Method and apparatus for correction of metabolomic data based on microbial similarity structure
By constructing a metabolomics data correction method based on microbial similarity structure, the systematic background interference problem introduced by microbial community structure is solved, the accuracy of metabolite identification and the interpretability of results are improved, and more stable identification of disease-related metabolic signals is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- EAST CHINA UNIV OF TECH
- Filing Date
- 2026-04-17
- Publication Date
- 2026-07-14
Smart Images
Figure CN122392648A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing, specifically to a method and apparatus for correcting metabolomics data based on microbial similarity structures. The research conducted in this invention was supported by the National Natural Science Foundation of China, "A New Method for Personalized Molecular Network Modeling Applied to the Study of Metabolic Heterogeneity in Colorectal Cancer" (Project No.: 82360363). Background Technology
[0002] In gut multi-omics studies, metabolomics and microbiome are often combined to analyze the interaction between host metabolic state and gut microbiota. The gut microbiota participates in host metabolic regulation through metabolite synthesis, transformation, and signal molecule exchange; therefore, changes in microbiome structure are usually accompanied by changes in metabolic profile. In studies of various gut-related diseases, such as inflammatory bowel disease (IBD) or colorectal cancer (CRC) patients, samples often show both microbiome imbalance and abnormal metabolite levels.
[0003] However, in actual data analysis, complex cross-omics coupling relationships exist between the microbiome and metabolome. The gut microbiota typically exhibits a stable co-occurrence structure, with multiple microorganisms changing synergistically at the community level. When different samples have similar microbial compositions, their metabolite profiles may also show overall similarity. This systematic similarity driven by community structure can easily be mapped into correlations between metabolites or between metabolites and diseases in statistical analysis, thus generating indirect associations or spurious correlation signals. These phenomena can mask true disease-related metabolic changes and may lead to the omission of key metabolites or the misidentification of irrelevant variables as differentially expressed metabolites.
[0004] Current metabolomics data analysis methods primarily rely on univariate statistical tests or multivariate classification models. For example, statistical tests based on mean differences and multivariate models based on projections typically assume that samples are independent and that the relationships between variables are relatively simple. When the overall structure of the microbial community changes, multiple metabolites may be simultaneously affected by community effects. This makes it difficult for these methods to distinguish between direct metabolic changes driven by disease and indirect changes caused by microbial community structure, thereby reducing the accuracy of differential metabolite identification and affecting model stability and the interpretability of results.
[0005] Furthermore, in metabolic pathway or metabolic network analyses, the aforementioned community structure effects can also introduce systematic bias. Some significant pathways may only reflect the co-occurrence patterns of the microbial community, rather than the actual disease-related metabolic processes, thus affecting the explanation of disease mechanisms.
[0006] To address the problem of background interference, some data analysis methods propose using the similarity structure between samples to identify and remove systematic changes driven by background factors. These methods construct sample structural information in the feature space by calculating the similarity relationships between samples, and then perform structured projection or reconstruction of the data. This allows them to identify systematic changes caused by background structures and remove them from the original data, thereby reducing the impact of indirect correlation signals introduced by background factors on the analysis results.
[0007] In the context of gut multi-omics data analysis, microbial community structure can also be considered a potential source of background structure. When there are significant differences in microbiome structure, metabolic data may contain systematic variations introduced by community co-occurrence patterns. Therefore, a method is needed to correct metabolomics data using microbial similarity structures to reduce the influence of cross-omics confounding factors and improve the accuracy and reliability of metabolomics data analysis results. Summary of the Invention
[0008] The purpose of this application is to propose a method and apparatus for correcting metabolomics data based on microbial similarity structures to address the aforementioned technical problems.
[0009] In a first aspect, the present invention provides a method for correcting metabolomics data based on microbial similarity structures, comprising the following steps:
[0010] We acquire microbiome and metabolome data of labeled samples collected and detected at the same time point in the same organism or environment, and construct matching microbiome and metabolome data matrices; we preprocess the microbiome and metabolome data matrices respectively to obtain preprocessed microbiome and metabolome data matrices.
[0011] The preprocessed microbiome data matrix is binarized to obtain a binarized microbiome data matrix. Based on the binarized microbiome data matrix, the similarity between any sample and any sample labeled as healthy is calculated. The top P samples labeled as healthy with the highest similarity to any sample are selected and a nearest neighbor weight matrix is constructed. A mapping matrix is constructed based on the nearest neighbor weight matrix.
[0012] The metabolomics data matrices of samples marked as healthy in the preprocessed metabolomics data matrix are linearly combined using the mapping matrix to obtain the projection matrix; the preprocessed metabolomics data matrix is then corrected using the projection matrix to obtain the corrected metabolomics data matrix.
