A method and system for metabolomics data analysis

Abstract

The application discloses a kind of metabolomics data analysis method and system, method includes: obtaining the metabolomics data of biological sample under multiple conditions, adopt the strategy of fusion data distribution characteristics to carry out missing value filling;The response intensity distribution difference value of each metabolite feature is calculated by sliding window local fluctuation analysis, and candidate differential metabolite feature is screened;The response value of each feature under different conditions is sorted by size and fitted as feature response curve, and clustering is carried out by calculating curve similarity to obtain feature cluster;Further based on the overall morphology and local synchronism fusion degree of measure between features, construct weighted network and carry out community division, from each subcommunity High coordination intensity feature pair is screened in community, and potential biomarker combination is formed.The application realizes the whole process analysis from data reconstruction, difference screening to collaborative marker identification, improves the accuracy, robustness and biological explainability of single sample metabolomics data mining.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and in particular relates to a metabolomics data analysis method and system. Background Technology

[0002] Metabolomics, as an important branch of systems biology, aims to qualitatively and quantitatively analyze all small-molecule metabolites in organisms, thereby directly reflecting the overall functional state of an organism under physiological, pathological, or external intervention conditions. Chromatography-mass spectrometry (GC-MS), with its high resolution and sensitivity, has become a core tool for acquiring large-scale, high-dimensional metabolomics data. However, accurately and efficiently identifying stable and highly interpretable potential biomarkers related to specific biological conditions (such as specific disease states, drug treatment responses, and different developmental stages) from massive amounts of complex raw data still faces a series of technical bottlenecks.

[0003] Current mainstream analytical workflows typically include data preprocessing, univariate differential analysis, and subsequent bioinformatics annotation. In the preprocessing stage, missing data is a common problem, and traditional methods often use simple statistical values ​​(such as minimum, mean, and median) or k-nearest neighbor algorithms for imputation. These methods often neglect the statistical distribution characteristics within metabolite features and the biological correlations between different features (such as co-expression or pathway co-occurrence relationships), which may lead to imputed values ​​deviating from true biological variation and introducing systematic bias into downstream analysis. Summary of the Invention

[0004] In view of this, the present invention provides a metabolomics data analysis method and system, which aims to solve the technical problem that neglecting the statistical distribution characteristics within metabolite features and the biological correlations between different features may lead to filled values ​​deviating from the true biological variation, introducing systematic biases into downstream analysis.

[0005] In a first aspect, the present invention provides a method for metabolomics data analysis, comprising:

[0006] A set of metabolomics data sequences of biological samples under different detection conditions is obtained, and the set of metabolomics data is filled with data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and a metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions.

[0007] The response intensity distribution difference value of each metabolite feature is determined based on the target metabolomics data sequence in the target metabolomics dataset, and it is determined whether the response intensity distribution difference value is greater than the preset difference value threshold.

[0008] If the difference in the distribution of the response intensity of a certain metabolite feature is greater than a preset difference threshold, then the certain metabolite feature is defined as a candidate differential metabolite feature, and a certain response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences is obtained. Each certain response intensity value is fitted to obtain a certain feature response curve of the candidate differential metabolite feature.

[0009] Calculate the curve similarity between a certain feature response curve of a candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and sum the curve similarities to obtain the similarity to a certain target curve of the candidate differential metabolite feature.

[0010] Cluster all candidate differential metabolite features based on the similarity of each target curve to obtain at least one candidate differential metabolite feature cluster.

[0011] Based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster, a preset pattern recognition strategy is used to identify potential biomarker combinations associated with the biological sample.

[0012] Secondly, the present invention provides a metabolomics data analysis system, comprising:

[0013] The filling module is configured to acquire a set of metabolomics data sequences of biological samples under different detection conditions, and fill the set of metabolomics data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and a metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions.

[0014] The judgment module is configured to determine the difference value of the response intensity distribution of each metabolite feature based on the target metabolomics data sequence in the target metabolomics dataset, and to determine whether the difference value of each response intensity distribution is greater than a preset difference value threshold.

[0015] The fitting module is configured to define a certain metabolite feature as a candidate differential metabolite feature if the difference value of the response intensity distribution of a certain metabolite feature is greater than a preset difference value threshold, and to obtain a certain response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences, and to fit each certain response intensity value to obtain a certain feature response curve of the candidate differential metabolite feature.

[0016] The calculation module is configured to calculate the curve similarity between a certain feature response curve of a certain candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and to sum the curve similarities to obtain the similarity to a certain target curve of the certain candidate differential metabolite feature.

[0017] The clustering module is configured to cluster all candidate differential metabolite features based on the similarity of each target curve, thereby obtaining at least one candidate differential metabolite feature cluster.

[0018] The identification module is configured to identify a combination of potential biomarkers associated with the biological sample by using a preset pattern recognition strategy based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster.

