readable codes, which, when executed by the one or more processors, may execute the automated analysis method integrating multiple immunoassay technologies as described above.

[0291] The method or system according to the embodiment of the present application can also be implemented by means of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. A storage device in an electronic device, such as a ROM or a hard disk, can store the automated analysis method for integrating multiple immunoassay techniques provided by the present application. An automated analysis method integrating multiple immune detection technologies may, for example, include: receiving raw data from sample detection using multiple immune detection technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data, and immunofluorescence data; extracting features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set, respectively, to obtain a first feature set, a second feature set, and a third feature set; constructing a mutual information matrix between the first feature set, the second feature set, and the third feature set, determining a binarization threshold, binarizing the mutual information matrix, and eliminating redundant features, to obtain a fourth feature set, a fifth feature set, and a sixth feature set; determining whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements, and if so, identifying samples with abnormal immune function based on the fourth feature set, the fifth feature set, and the sixth feature set; if not, dynamically adjusting the binarization threshold to obtain an updated binarization threshold, and re-binarizing the mutual information matrix and eliminating redundant features according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements.

[0292] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0293] Example 4

[0294] The present embodiment discloses a computer-readable storage medium, on which computer-readable storage medium is stored computer-readable instructions. When the computer-readable instructions are executed by a processor, the automated analysis method of integrating multiple immunoassay techniques according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may, for example, include random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0295] In addition, according to the implementation mode of the present application, the process described in the above reference flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: receiving raw data from sample detection using multiple immunoassay technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data, and immunofluorescence data; extracting features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set, respectively, to obtain a first feature set, a second feature set, and a third feature set; constructing the first feature set, the second feature set, and the third feature set in pairs; The mutual information matrix between the four features is determined, the binarization threshold is determined, the mutual information matrix is ​​binarized and redundant features are eliminated, and the fourth feature set, the fifth feature set and the sixth feature set are obtained; it is determined whether the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements, if they meet, then based on the fourth feature set, the fifth feature set and the sixth feature set, the samples with abnormal immune function are identified; if not, the binarization threshold is dynamically adjusted to obtain an updated binarization threshold, and the mutual information matrix is ​​binarized and redundant features are eliminated again according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0296] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0297] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0298] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An automated analysis method integrating multiple immunoassay techniques, characterized in that: The method comprises: Receiving raw data from sample detection using a plurality of immunoassay technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data, and immunofluorescence data; extracting features from the ELISA data, flow cytometry data, and immunofluorescence data in the first data set, respectively, to obtain a first feature set, a second feature set, and a third feature set; Constructing mutual information matrices between the first feature set, the second feature set, and the third feature set, determining a binarization threshold, binarizing the mutual information matrix and eliminating redundant features to obtain a fourth feature set, a fifth feature set, and a sixth feature set; Determine whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements. If so, identify samples with abnormal immune function based on the fourth feature set, the fifth feature set, and the sixth feature set; if not, dynamically adjust the binarization threshold to obtain an updated binarization threshold, and re-binarize the mutual information matrix and eliminate redundant features according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements.

2. The automated analysis method integrating multiple immunoassay techniques according to claim 1, characterized in that: The extracting features of the ELISA data, the flow cytometry data and the immunofluorescence data in the first data set respectively comprises: For ELISA data, extract the slope, intercept and goodness of fit R of the ELISA standard curve 2 As the first quantitative feature, the ratio of the sample absorbance to the negative control absorbance and the ratio of the sample absorbance to the positive control absorbance are extracted as the first qualitative feature to obtain a first feature set; For the flow cytometry data, the mean fluorescence intensity MFI and cell ratio of each cell population are extracted as the second quantitative feature, and the peak type and peak position of the fluorescence signal distribution are extracted as the second qualitative feature to obtain the second feature set; For immunofluorescence data, the mean and standard deviation of fluorescence intensity were extracted as the third quantitative feature, and the positive rate and colocalization coefficient of cell staining were extracted as the third qualitative feature to obtain the third feature set.

3. The automated analysis method integrating multiple immunoassay techniques according to claim 1, characterized in that: The mutual information matrix includes a first mutual information matrix M 12 , the second mutual information matrix M 13 And the third mutual information matrix M 23 ; The constructing of the mutual information matrix between the first feature set, the second feature set and the third feature set comprises: Pair the features in the first feature set and the second feature set in pairs to construct a first feature pair set; pair the features in the first feature set and the third feature set in pairs to construct a second feature pair set; pair the features in the second feature set and the third feature set in pairs to construct a third feature pair set; Calculate the mutual information between each pair of features in the first feature pair set, the second feature pair set, and the third feature pair set respectively; According to the calculated mutual information, the first mutual information matrix M between the first feature set and the second feature set is constructed respectively. 12 , the second mutual information matrix M between the first feature set and the third feature set 13 , the third mutual information matrix M between the second feature set and the third feature set 23 .