[0013] Preferably, the preprocessing procedure includes:
[0014] The microbiome data matrix and metabolome data matrix are filtered based on the proportion of missing values. Microbial variables and metabolite variables with a zero value proportion exceeding a preset threshold are deleted to obtain the preprocessed microbiome data matrix and the filtered metabolome data matrix.
[0015] The zero values in the filtered metabolomics data matrix were imputed using the median imputation strategy, followed by log2 logarithmic transformation and unit variance standardization to obtain the preprocessed metabolomics data matrix.
[0016] Preferably, the preprocessed microbiome data matrix is binarized to obtain a binarized microbiome data matrix, specifically including:
[0017] Binarization is performed using the following formula:
[0018] ;
[0019] in, This represents the first element in the pretreated microbiome data matrix. The first sample Abundance values of individual microbial variables This represents the first element in the binarized microbiome data matrix. The first sample Binarized values of individual microbial variables .
[0020] As a preferred method, similarity includes Jaccard similarity, calculated using the following formula:
[0021] ;
[0022] in, Represents any sample Compared with any sample marked as healthy Similarity between them Represents any sample labeled as healthy. The set of binarized values in the binarized microbiome data matrix. Represents any sample The set of binarized values in the binarized microbiome data matrix. .
[0023] As a preferred method, the construction process of the nearest neighbor weight matrix is as follows:
[0024] ;
[0025] in, Represents any sample The nearest neighbor set is constructed from the top P samples with the highest similarity, which are labeled as healthy. The nearest neighbor weight matrix represents the th Line number Column elements, , ;
[0026] The process of constructing the mapping matrix is as follows:
[0027] By performing an exponential transformation on each element of the nearest neighbor weight matrix, we obtain the transformed nearest neighbor weight matrix, as shown in the following equation:
[0028] ;
[0029] in, Represents the nearest neighbor weight matrix after transformation. Line number Column elements, For exponential scaling parameters;
[0030] The transformed nearest neighbor weight matrix is normalized to obtain the mapping matrix, as shown in the following equation:
[0031]
[0032] in, Represents the first in the mapping matrix Line number The elements of the column.
[0033] As a preferred method, the calculation process for the projection matrix is as follows:
[0034] ;
[0035] in, This represents the metabolomics data matrix of samples labeled as healthy in the preprocessed metabolomics data matrix. Represents the projection matrix. Represents the mapping matrix;
[0036] The calculation process for the corrected metabolomics data matrix is as follows:
[0037] ;
[0038] in, This represents the preprocessed metabolomics data matrix. This represents the corrected metabolomics data matrix.
[0039] Secondly, the present invention provides a metabolomics data correction device based on microbial similarity structure, comprising:
[0040] The preprocessing module is configured to acquire microbiome and metabolome data of labeled samples collected and detected at the same time point in the same organism or environment, and construct a matching microbiome data matrix and metabolome data matrix; and preprocess the microbiome data matrix and metabolome data matrix respectively to obtain the preprocessed microbiome data matrix and the preprocessed metabolome data matrix.
[0041] The mapping matrix construction module is configured to binarize the preprocessed microbiome data matrix to obtain a binarized microbiome data matrix, calculate the similarity between any sample and any sample labeled as healthy based on the binarized microbiome data matrix, select the top P samples labeled as healthy with the highest similarity to any sample and construct a nearest neighbor weight matrix; and construct a mapping matrix based on the nearest neighbor weight matrix.
[0042] The correction module is configured to use a mapping matrix to linearly combine the metabolomics data matrices of samples marked as healthy in the preprocessed metabolomics data matrix to obtain a projection matrix; and to correct the preprocessed metabolomics data matrix using the projection matrix to obtain a corrected metabolomics data matrix.
[0043] Thirdly, the present invention provides an electronic device including one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any implementation of the first aspect.
[0044] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the implementations of the first aspect.
[0045] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in any of the implementations in the first aspect.
[0046] Compared with the prior art, the present invention has the following beneficial effects:
[0047] (1) The metabolomics data correction method based on microbial similarity structure proposed in this invention is based on paired microbial and metabolomics data. It constructs microbial similarity structure between samples by utilizing microbial information, and reconstructs and corrects metabolomics data based on the microbial background information of healthy samples. This effectively removes the systematic background interference introduced by the microbial community structure, improves the stability and interpretability of disease-related metabolic signals, and provides a reliable data basis for the identification of differential metabolites.