[0019] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the metabolomics data analysis method according to any embodiment of the present invention.

[0020] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the metabolomics data analysis method of any embodiment of the present invention.

[0021] The metabolomics data analysis method and system of this application significantly improve the accuracy of missing data recovery by combining an intelligent imputation strategy with data distribution characteristics, ensuring the reliability of subsequent analysis. Secondly, by using a sliding window local fluctuation algorithm, robust and accurate screening of differential metabolites is achieved under conditions without biological replication. Furthermore, by converting metabolite response patterns into comparable curves and performing clustering, functional groups of metabolites with similar response behaviors are effectively identified. Finally, networked collaborative analysis is introduced to automatically mine highly collaborative feature pairs from metabolite clusters, forming biomarker combinations with biological significance, greatly enhancing the interpretability and application value of the findings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1A flowchart illustrating a metabolomics data analysis method provided in an embodiment of the present invention;

[0024] Figure 2 This is a structural block diagram of a metabolomics data analysis system provided in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figure 1 The diagram shows a flowchart of a metabolomics data analysis method according to this application.

[0028] like Figure 1 As shown, the metabolomics data analysis method specifically includes the following steps:

[0029] Step S101: Obtain a set of metabolomics data sequences of biological samples under different detection conditions, and fill the set of metabolomics data with data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and a metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions.

[0030] In this step, it is determined whether there are missing response intensity values ​​in each metabolomics data sequence in the metabolomics dataset;

[0031] If a target response intensity value is missing in a certain metabolomics data sequence, a value is randomly selected from a non-missing response intensity value in other metabolomics data sequences as the baseline value. The other metabolomics data sequences are any metabolomics data sequences in the metabolomics dataset excluding the specific metabolomics data sequence, and the specific non-missing response intensity value and the specific target response intensity value both correspond to a specific metabolite feature.

[0032] The base value Initially fill in the missing position of a target response intensity value in a certain metabolomics data sequence to obtain the sequence. and according to the sequence Generate scaling factors for filling. ;

[0033] Based on the aforementioned base value The scaling factor is used to generate the final padding value using a preset transformation strategy, and this value is used as a response intensity value for a certain metabolomics data sequence, thus obtaining a complete metabolomics data sequence.

[0034] Wherein, according to the sequence Generate scaling factors for filling. include:

[0035] Determine the sequence Cumulative frequency distribution ,sequence skewness and sequences peak , where the calculation sequence Cumulative frequency distribution The expression is:

[0036] ,

[0037] In the formula, Base value In sequence rank in For sequence The total amount of elements;

[0038] Calculate sequence skewness The expression is:

[0039] ,

[0040] In the formula, For sequence The element mean, For sequence The i-th element in the equation, i.e., the i-th response intensity value. For sequence The sample standard deviation;

[0041] Calculate sequence peak The expression is:

[0042] ,

[0043] Calculate the base value In the sequence Local sample density The expression is:

[0044] ,

[0045] In the formula, The number of elements in the set. For sequence-based The preset half-width of the value range;

[0046] According to the cumulative frequency distribution The aforementioned skewness The kurtosis and the local sample density Calculate the scaling factor The expression is:

[0047] .

[0048] Specifically, the expression for the transformation strategy is:

[0049] ,

[0050] ,

[0051] In the formula, For fill value, The noise intensity coefficient, For random noise that follows a standard Gumbel distribution, For sequence standard deviation This represents the maximum standard deviation among other metabolomics data sequences. For sequence skewness.

[0052] Step S102: Determine the difference value of the response intensity distribution of each metabolite feature based on the target metabolomics data sequence in the target metabolomics dataset, and determine whether the difference value of each response intensity distribution is greater than a preset difference value threshold.

[0053] In this step, the target metabolomics data sequences in the target metabolomics dataset are aligned; a preset sliding window is slid across each target metabolomics data sequence, and a sliding window position corresponding to a certain metabolite feature is determined; at the certain sliding window position, the target response intensity values ​​in the sliding window are averaged to obtain a certain average target response intensity value; the difference between each target response intensity value in the sliding window and the certain average target response intensity value is calculated, and the difference results are summed to obtain a certain response intensity distribution difference value of a certain metabolite feature.

[0054] After determining whether the difference value of each response intensity distribution is greater than the preset difference value threshold, if the difference value of the response intensity distribution of a certain metabolite feature is not greater than the preset difference value threshold, then the certain metabolite feature is not defined as a candidate differential metabolite feature.

[0055] In one specific embodiment, after acquiring and completing data filling, step S102 is executed to determine the response stability or significance of each metabolite feature under different detection conditions. The core of this step lies in quantifying the dispersion of a metabolite feature's response intensity across all detection conditions, rather than simply comparing the mean. The specific implementation is as follows:

[0056] First, the target metabolomics data sequences (each sequence representing the response value of all metabolite features under a detection condition) in the padded target metabolomics dataset are aligned according to the order of the detection conditions to form a two-dimensional matrix, where rows represent metabolite features and columns represent detection conditions.