4. The automated analysis method integrating multiple immunoassay techniques according to claim 3, characterized in that: The binarization threshold includes a first binarization threshold t 12 , the second binarization threshold t 13 and the third binarization threshold t 23 ; The binarization processing of the mutual information matrix and the elimination of redundant features include: M 12 Medium greater than t 12 Set the elements of 1 and the rest of the elements to 0 to get the first binarized matrix B 12 ; M 13 Medium greater than t 13 Set the elements of 1 and the rest of the elements to 0 to get the second binarized matrix B 13 ; M 23 Medium greater than t 23 Set the elements of 1 and the rest of the elements to 0 to get the third binarized matrix B 23 ; To B 12 and B 13 Perform a logical OR operation to obtain a first redundant feature matrix R1; To B 12 and B 23 Perform a logical OR operation to obtain a second redundant feature matrix R2; To B 13 and B 23 Perform a logical OR operation to obtain a third redundant feature matrix R3; Performing a logical OR operation on the first redundant feature matrix R1, the second redundant feature matrix R2, and the third redundant feature matrix R3 to obtain a final redundant feature matrix R; Traverse each element of the redundant feature matrix R, collect the row and column numbers corresponding to the elements with values ​​of 1, and obtain a set of redundant feature numbers; According to the redundant feature number set, the features with corresponding numbers in the first feature set are removed; According to the redundant feature number set, the features with corresponding numbers in the second feature set are removed; According to the redundant feature number set, the features with corresponding numbers in the third feature set are removed.

5. The automated analysis method integrating multiple immunoassay techniques according to claim 1, characterized in that: The determining whether the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set meet the preset dimension requirements includes: Calculate the feature dimensions of the fourth feature set, the fifth feature set, and the sixth feature set, respectively, and record them as dim4, dim5, and dim6; Determine whether dim4, dim5, and dim6 are all less than or equal to the preset dimension threshold d; if so, it is considered that the feature dimension meets the preset dimension requirement; if not, it is considered that the feature dimension does not meet the preset dimension requirement; The dynamically adjusting the binarization threshold comprises: setting the binarization threshold t 12 ,t 13 and t 23 Multiply them by the attenuation factor γ, 0<γ<1 respectively.

6. The automated analysis method integrating multiple immunoassay techniques according to claim 1, characterized in that: The identifying of samples with abnormal immune function based on the fourth feature set, the fifth feature set and the sixth feature set includes: Perform principal component analysis on the fourth feature set and extract the first principal component PC E1 , the second principal component PC E2 and the third principal component PC E3 , as the first immune score indicator; Perform principal component analysis on the fifth feature set and extract the first principal component PC C1 , the second principal component PC C2 and the third principal component PC C3 , as the second immune score indicator; Perform principal component analysis on the sixth feature set and extract the first principal component PC F1 , the second principal component PC F2 and the third principal component PC F3 , as the third immune score indicator; According to the first immune score index, the second immune score index and the third immune score index, the immune score IS of the sample is calculated; the immune score IS is normalized to obtain a normalized immune score IS'; IS'∈[0,1]; According to the normalized immune score IS', the samples were graded for immune grade and samples with abnormal immune function were identified.

7. The automated analysis method integrating multiple immunoassay techniques according to claim 6, characterized in that: The immune grade of the samples is graded according to the normalized immune score IS', and the samples with abnormal immune function are identified, including: A first immune threshold IMM1, a second immune threshold IMM2 and a third immune threshold IMM3 are set; wherein IMM3>IMM2>IMM1; If 0≤IS'<IMM1, the sample level is immunosuppression level; If IMM1≤IS'<IMM2, the sample level is low immunity; If IMM2≤IS'<IMM3, the sample level is immune normal level; If IMM3≤IS'≤1, the sample level is immune enhancement level; Samples at the immunosuppressed and immunocompromised levels were marked as immune dysfunction.

8. An automated analysis platform integrating multiple immunoassay technologies, which is used to implement the automated analysis method integrating multiple immunoassay technologies according to any one of claims 1 to 7, characterized in that: The platform includes: Feature extraction module: used to receive raw data from sample detection using multiple immune detection technologies to obtain a first data set; the first data set includes ELISA data, flow cytometry data and immunofluorescence data; extract features from the ELISA data, flow cytometry data and immunofluorescence data in the first data set respectively to obtain a first feature set, a second feature set and a third feature set; Redundancy elimination module: used to construct mutual information matrices between the first feature set, the second feature set and the third feature set, determine the binarization threshold, perform binarization processing on the mutual information matrix and eliminate redundant features to obtain the fourth feature set, the fifth feature set and the sixth feature set; Dimension judgment module: used to judge whether the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements. If so, samples with abnormal immune function are identified based on the fourth feature set, the fifth feature set and the sixth feature set; if not, the binarization threshold is dynamically adjusted to obtain an updated binarization threshold, and the mutual information matrix is ​​re-binarized and redundant features are eliminated according to the updated binarization threshold until the feature dimensions of the fourth feature set, the fifth feature set and the sixth feature set meet the preset dimension requirements.

9. An electronic device comprising a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit, characterized in that: When the central processing unit executes the computer program, the automated analysis method integrating multiple immunoassay technologies described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed, implements the automated analysis method integrating multiple immunoassay technologies according to any one of claims 1 to 7.

Images