[0048] (2) The metabolomics data correction method based on microbial similarity structure proposed in this invention performs binarization processing on the preprocessed microbial data sentences, determines the nearest neighbor set of each sample by calculating the Jaccard similarity between samples, and constructs a mapping matrix that reflects the co-occurrence structure of the microbial community. This solves the problem that the cross-omics confounding effect introduced by the difference in microbial community structure in the joint analysis of microbiome and metabolome can easily lead to metabolite association bias and generate spurious correlation signals.
[0049] (3) The metabolomics data correction method based on microbial similarity structure proposed in this invention introduces a microbial similarity metabolic correction method. The metabolomics data is projected onto the mapping matrix to obtain the projection matrix. By subtracting the projection matrix from the preprocessed metabolomics data matrix, the systematic confounding components introduced by the microbial community similarity structure are removed, thereby obtaining the corrected metabolic data matrix. The background variation in the metabolomics data is corrected by the microbial community similarity structure, reducing the impact of the systematic bias driven by the community structure on the statistical analysis, thereby improving the accuracy of differential metabolite identification and enhancing the accuracy and separability of differential metabolite identification results in gut multi-omics research. Attached Figure Description
[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0051] Figure 1 This is a schematic flowchart illustrating the metabolomics data correction method based on microbial similarity structure, as an embodiment of this application.
[0052] Figure 2 A comparison chart of PCA scores of metabolomics data before and after correction using the metabolomics data correction method based on microbial similarity structure, as shown in the embodiments of this application;
[0053] Figure 3 A comparison chart of PCA scores of metabolomics data before and after correction using the metabolomics data correction method based on microbial similarity structure, as shown in the embodiments of this application;
[0054] Figure 4 This is a Venn diagram comparing the results obtained using three different statistical strategies based on the original metabolomics data before correction.
[0055] Figure 5This is a Venn diagram comparing the results obtained using three different statistical strategies based on the metabolomics data correction method based on microbial similarity structure according to the embodiments of this application.
[0056] Figure 6 This is a schematic diagram of a metabolomics data correction device based on microbial similarity structure, as an embodiment of this application.
[0057] Figure 7 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0059] Figure 1 An embodiment of this application provides a method for correcting metabolomics data based on microbial similarity structures, comprising the following steps:
[0060] S1. Acquire microbiome and metabolome data of labeled samples collected and detected at the same time point in the same organism or environment, and construct a matching microbiome data matrix and metabolome data matrix; preprocess the microbiome data matrix and metabolome data matrix respectively to obtain the preprocessed microbiome data matrix and the preprocessed metabolome data matrix.
[0061] In a specific embodiment, the preprocessing process includes:
[0062] The microbiome data matrix and metabolome data matrix are filtered based on the proportion of missing values. Microbial variables and metabolite variables with a zero value proportion exceeding a preset threshold are deleted to obtain the preprocessed microbiome data matrix and the filtered metabolome data matrix.
[0063] The zero values in the filtered metabolomics data matrix were imputed using the median imputation strategy, followed by log2 logarithmic transformation and unit variance standardization to obtain the preprocessed metabolomics data matrix.
[0064] Specifically, in the embodiments of this application, a sample refers to a biological entity independently collected and detected at the same time point from the same organism or environment (such as intestinal contents, feces, etc.). Each sample number corresponds to a unique set of observations, including microbiome data and metabolome data. Metabolome data is obtained through high-throughput methods such as mass spectrometry (MS) or nuclear magnetic resonance (NMR) to reflect the intensity of metabolite variables within the sample; while microbiome data is obtained through metagenomics or 16S rRNA sequencing technology to characterize the composition and abundance of the microbial community. To construct topological associations between samples and perform subsequent metabolic corrections, the two types of data must be paired and matched, ensuring they originate from the same sample with the same number, and are strictly sorted and aligned according to sample number before analysis.
[0065] Let the total Each sample is labeled as healthy or diseased, and each sample is tested. Metabolic variables and One microbial variable. The metabolomics data matrix is denoted as... ,in Indicates the first The first sample Intensity of each metabolite variable A matrix of metabolite data for samples labeled as healthy. A matrix of metabolite data for samples labeled as disease. Let represent the set of real numbers. The microbiome data matrix is denoted as . ,in Indicates the first In the nth sample Abundance values of individual microbial variables The microbiome matrix of samples labeled as healthy. A microbiome matrix for samples labeled as disease.
[0066] Obtain the matching microbiome data matrix With metabolomics data matrix The samples were then sorted and aligned according to their numbers. Subsequently, the microbiome and metabolome data were filtered for missing values. Specifically, the proportion of zero values in each variable was calculated to represent the total number of samples, and variables with a proportion of zero values exceeding a preset threshold were deleted, resulting in the preprocessed microbiome data matrix and the filtered metabolome data matrix.