[0057] For each row in the matrix (i.e., a specific metabolite feature), the algorithm uses a sliding window of a preset width (e.g., the width of three consecutive detection conditions) to slide along the column direction (i.e., the detection condition dimension). At a certain position of the sliding window, the window covers the target response intensity values ​​of the metabolite feature under several consecutive detection conditions. The average of each target response intensity value in the sliding window is calculated to obtain a certain average target response intensity value. The difference between each target response intensity value in the sliding window and the certain average target response intensity value is calculated, and the difference results are summed to obtain a certain response intensity distribution difference value for a certain metabolite feature.

[0058] The difference in response intensity distribution reflects the degree of fluctuation in the characteristic response intensity of the metabolite under different detection conditions. The larger the value, the more unstable the response of the characteristic or the more sensitive it is to the detection conditions, and the more likely it is to be a condition-specific differential metabolite.

[0059] Finally, the calculated differences in the response intensity distribution of all metabolite features are compared with a preset difference threshold. This threshold can be a statistical value set based on the distribution of differences in all features, such as a function of the median and standard deviation of all response intensity distribution differences, or it can be an empirical value set based on prior knowledge. If the difference in the response intensity distribution is greater than the difference threshold, the metabolite feature corresponding to that difference in response intensity distribution is determined to be a candidate differentially expressed metabolite feature and proceeds to the next analysis step.

[0060] Step S103: If the difference in the distribution of the response intensity of a certain metabolite feature is greater than a preset difference threshold, then the certain metabolite feature is defined as a candidate differential metabolite feature, and a certain response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences is obtained. Each certain response intensity value is fitted to obtain a certain feature response curve of the candidate differential metabolite feature.

[0061] In this step, a target response intensity value of a candidate differential metabolite feature is obtained in different target metabolomics data sequences, and the target response intensity values ​​are sorted according to their numerical values ​​to obtain a target response intensity value sequence. The target response intensity value sequence is set in a preset two-dimensional coordinate system with the sequence number as the x-axis and the target response intensity value as the y-axis, and a smooth spline function is used to fit each coordinate point in the two-dimensional coordinate system to obtain a feature response curve of the candidate differential metabolite feature.

[0062] In one specific embodiment, after screening candidate differential metabolite features, step S103 aims to convert the response pattern of each feature under all detection conditions into a continuous and comparable curve for subsequent pattern similarity analysis. The specific implementation is as follows:

[0063] For each feature identified as a candidate differential metabolite characteristic First, this feature is extracted from the target metabolomics dataset under each detection condition. ( , The response intensity value under the total number of detection conditions. .

[0064] All according to the magnitude of the response intensity value Sort the values ​​in ascending order to generate a new sequence called the "target response intensity value sequence", denoted as . , For the first Each response intensity value, this sorting operation strips away the specific order of detection conditions (such as time order, concentration order), so that the curve only reflects the distribution pattern of the characteristic response value, rather than its change trajectory in a specific experimental procedure.

[0065] Next, using the sorted sequence indices (numbers 1, 2, ..., N) as the x-axis and the corresponding sorted response intensity values... As the ordinate, it forms a set of points in a two-dimensional coordinate system. .

[0066] Finally, a smooth spline function is used for the point set. Perform nonparametric fitting. The goal of smooth spline functions is to find a continuous and smooth function curve. While fitting the data points as closely as possible, a smoothing penalty term is introduced to control the curvature of the curve and prevent overfitting. This optimization problem is typically expressed as minimizing the following objective function:

[0067] ,

[0068] in, It is a smoothing parameter used to balance goodness of fit and curve smoothness. For the i-th response intensity value, The index of the i-th data point. For the fitting function On the horizontal axis The function value calculated at that point, For the fitting function The second derivative;

[0069] By solving, a smooth and continuous characteristic response curve is obtained. This curve is "a characteristic response curve of a certain candidate differential metabolite". The shape of the curve on the horizontal axis (sort number) intuitively shows whether the "path" of the characteristic response value from minimum to maximum is gentle, steep or has a plateau.

[0070] In summary, firstly, by stripping away the sequence information of detection conditions, the focus of analysis shifts from "condition-dependent instantaneous changes" to "global distribution patterns of response intensity," making the metabolite characteristics obtained under different experimental designs or disordered conditions (such as different reagents or pH levels) comparable, greatly enhancing the method's versatility and robustness. Secondly, using smooth splines for fitting effectively suppresses the interference of random noise in the original data and reveals the inherent continuous trend of the response intensity distribution (such as monotonically increasing, S-shaped growth, or the existence of a saturation plateau), thus transforming discrete data points, which may contain measurement errors, into a robust curve object containing mathematical characteristics. Finally, this curve representation provides an ideal data form for subsequent steps (such as calculating the similarity between curves), enabling the use of mature function data analysis or shape matching algorithms to quantitatively compare whether the global response behaviors of different metabolite characteristics are similar, laying a solid foundation for metabolite clustering based on functional or regulatory similarity.