[0067] For zero values in the filtered metabolomics data matrix, a median imputation strategy was used to fill in missing values. Then, a log2 logarithmic transformation was performed on the processed metabolomics data. Finally, unit variance standardization was applied to the variables to obtain the preprocessed metabolomics data matrix.
[0068] S2, the preprocessed microbiome data matrix is binarized to obtain a binarized microbiome data matrix. The similarity between any sample and any sample marked as healthy is calculated based on the binarized microbiome data matrix. The top P samples marked as healthy with the highest similarity to any sample are selected and a nearest neighbor weight matrix is constructed. A mapping matrix is constructed based on the nearest neighbor weight matrix.
[0069] In a specific embodiment, the preprocessed microbiome data matrix is binarized to obtain a binarized microbiome data matrix, specifically including:
[0070] Binarization is performed using the following formula:
[0071] ;
[0072] in, This represents the first element in the pretreated microbiome data matrix. The first sample Abundance values of individual microbial variables This represents the first element in the binarized microbiome data matrix. The first sample Binarized values of individual microbial variables .
[0073] Specifically, the microbial data in the preprocessed microbial data matrix is binarized, that is, the binarized value corresponding to the abundance value greater than 0 is set to 1, and the binarized value corresponding to the abundance value of 0 is set to 0, thereby constructing a binarized microbial data matrix. The elements in this binarized microbial data matrix are all binarized values obtained by converting the microbial data in the preprocessed microbial data matrix.
[0074] In a specific embodiment, the similarity includes Jaccard similarity, calculated using the following formula:
[0075] ;
[0076] in, Represents any sample Compared with any sample marked as healthy Similarity between them Represents any sample labeled as healthy. The set of binarized values in the binarized microbiome data matrix. Represents any sample The set of binarized values in the binarized microbiome data matrix. .
[0077] In a specific embodiment, the process of constructing the nearest neighbor weight matrix is as follows:
[0078] ;
[0079] in, Represents any sample The nearest neighbor set is constructed from the top P samples with the highest similarity, which are labeled as healthy. The nearest neighbor weight matrix represents the th Line number Column elements, , ;
[0080] The process of constructing the mapping matrix is as follows:
[0081] By performing an exponential transformation on each element of the nearest neighbor weight matrix, we obtain the transformed nearest neighbor weight matrix, as shown in the following equation:
[0082] ;
[0083] in, Represents the nearest neighbor weight matrix after transformation. Line number Column elements, For exponential scaling parameters;
[0084] The transformed nearest neighbor weight matrix is normalized to obtain the mapping matrix, as shown in the following equation:
[0085]
[0086] in, Represents the first in the mapping matrix Line number The elements of the column.
[0087] Specifically, for the binarized microbiome data matrix, for any sample in the sample set... The set of binarized values of the microbiome in the binarized microbiome data matrix is denoted as For any sample in the sample set labeled as healthy The set of binarized values of the microbiome in the binarized microbiome data matrix is denoted as .pass and Calculate samples Compared with samples marked as healthy Jaccard similarity between This yields any sample. Similarity vector .
[0088] In any sample In the similarity vector, represents any sample. Select the top ones with the highest similarity. Each sample labeled as healthy is used as a nearest neighbor sample to construct a corresponding nearest neighbor set. In one example, Take the total number of samples marked as healthy 5%.
[0089] Based on nearest neighbor set The nearest neighbor weight matrix can be constructed. The specific process is as follows: for each element in the nearest neighbor weight matrix Based on whether the sample corresponding to each row belongs to the nearest neighbor set Perform the assignment.
[0090] Further adjustments to the nearest neighbor weight matrix Each element in An exponential transformation is performed to non-linearly increase the weight differences between different similarity metrics, where the exponential scaling parameter... This step is used to adjust for weight differences. It aims to highlight the contributions of strongly correlated neighbors and suppress the influence of weakly correlated background. To ensure that the projection of each sample into the healthy sample space has a consistent order of magnitude, the transformed neighbor weight matrix is row-normalized, thus obtaining the mapping matrix. The mapping matrix represents the linear combination weights of each sample in the healthy sample space.
[0091] S3. The metabolomics data matrix of the preprocessed metabolomics data matrix marked as healthy samples is linearly combined using the mapping matrix to obtain the projection matrix; the preprocessed metabolomics data matrix is then corrected using the projection matrix to obtain the corrected metabolomics data matrix.
[0092] In a specific embodiment, the calculation process of the projection matrix is as follows:
[0093] ;
[0094] in, This represents the metabolomics data matrix of samples labeled as healthy in the preprocessed metabolomics data matrix. Represents the projection matrix. Represents the mapping matrix;
[0095] The calculation process for the corrected metabolomics data matrix is as follows:
[0096] ;
[0097] in, This represents the preprocessed metabolomics data matrix. This represents the corrected metabolomics data matrix.
[0098] Specifically, using a mapping matrix Metabolomics data matrix of samples labeled as healthy A linear combination is performed to obtain the projection matrix, which represents the metabolomics background component. Then, the preprocessed metabolomics data matrices corresponding to all samples are generated. Subtract its corresponding projection matrix The corrected metabolomics data matrix was obtained. The corrected metabolomics data matrix represents metabolic signals after removing the influence of microbial background.
[0099] The present invention will be further described below with reference to specific embodiments, but the scope of protection of the present invention is not limited thereto.
[0100] 1. Data Acquisition
[0101] The data used in the embodiments of this application comes from publicly available gut microbiome-metabolome integrated analysis resources. These resources have uniformly organized and standardized multi-omics data from multiple publicly available studies, and the data complies with legal and regulatory requirements. The embodiments of this application select the FRANZOSA_IBD_2019 dataset for method validation.
[0102] This dataset contains paired fecal metagenomic and metabolomical data, along with corresponding sample grouping information. The data was obtained from the project's official code repository. In this embodiment, a total of 132 samples were used, including 56 samples labeled as healthy controls and 76 samples labeled as disease. The disease-labeled samples were all from Ulcerative Colitis (UC) patients. Each sample contains corresponding gut microbiome abundance data and fecal metabolomical data, which can be used to analyze the relationship between microbial community structure and host metabolic profile, and to verify the effectiveness of the method in metabolomical background correction.
[0103] 2. Data Preprocessing
[0104] Microbiome data preprocessing: Variable screening was performed on the raw microbial data. First, the proportion of zero values (i.e., the missing rate) for each microorganism in all samples was calculated. Microbial variables with a missing rate greater than 50% were removed, retaining only those with a missing rate less than 50%. After screening, 2173 microbial variables were ultimately retained for subsequent analysis.
[0105] Preprocessing of metabolomics data: First, the proportion of zero values for each metabolite in all samples was calculated, and metabolite variables with a zero value proportion exceeding 5% were removed to reduce the impact of highly missing variables on the analysis results. After screening, 268 metabolite variables were retained. Then, the zero values in the data were processed, treating them as missing values. The median of non-zero values for each metabolite variable was calculated, and this median was used to impute the corresponding missing values. After missing value processing, a logarithmic transformation was performed on the metabolite data to compress the dynamic range of the data and reduce the impact of extreme values. Finally, the logarithmically transformed metabolite data matrix was standardized using unit-variance (UV) to bring the metabolite variables to a comparable scale.
[0106] 3. Metabolic data correction based on microbial similarity structures
[0107] The preprocessed microbial data matrix was binarized by setting values greater than 0 to 1 and all others to 0, resulting in a binarized microbiome data matrix used to identify the presence or absence of microorganisms. Subsequently, Jaccard similarity between samples was calculated based on this binarized microbiome data matrix to construct a sample similarity matrix. For each sample Based on the similarity, the top 5% of healthy samples that are closest to it are selected to form a nearest neighbor set. After obtaining the nearest neighbor set, an exponential mapping transformation is performed based on the neighbor weight matrix, and the transformed result is normalized to obtain the mapping matrix of the sample in the healthy sample space. Based on this, the metabolomics data matrix labeled as healthy is used... The preprocessed metabolomics data matrix is projected to obtain the projection matrix. And calculate the corrected metabolomics data matrix. .
[0108] 4. Method effectiveness evaluation and performance verification
[0109] (1) Correlation change analysis: Principal component analysis (PCA) was performed on the metabolome data matrices before and after correction to reduce the dimensionality, observe the distribution of samples in the low-dimensional projection space, and compare the changes in the main direction of variation, between-group separation and within-group concentration.