[0071] Step S104: Calculate the curve similarity between a certain feature response curve of a certain candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and sum the curve similarities to obtain the similarity to a certain target curve of the certain candidate differential metabolite feature.

[0072] In this step, uniform sampling is performed on the certain characteristic response curve and other characteristic response curves to obtain a certain discrete sequence corresponding to the certain characteristic response curve, and other discrete sequences corresponding to the other characteristic response curves; the Pearson correlation coefficient between the certain discrete sequence and the other discrete sequences is calculated, and the curve similarity between the certain characteristic response curve of a certain candidate differential metabolite and other characteristic response curves of other candidate differential metabolites is calculated.

[0073] Step S105: Cluster all candidate differential metabolite features based on the similarity of each target curve to obtain at least one candidate differential metabolite feature cluster.

[0074] In this step, it is assumed that there are Each candidate differentially expressed metabolite has a characteristic, and the similarity of each characteristic to the target curve has been calculated. ( ). All Values ​​form a one-dimensional dataset , For the first Similarity of target curves.

[0075] For one-dimensional datasets Z-score standardization is performed to eliminate the influence of dimensions, resulting in a standardized value. .

[0076] K-means clustering: Set the number of clusters C (which can be determined by the elbow rule or silhouette coefficient), and... Using one-dimensional coordinates, the K-means algorithm is used for iterative optimization to divide the features into C clusters.

[0077] Based on the cluster labels output by the K-means clustering method, the original candidate differential metabolite features are divided into different clusters to obtain at least one candidate differential metabolite feature cluster.

[0078] Step S106: Based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster, a preset pattern recognition strategy is used to identify potential biomarker combinations associated with the biological sample.

[0079] In this step, based on the at least one candidate differential metabolite feature cluster, the response intensity values ​​of each candidate differential metabolite feature within each candidate differential metabolite feature cluster are extracted under the different detection conditions to form multiple detection condition-feature response curves.

[0080] Calculate the intensity of co-change between any two detection condition-feature response curves within the same candidate differential metabolite feature cluster. The intensity of co-change is based on a fusion measurement of the overall morphological similarity and local change synchronicity of the two detection condition-feature response curves across all detection conditions.

[0081] Based on the intensity of the coordinated change, an undirected weighted network is constructed with candidate differential metabolite features as nodes and coordinated change intensity as edge weights;

[0082] The undirected weighted network is divided into communities to identify at least one feature sub-community, wherein the features within each feature sub-community have a cooperative change intensity higher than a preset threshold.

[0083] For each feature sub-community, feature pairs with edge weights greater than a preset collaboration threshold or ranked in the top preset number are selected as candidate collaboration feature pairs.

[0084] Candidate collaborative feature pairs selected from all feature sub-communities are aggregated to form a combination of potential biomarkers related to the biological sample.

[0085] In one specific embodiment, for each candidate differential metabolite feature cluster Extract features of all candidate differentially metabolites within the cluster. The original response intensity values ​​under different detection conditions.

[0086] Cluster Include The first feature. For the first feature. Features , its in The response intensity values ​​under each detection condition constitute a vector. :

[0087]

[0088] in, Representation of features In the The response intensity under each detection condition.

[0089] For all For each feature, plot the curves of its response intensity as a function of the detection conditions, ordered by the detection conditions (rather than by intensity value), to form... Each curve is a "detection condition-feature response curve". To eliminate the magnitude difference in absolute response values ​​between different features and to facilitate synergistic analysis, each curve is Z-score standardized.

[0090] For any two features within the same feature cluster and Based on its standardized response curve and Calculate the intensity of their coordinated change. This intensity is defined as the overall morphological similarity. Synchronicity with local changes Weighted fusion:

[0091] ,

[0092] in, Weight and It can be set based on experience or through cross-validation, for example , To emphasize the overall form.

[0093] Overall morphological similarity The distance is measured using the reciprocal of the Dynamic Time Warping (DTW) distance. DTW can handle the non-linear scaling of two curves on the time (detection condition) axis, thus better capturing the similarity of the overall shape.