[0110] (2) A robust metabolite association network was constructed using the Probabilistic Context Likelihood of Relatedness (PCLRC) algorithm to assess the impact of the correction process on the interaction patterns between metabolite systems. First, the metabolome data was iterated 1000 times using Bootstrap resampling (randomly selecting 25%–75% of the samples each time), and the Spearman rank correlation coefficient between metabolites was calculated in each sampling to construct a probability distribution model of the correlation. Subsequently, probability estimation was performed on all iteration results, strictly retaining only strongly correlated terms with a probability value greater than 0.95 as stable edges, and the metabolite association matrices before and after correction were defined accordingly. Based on this, the maximum connected component in each network was extracted, and the number and proportion of newly added and disappeared associations in the statistical network were analyzed by comparison, and the differential connectivity of each metabolite was calculated. To ensure the statistical significance of the differences, 1000 permutation tests were further implemented to construct an empirical distribution under the null hypothesis, ultimately... Metabolites were identified as key variables that showed significant differences in network connectivity patterns.
[0111] (3) Discriminant performance verification: To evaluate the discriminant ability of the proposed method under small sample constraints and the stability of variable selection results, the embodiments of this application use Monte Carlo cross-validation (MCCV) strategy to construct a partial least squares discriminant (PLS-DA) classification model. The specific process is as follows:
[0112] Random performance evaluation: The entire dataset was randomly divided into a training set (70%) and a test set (30%) 1000 times. In each split, a classification model was built using the training set and made predictions on the independent test set, and the area under the receiver operating characteristic (AUC) was calculated. Finally, the mean AUC ± standard deviation obtained from 1000 simulations was used as a robust indicator to evaluate the model's classification performance and generalization ability.
[0113] Multi-criteria variable screening comparison: To verify the consistency of the variable screening results, this study compared the differentially identified metabolites by the proposed method with three other commonly used screening strategies. The criteria for determining the significance of each method are as follows:
[0114] Univariate statistics: using Test, with The significance threshold;
[0115] Multivariate Projective Importance: The projected importance values of variables are calculated using the PLS-DA model. As the screening criteria;
[0116] Topology differences: Permutation verification was performed using the PCLRC algorithm, with empirical values... This serves as evidence of a significant change in network connectivity patterns.
[0117] Consistency assessment: The number of metabolites commonly identified under different screening criteria was quantified using Venn diagrams. By analyzing the overlap of screening results across different dimensions, the consistency and reliability of this study's method in identifying key metabolic biomarkers were comprehensively evaluated.
[0118] 5. Results (1): Changes in data structure and correlation after correction
[0119] To assess the impact of correction on data structure, PCA dimensionality reduction was performed on the data before and after correction, and the differences in their distribution in the lower-dimensional space were compared. PCA analysis results showed that the correction process significantly affected the overall structure of the metabolomics data. (Reference) Figure 2 and Figure 3 Before correction, the first principal component explained 19.1% of the variance, which increased to 21.2% after correction, indicating an increase in the amount of information contained in the main directions of variation. In the low-dimensional projected space, the sample distribution was relatively discrete before correction, with some extreme values and uneven distance distribution between samples; after correction, the sample dispersion decreased, the influence of extreme values weakened, the sample distribution became more concentrated, and the principal axis directions became more stable. In the projected space composed of the first two principal components, the samples are more densely distributed in the low-dimensional space, and the ability to distinguish the main directions of variation is improved.
[0120] 6. Results (2): Analysis of changes in the structure of metabolite association networks
[0121] To evaluate the optimization effect of the PCLRC correction algorithm on the metabolite association network structure, the embodiments of this application compared and analyzed the evolution of topological features before and after correction for the healthy group and the disease group, as shown in Table 1. The results show that, while maintaining a constant node size, the correction process significantly enhanced the robustness and connectivity of the network: the number of edges in the healthy group increased from 409 to 512 (an increase of 25.2%), and the number of edges in the disease group increased from 472 to 549 (an increase of 16.3%). This change indicates that the correction process effectively eliminated background interference in the original data and identified more stable metabolic associations with high confidence. Correspondingly, the average degree of both groups of networks improved (from 8.26 to 10.34 in the healthy group and from 7.49 to 8.71 in the disease group), demonstrating that the corrected network construction method can reveal the synergistic interactions between metabolites at a deeper level, providing a more complete network topology foundation for the accurate identification of subsequent differential modules.