[0094]

[0095] in, , represent the standardized response intensity vectors of candidate differential metabolite feature i and candidate differential metabolite feature j, respectively, from the same feature cluster. DTW distance is the dynamic time-warped distance between the standardized response intensity vectors of candidate differential metabolite feature i and candidate differential metabolite feature j. The smaller the DTW distance, the more similar the overall morphology. The closer to 1;

[0096] Synchronicity of local changes Calculate the Pearson correlation coefficient between the first-order difference sequences of the two curves to measure the consistency of their pace of change. First, calculate the difference sequences of each curve:

[0097] ,

[0098] In the formula, for The first-order difference sequence, For candidate differential metabolite characteristics i in the first Standardized response intensity values ​​under each detection condition;

[0099] Similarly, we can obtain First-order difference sequence Then, the synchronicity of local changes is calculated. , This is the Pearson correlation coefficient.

[0100] With cluster Each candidate differential metabolite within the node is characterized as a node. For any two nodes i and j, if their cooperative change intensity... Greater than the preset connection threshold (For example, 0.3), then an edge is created between them, and the weight of the edge is... .like If no obvious cooperative relationship is found between the two, then no connection is established. This constructs an undirected weighted network. ,in It is a set of nodes. It is an edge set. It is a weight set.

[0101] For undirected weighted networks Community Detection aims to divide the network into several "communities" or "modules" with tight internal connections and sparse external connections, i.e., feature sub-communities.

[0102] This implementation uses the Louvain algorithm, which is an efficient heuristic algorithm based on modularity optimization.

[0103] Louvain's algorithm maximizes the modularity Q-value through iterative optimization, ultimately optimizing the network. Divided into Each feature sub-community has an average edge weight (cooperative change intensity) between nodes within it that is higher than the average level of the entire network, indicating that it has internal cooperativeness above a preset threshold.

[0104] Extract the actual edges and their weights between all nodes (features) within the feature sub-community.

[0105] Set a collaboration threshold (e.g., 0.6) and filter out feature pairs with edge weights greater than the collaboration threshold as high-confidence candidate collaboration feature pairs.

[0106] Alternatively, to avoid the subjectivity of threshold setting, a relative ranking method can be used: sort all edges within the feature sub-community from high to low weight, and select the top P pairs (e.g., the top 3 pairs or the top 10%) as candidate collaborative feature pairs.

[0107] Record the two feature identifiers contained in each candidate co-feature pair and the strength of their co-variation.

[0108] Traverse all feature clusters and all feature sub-communities within them, summarize and deduplicate all selected candidate co-operating feature pairs (if the same pair of features is repeatedly selected in different sub-communities or different clusters, retain its highest co-operating strength record), and finally form a global list of "candidate co-operating feature pairs".

[0109] This list represents a combination of potential biomarkers identified from the metabolomics data of this biological sample. The combination is centered on "metabolite feature pairs" and their synergistic strength, reflecting highly synergistic patterns of change in the response behavior of these metabolites under various given detection conditions. This suggests that they may participate in the same metabolic pathways, be co-regulated, or have functional associations, thus serving as a potential set of combined biomarkers to indicate a specific physiological or pathological state of the sample.

[0110] In summary, the method of this application, firstly, in the data preprocessing stage, effectively overcomes the filling distortion problem caused by neglecting data structure in traditional methods by introducing a dynamic filling strategy that integrates data distribution patterns, local density, and statistical characteristics. This significantly improves the fidelity and reliability of missing value restoration, laying a high-quality data foundation for subsequent analysis. Secondly, in the differential feature screening stage, a local fluctuation accumulation algorithm based on a sliding window is used instead of a simple global variance comparison. This can sensitively capture the mutation patterns of metabolite responses under specific condition combinations and remain robust to outliers, thereby achieving accurate and robust identification of condition-specific differential metabolites without the need for biological replication. Furthermore, by... The selected differential features were curve-fitted and similarity calculated based on the distribution pattern of their response intensity values ​​(rather than conditional order), successfully removing interference from experimental sequences and transforming metabolite response patterns into quantifiable and comparable mathematical objects. Further clustering analysis was used to group these features into several feature clusters with similar global response behaviors, achieving efficient and reasonable classification of metabolite functional groups. Finally, in the biomarker identification stage, a "co-variation intensity" metric was introduced. Combining overall morphology and local synchronicity, network construction and community division were performed, automatically mining highly co-operative feature sub-communities from each feature cluster and selecting metabolite feature pairs with high co-operative intensity as potential biomarker combinations. The final output of this method is no longer an isolated list of metabolites, but a combination of metabolite pairs containing functional association information, greatly enhancing the biological interpretability of the findings. Overall, this method systematically solves a series of technical challenges in handling missing values, screening differential markers, and discovering synergistic biomarkers in single-sample multi-condition metabolomics data. It provides an automated, high-fidelity analysis path from raw data to high-value biomarker combinations, and offers a powerful data mining tool for research in precision medicine, disease subtyping, and drug response.

[0111] Please see Figure 2 The diagram shows a structural block diagram of a metabolomics data analysis system according to this application.

[0112] like Figure 2 As shown, the metabolomics data analysis system 200 includes a filling module 210, a judgment module 220, a fitting module 230, a calculation module 240, a clustering module 250, and an identification module 260.

[0113] The filling module 210 is configured to acquire a set of metabolomics data sequences of biological samples under different detection conditions, and fill the set of metabolomics data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and each metabolomics data sequence contains response intensity values ​​of different metabolite features under the same detection conditions. The judgment module 220 is configured to determine the difference value of the response intensity distribution of each metabolite feature based on the target metabolomics data sequence in the target metabolomics data set, and determine whether the difference value of each response intensity distribution is greater than a preset difference value threshold. The fitting module 230 is configured to define the metabolite feature as a candidate differential metabolite feature if the difference value of the response intensity distribution of a certain metabolite feature is greater than the preset difference value threshold, and obtain the candidate differential metabolite feature. The system uses a response intensity value of a metabolite feature in different target metabolomics data sequences, fits each response intensity value to obtain a feature response curve of the candidate differential metabolite feature; a calculation module 240 is configured to calculate the curve similarity between the feature response curve of the candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and sums the curve similarities to obtain a target curve similarity with the candidate differential metabolite feature; a clustering module 250 is configured to cluster all candidate differential metabolite features according to the target curve similarity to obtain at least one candidate differential metabolite feature cluster; and an identification module 260 is configured to identify potential biomarker combinations associated with the biological sample based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster using a preset pattern recognition strategy.

[0114] It should be understood that Figure 2 The modules and references described in the document Figure 1 The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0115] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the metabolomics data analysis method in any of the above method embodiments.

[0116] In one embodiment, the computer-readable storage medium of the present invention stores computer-executable instructions, which are configured as follows:

[0117] A set of metabolomics data sequences of biological samples under different detection conditions is obtained, and the set of metabolomics data is filled with data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and a metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions.

[0118] The response intensity distribution difference value of each metabolite feature is determined based on the target metabolomics data sequence in the target metabolomics dataset, and it is determined whether the response intensity distribution difference value is greater than the preset difference value threshold.

[0119] If the difference in the distribution of the response intensity of a certain metabolite feature is greater than a preset difference threshold, then the certain metabolite feature is defined as a candidate differential metabolite feature, and a certain response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences is obtained. Each certain response intensity value is fitted to obtain a certain feature response curve of the candidate differential metabolite feature.

[0120] Calculate the curve similarity between a certain feature response curve of a candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and sum the curve similarities to obtain the similarity to a certain target curve of the candidate differential metabolite feature.

[0121] Cluster all candidate differential metabolite features based on the similarity of each target curve to obtain at least one candidate differential metabolite feature cluster.

[0122] Based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster, a preset pattern recognition strategy is used to identify potential biomarker combinations associated with the biological sample.

[0123] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the metabolomics data analysis system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the metabolomics data analysis system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0124] Figure 3This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the metabolomics data analysis method described in the above embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the metabolomics data analysis system. The output device 340 may include a display screen or other display device.

[0125] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0126] In one implementation, the above-described electronic device is used in a metabolomics data analysis system as a client, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0127] A set of metabolomics data sequences of biological samples under different detection conditions is obtained, and the set of metabolomics data is filled with data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and a metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions.

[0128] The response intensity distribution difference value of each metabolite feature is determined based on the target metabolomics data sequence in the target metabolomics dataset, and it is determined whether the response intensity distribution difference value is greater than the preset difference value threshold.

[0129] If the difference in the distribution of the response intensity of a certain metabolite feature is greater than a preset difference threshold, then the certain metabolite feature is defined as a candidate differential metabolite feature, and a certain response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences is obtained. Each certain response intensity value is fitted to obtain a certain feature response curve of the candidate differential metabolite feature.

[0130] Calculate the curve similarity between a certain feature response curve of a candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and sum the curve similarities to obtain the similarity to a certain target curve of the candidate differential metabolite feature.

[0131] Cluster all candidate differential metabolite features based on the similarity of each target curve to obtain at least one candidate differential metabolite feature cluster.

[0132] Based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster, a preset pattern recognition strategy is used to identify potential biomarker combinations associated with the biological sample.