[0122] Table 1 Comparison of topological variables before and after metabolite association network correction
[0123]
[0124] 7. Results (3): Robustness assessment based on microbial correction
[0125] To evaluate the impact of calibration methods on model stability, Monte Carlo cross-validation was used to assess the disease classification models. The results showed that the PLS-DA models constructed from each variable set after calibration exhibited higher classification performance. The VIP metabolite variable set performed best, with AUC increasing from 0.9380 to 0.9579; the T-test variable set increased from 0.9131 to 0.9386; the PCLRC variable set increased from 0.8630 to 0.8979; the total metabolite variable set increased from 0.9080 to 0.9213; and the common metabolite variable set increased from 0.8900 to 0.9175, as shown in Table 2. Meanwhile, the number of metabolites screened by different methods also changed: the T-test method increased from 126 to 147, the VIP method increased from 98 to 116, the PCLRC method decreased from 154 to 133, and the common metabolite method increased from 52 to 60. Overall, the results indicate that the discriminative performance of the models was generally improved after calibration. refer to Figure 4 and Figure 5 Before correction, the three methods jointly identified 52 differentially identified metabolites, while after correction, this intersection increased to 60. Furthermore, taking the VIP method as an example, the number of differentially identified metabolites increased from 98 to 116 after correction. Further comparison revealed 32 key metabolites that appeared only in the corrected data.
[0126] Table 2. Mean AUC ± Standard Deviation Scores of PLS-DA Disease Prediction Models Constructed Based on Different Variable Screening Methods
[0127]
[0128] Further reference Figure 6 As an implementation of the methods shown in the above figures, this application provides an embodiment of a metabolomics data correction device based on microbial similarity structures. This device embodiment is similar to... Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.
[0129] This application provides a metabolomics data correction device based on microbial similarity structure, including:
[0130] Preprocessing module 1 is configured to acquire microbiome data and metabolome data of labeled samples collected and detected at the same time point in the same organism or environment, and construct a matching microbiome data matrix and metabolome data matrix; preprocess the microbiome data matrix and metabolome data matrix respectively to obtain the preprocessed microbiome data matrix and the preprocessed metabolome data matrix.
[0131] The mapping matrix construction module 2 is configured to binarize the preprocessed microbiome data matrix to obtain a binarized microbiome data matrix, calculate the similarity between any sample and any sample labeled as healthy based on the binarized microbiome data matrix, select the top P samples labeled as healthy with the highest similarity to any sample and construct a nearest neighbor weight matrix; and construct a mapping matrix based on the nearest neighbor weight matrix.
[0132] The correction module 3 is configured to use a mapping matrix to linearly combine the metabolomics data matrices of samples marked as healthy in the preprocessed metabolomics data matrix to obtain a projection matrix; and to correct the preprocessed metabolomics data matrix using the projection matrix to obtain a corrected metabolomics data matrix.
[0133] Figure 7 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. For example... Figure 7 As shown, the electronic device of this embodiment includes a processor 701 and a memory 702; wherein the memory 702 is used to store computer execution instructions; and the processor 701 is used to execute the computer execution instructions stored in the memory to implement the various steps performed by the electronic device in the above embodiment. For details, please refer to the relevant descriptions in the foregoing method embodiments.
[0134] Alternatively, the memory 702 can be either standalone or integrated with the processor 701.
[0135] When the memory 702 is set up independently, the electronic device also includes a bus 703 for connecting the memory 702 and the processor 701.
[0136] This invention also provides a computer storage medium storing computer execution instructions, which, when executed by processor 701, implement the above method.
[0137] This invention also provides a computer program product, including a computer program that, when executed by a processor 701, implements the above-described method.
[0138] In the embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.
[0139] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.
[0140] Furthermore, the functional modules in the various embodiments of this invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit formed by the above modules can be implemented in hardware or in the form of hardware plus software functional units.
[0141] The integrated modules implemented as software functional modules described above can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor 701 to execute some steps of the methods of the various embodiments of this application.
[0142] It should be understood that the processor 701 described above can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor, or the processor 701 can be any conventional processor 701. The steps of the method disclosed in this invention can be directly manifested as the hardware processor 701 executing the steps, or as a combination of hardware and software modules within the processor 701 executing the steps.
[0143] The memory 702 may include high-speed RAM memory, and may also include non-volatile memory NVM, such as at least one disk storage device, and may also be a USB flash drive, portable hard drive, read-only memory, disk or optical disc, etc.
[0144] Bus 703 can be an Industry Standard Architecture (ISA), a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 703 can be divided into address bus, data bus, control bus, etc. For ease of illustration, the bus 703 in the accompanying drawings of this application is not limited to only one bus 703 or one type of bus 703.
[0145] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.
[0146] An exemplary storage medium is coupled to a processor 701, enabling the processor 701 to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor 701. The processor 701 and the storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor 701 and the storage medium can exist as discrete components in an electronic device or a host device.