[0133] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0134] 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 of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing metabolomics data, characterized in that, include: A set of metabolomics data sequences of biological samples under different detection conditions is obtained, and the set of metabolomics data is filled according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and a metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions, specifically including: Determine whether there are missing response intensity values ​​for each metabolomics data sequence in the metabolomics dataset; If a target response intensity value is missing in a certain metabolomics data sequence, a value is randomly selected from a non-missing response intensity value in other metabolomics data sequences as the baseline value. The other metabolomics data sequences are any metabolomics data sequences in the metabolomics dataset excluding the specific metabolomics data sequence, and the specific non-missing response intensity value and the specific target response intensity value both correspond to a specific metabolite feature. The base value Initially fill in the missing position of a target response intensity value in a certain metabolomics data sequence to obtain the sequence. and according to the sequence Generate scaling factors for filling. ; Based on the aforementioned base value The scaling factor is used to generate the final padding value using a preset transformation strategy, and this value is used as a response intensity value of a certain metabolomics data sequence to obtain a complete metabolomics data sequence. The response intensity distribution difference value of each metabolite feature is determined based on the target metabolomics data sequence in the target metabolomics dataset, and it is determined whether the response intensity distribution difference value is greater than the preset difference value threshold. If the difference in the response intensity distribution of a certain metabolite feature is greater than a preset difference threshold, then the certain metabolite feature is defined as a candidate differential metabolite feature. A certain response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences is obtained, and each certain response intensity value is fitted to obtain a certain feature response curve of the candidate differential metabolite feature. Specifically, this includes: Obtain the target response intensity value of a certain candidate differential metabolite feature in different target metabolomics data sequences, and sort the target response intensity values ​​according to their numerical values ​​to obtain a target response intensity value sequence; Using the sequence number in the target response intensity value sequence as the horizontal axis and the target response intensity value in the target response intensity value sequence as the vertical axis, the target response intensity value sequence is set in a preset two-dimensional coordinate system, and a smooth spline function is used to fit each coordinate point in the two-dimensional coordinate system to obtain a characteristic response curve of the candidate differential metabolite. Calculate the curve similarity between a certain feature response curve of a candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and sum the curve similarities to obtain the similarity to a certain target curve of the candidate differential metabolite feature. Cluster all candidate differential metabolite features based on the similarity of each target curve to obtain at least one candidate differential metabolite feature cluster. Based on the response intensity values ​​of all candidate differentially expressed metabolite features within the at least one candidate differentially expressed metabolite feature cluster, a preset pattern recognition strategy is used to identify potential biomarker combinations associated with the biological sample, specifically including: Based on the at least one candidate differential metabolite feature cluster, the response intensity values ​​of each candidate differential metabolite feature within each candidate differential metabolite feature cluster are extracted under different detection conditions to form multiple detection condition-feature response curves. Calculate the intensity of co-change between any two detection condition-feature response curves within the same candidate differential metabolite feature cluster. The intensity of co-change is based on a fusion measurement of the overall morphological similarity and local change synchronicity of the two detection condition-feature response curves across all detection conditions. Based on the intensity of the coordinated change, an undirected weighted network is constructed with candidate differential metabolite features as nodes and coordinated change intensity as edge weights; The undirected weighted network is divided into communities to identify at least one feature sub-community, wherein the features within each feature sub-community have a cooperative change intensity higher than a preset threshold. For each feature sub-community, feature pairs with edge weights greater than a preset collaboration threshold or ranked in the top preset number are selected as candidate collaboration feature pairs. Candidate collaborative feature pairs selected from all feature sub-communities are aggregated to form a combination of potential biomarkers related to the biological sample.

2. The metabolomics data analysis method according to claim 1, characterized in that, in, According to the sequence Generate scaling factors for filling. include: Determine the sequence Cumulative frequency distribution ,sequence skewness and sequences peak , where the calculation sequence Cumulative frequency distribution The expression is: ， In the formula, Base value In sequence rank in For sequence The total amount of elements; Calculate sequence skewness The expression is: ， In the formula, For sequence The element mean, For sequence The i-th element in the equation, i.e., the i-th response intensity value. For sequence The sample standard deviation; Calculate sequence peak The expression is: ， Calculate the base value In the sequence Local sample density The expression is: ， In the formula, The number of elements in the set. For sequence-based The preset half-width of the value range; According to the cumulative frequency distribution The aforementioned skewness The kurtosis and the local sample density Calculate the scaling factor The expression is: 。 3. The metabolomics data analysis method according to claim 1, characterized in that, in, The expression for the transformation strategy is: ， ， In the formula, For fill value, The noise intensity coefficient, For random noise that follows a standard Gumbel distribution, For sequence standard deviation This represents the maximum standard deviation among other metabolomics data sequences. For sequence skewness.

4. The metabolomics data analysis method according to claim 1, characterized in that, The step of determining the difference in response intensity distribution of each metabolite feature based on the target metabolomics data sequence in the target metabolomics dataset includes: Align the individual target metabolomics data sequences in the target metabolomics dataset; Based on a preset sliding window, slide it on each target metabolomics data sequence and determine a sliding window position corresponding to a certain metabolite feature. At the certain sliding window position, the target response intensity values ​​of each target in the sliding window are averaged to obtain a certain average target response intensity value. The difference between each target response intensity value in the sliding window and a certain average target response intensity value is calculated, and the difference results are summed to obtain the difference value of a certain response intensity distribution of a certain metabolite characteristic.