[0147] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for correcting metabolomics data based on microbial similarity structures, characterized in that, Includes the following steps: Acquire microbiome and metabolome data of labeled samples collected and detected at the same time point in the same organism or environment, and construct a matching microbiome data matrix and metabolome data matrix; The microbiome data matrix and metabolome data matrix are preprocessed to obtain preprocessed microbiome data matrix and preprocessed metabolome data matrix, respectively. The preprocessed microbiome data matrix is binarized to obtain a binarized microbiome data matrix. Based on the binarized microbiome data matrix, the similarity between any sample and any sample marked as healthy is calculated. The top P samples marked as healthy with the highest similarity to any sample are selected and a nearest neighbor weight matrix is constructed. A mapping matrix is constructed based on the nearest neighbor weight matrix. The mapping matrix is used to linearly combine the metabolomics data matrices of samples marked as healthy in the preprocessed metabolomics data matrix to obtain the projection matrix. The preprocessed metabolomics data matrix is corrected by using a projection matrix to obtain a corrected metabolomics data matrix.
2. The method for correcting metabolomics data based on microbial similarity structures according to claim 1, characterized in that, The preprocessing process includes: The microbiome data matrix and metabolome data matrix are filtered based on the proportion of missing values. Microbial variables and metabolite variables with a zero value proportion exceeding a preset threshold are deleted to obtain the preprocessed microbiome data matrix and the filtered metabolome data matrix. The zero values in the filtered metabolomics data matrix are imputed using the median imputation strategy, followed by log2 logarithmic transformation and unit variance standardization to obtain the preprocessed metabolomics data matrix.
3. The method for correcting metabolomics data based on microbial similarity structures according to claim 1, characterized in that, The preprocessed microbiome data matrix is binarized to obtain a binarized microbiome data matrix, specifically including: Binarization is performed using the following formula: ； in, This indicates the first element in the pretreated microbiome data matrix. The first sample Abundance values of individual microbial variables This represents the first element in the binarized microbiome data matrix. The first sample Binarized values of individual microbial variables .
4. The method for correcting metabolomics data based on microbial similarity structures according to claim 3, characterized in that, The similarity includes Jaccard similarity, calculated using the following formula: ； in, Represents any sample Compared with any sample marked as healthy Similarity between them Represents any sample labeled as healthy. The set of binarized values in the binarized microbiome data matrix. Represents any sample The set of binarized values in the binarized microbiome data matrix. .
5. The method for correcting metabolomics data based on microbial similarity structures according to claim 4, characterized in that, The process of constructing the nearest neighbor weight matrix is as follows: ； in, Indicates any sample The nearest neighbor set is constructed from the top P samples with the highest similarity, which are labeled as healthy. The nearest neighbor weight matrix represents the first nearest neighbor weight matrix. Line 1 Column elements, , ; The process of constructing the mapping matrix is as follows: Perform an exponential transformation on each element of the nearest neighbor weight matrix to obtain the transformed nearest neighbor weight matrix, as shown in the following equation: ； in, The nearest neighbor weight matrix after the transformation represents the first nearest neighbor weight matrix. Line 1 Column elements, For exponential scaling parameters; The transformed nearest neighbor weight matrix is normalized to obtain the mapping matrix, as shown in the following equation: in, Represents the first in the mapping matrix Line 1 The elements of the column.
6. The method for correcting metabolomics data based on microbial similarity structures according to claim 1, characterized in that, The calculation process of the projection matrix is as follows: ； in, This represents the metabolomics data matrix of samples marked as healthy in the preprocessed metabolomics data matrix. Represents the projection matrix. Represents the mapping matrix; The calculation process for the corrected metabolomics data matrix is as follows: ； in, This represents the preprocessed metabolomics data matrix. This represents the corrected metabolomics data matrix.
7. A metabolomics data correction device based on microbial similarity structures, characterized in that, include: The preprocessing module is configured to acquire microbiome and metabolome data of labeled samples collected and detected at the same time point in the same organism or environment, and to construct a matching microbiome data matrix and metabolome data matrix. The microbiome data matrix and metabolome data matrix are preprocessed to obtain preprocessed microbiome data matrix and preprocessed metabolome data matrix, respectively. The mapping matrix construction module is configured to binarize the preprocessed microbiome data matrix to obtain a binarized microbiome data matrix, calculate the similarity between any sample and any sample marked as healthy based on the binarized microbiome data matrix, select the top P samples marked as healthy with the highest similarity to any sample and construct a nearest neighbor weight matrix; and construct a mapping matrix based on the nearest neighbor weight matrix. The correction module is configured to use the mapping matrix to linearly combine the metabolomics data matrices of samples marked as healthy in the preprocessed metabolomics data matrix to obtain a projection matrix; The preprocessed metabolomics data matrix is corrected by using a projection matrix to obtain a corrected metabolomics data matrix.
8. An electronic device, comprising: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.