5. The metabolomics data analysis method according to claim 1, characterized in that, The curve similarity between a certain feature response curve of a candidate differential metabolite and other feature response curves of other candidate differential metabolites includes: Uniform sampling is performed on the certain characteristic response curve and other characteristic response curves to obtain a certain discrete sequence corresponding to the certain characteristic response curve, and other discrete sequences corresponding to the other characteristic response curves; Calculate the Pearson correlation coefficient between a certain discrete sequence and the other discrete sequences, and the curve similarity between a certain feature response curve of a candidate differential metabolite and other feature response curves of other candidate differential metabolites.

6. A metabolomics data analysis system, characterized in that, include: The data filling module is configured to acquire a set of metabolomics data sequences of biological samples under different detection conditions, and to fill the set of metabolomics data according to a preset data filling rule to obtain a target metabolomics data set. The set of metabolomics data sequences contains at least one metabolomics data sequence, and each metabolomics data sequence contains response intensity values ​​of different metabolite characteristics under the same detection conditions. Specifically, it includes: Determine whether there are missing response intensity values ​​for each metabolomics data sequence in the metabolomics dataset; If a target response intensity value is missing in a certain metabolomics data sequence, a value is randomly selected from a non-missing response intensity value in other metabolomics data sequences as the baseline value. The other metabolomics data sequences are any metabolomics data sequences in the metabolomics dataset excluding the specific metabolomics data sequence, and the specific non-missing response intensity value and the specific target response intensity value both correspond to a specific metabolite feature. The base value Initially fill in the missing position of a target response intensity value in a certain metabolomics data sequence to obtain the sequence. and according to the sequence Generate scaling factors for filling. ; Based on the aforementioned base value The scaling factor is used to generate the final padding value using a preset transformation strategy, and this value is used as a response intensity value of a certain metabolomics data sequence to obtain a complete metabolomics data sequence. The judgment module is configured to determine the difference value of the response intensity distribution of each metabolite feature based on the target metabolomics data sequence in the target metabolomics dataset, and to determine whether the difference value of each response intensity distribution is greater than a preset difference value threshold. The fitting module is configured to define a metabolite feature as a candidate differential metabolite feature if the difference in the distribution of the response intensity of a certain metabolite feature is greater than a preset difference threshold. It then acquires a response intensity value of the candidate differential metabolite feature in different target metabolomics data sequences, fits each response intensity value to obtain a feature response curve of the candidate differential metabolite feature, specifically including: Obtain the target response intensity value of a certain candidate differential metabolite feature in different target metabolomics data sequences, and sort the target response intensity values ​​according to their numerical values ​​to obtain a target response intensity value sequence; Using the sequence number in the target response intensity value sequence as the horizontal axis and the target response intensity value in the target response intensity value sequence as the vertical axis, the target response intensity value sequence is set in a preset two-dimensional coordinate system, and a smooth spline function is used to fit each coordinate point in the two-dimensional coordinate system to obtain a characteristic response curve of the candidate differential metabolite. The calculation module is configured to calculate the curve similarity between a certain feature response curve of a certain candidate differential metabolite feature and other feature response curves of other candidate differential metabolite features, and to sum the curve similarities to obtain the similarity to a certain target curve of the certain candidate differential metabolite feature. The clustering module is configured to cluster all candidate differential metabolite features based on the similarity of each target curve, thereby obtaining at least one candidate differential metabolite feature cluster. The identification module is configured to identify potential biomarker combinations associated with the biological sample based on the response intensity values ​​of all candidate differential metabolite features within the at least one candidate differential metabolite feature cluster, using a preset pattern recognition strategy. Specifically, this includes: Based on the at least one candidate differential metabolite feature cluster, the response intensity values ​​of each candidate differential metabolite feature within each candidate differential metabolite feature cluster are extracted under different detection conditions to form multiple detection condition-feature response curves. Calculate the intensity of co-change between any two detection condition-feature response curves within the same candidate differential metabolite feature cluster. The intensity of co-change is based on a fusion measurement of the overall morphological similarity and local change synchronicity of the two detection condition-feature response curves across all detection conditions. Based on the intensity of the coordinated change, an undirected weighted network is constructed with candidate differential metabolite features as nodes and coordinated change intensity as edge weights; The undirected weighted network is divided into communities to identify at least one feature sub-community, wherein the features within each feature sub-community have a cooperative change intensity higher than a preset threshold. For each feature sub-community, feature pairs with edge weights greater than a preset collaboration threshold or ranked in the top preset number are selected as candidate collaboration feature pairs. Candidate collaborative feature pairs selected from all feature sub-communities are aggregated to form a combination of potential biomarkers related to the biological sample.

Images