A method and system for screening for markers of differential lipid metabolism in human tissues

Abstract

The application relates to the technical field of medical data processing, and discloses a screening method and system for differential lipid metabolism markers in human tissues. The absolute value of the grade correlation coefficient of a differential lipid expression sequence in a standardization data matrix is calculated; a lipid coordination adjacency matrix and a graph Laplacian matrix are constructed by fusing node diagnosis penalty factors; a core hub marker cluster is determined by extracting a Fiedler value parameter and a feature vector centrality parameter; a diagnosis model is established, and a diagnosis error reverse gradient is extracted when a threshold is not reached; an absolute safe perturbation step threshold is defined in combination with an eigenvalue offset radius; and an optimization model is constructed to update the penalty factor and reconstruct the matrix for iteration. The method can at least solve the problem that traditional marker screening mostly adopts static network one-way extraction, and due to the lack of a bottom topology correction mechanism for clinical error feedback, a high false positive rate is caused in clinical verification.

Description

Technical Field

[0001] This application relates to the field of medical data processing technology, and in particular to a method and system for screening differential lipid metabolism biomarkers in human tissues. Background Technology

[0002] With the development of metabolomics technology, graph-based biological network analysis has been widely used in the screening of differential lipid metabolism biomarkers in human tissues. Traditional lipid metabolism biomarker screening methods mostly use static network feature extraction and one-way output mode for calculation.

[0003] However, the inventors have found at least the following technical problems in the related technologies: traditional screening networks lack a clinical error closed-loop feedback reconstruction mechanism with boundary constraints to prevent topological breakage. When blindly performing weight parameter tuning, it is very easy to cause structural collapse of the underlying pathological metabolic pathways, resulting in a high false positive rate of clinical marker combinations output by screening. Summary of the Invention

[0004] To address the shortcomings of existing technologies, one objective of this application is to provide a method for screening differential lipid metabolism biomarkers in human tissues, at least to solve the problem that traditional biomarker screening mostly uses static network unidirectional extraction, which results in a very high false positive rate in clinical validation due to the lack of a low-level topology correction mechanism for clinical error feedback.

[0005] To achieve the above objectives, some embodiments of this application provide the following aspects:

[0006] In a first aspect, some embodiments of this application provide a method for screening differential lipid metabolism biomarkers in human tissues, comprising the following steps:

[0007] S1. Calculate the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the standardized data matrix;

[0008] S2. The absolute value of the rank correlation coefficient is fused with the node diagnostic penalty factor to construct a lipid cooperative adjacency matrix, and a graph Laplacian matrix is ​​generated.

[0009] S3. Extract the Fiedler value parameter of the graph Laplacian matrix and the eigenvector centrality parameter of the lipid co-adjacency matrix, and extract the core hub marker cluster;

[0010] S4. Establish a diagnostic model based on the core hub marker cluster to obtain false positive rate parameters and diagnostic evaluation indicators. If the preset qualified threshold is not reached, calculate the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor to extract the diagnostic error back gradient.

[0011] S5. Combine the eigenvalue offset radius, the Fiedler value parameter, and the preset minimum value to define the absolute safety perturbation step size threshold.

[0012] S6. Construct an optimization model with the goal of minimizing the false positive rate parameter and constrained by the absolute safety perturbation step size threshold and the preset minimum value. Solve the incremental update of the node diagnostic penalty factor, reconstruct the lipid collaborative adjacency matrix and iterate until the diagnostic evaluation index reaches the preset qualified threshold, and output the core hub marker cluster.

[0013] Secondly, some embodiments of this application also provide a screening system for differential lipid metabolism biomarkers in human tissues, including the following modules:

[0014] The basic correlation calculation module is used to calculate the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the standardized data matrix;

[0015] The topology network initialization module is used to fuse the absolute value of the rank correlation coefficient with the node diagnostic penalty factor to construct a lipid cooperative adjacency matrix and generate a graph Laplacian matrix.

[0016] The centrality feature extraction module is used to extract the Fiedler value parameter of the graph Laplacian matrix and the eigenvector centrality parameter of the lipid co-adjacency matrix, and to extract the core hub marker cluster;

[0017] The diagnostic error extraction module is used to establish a diagnostic model based on the core hub marker cluster to obtain false positive rate parameters and diagnostic evaluation indicators. If the preset qualified threshold is not reached, the module calculates the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor to extract the back gradient of the diagnostic error.

[0018] The disturbance boundary definition module is used to define an absolute safe disturbance step size threshold by combining the eigenvalue offset radius, the Fiedler value parameter and the preset minimum value.

[0019] The constrained optimization iteration module is used to construct an optimization model with the objective of minimizing the false positive rate parameter and constrained by the absolute safety perturbation step size threshold and the preset minimum value. It solves the incremental update of the node diagnostic penalty factor, reconstructs the lipid collaborative adjacency matrix and iterates until the diagnostic evaluation index reaches the preset qualified threshold, and outputs the core hub marker cluster.

[0020] Thirdly, some embodiments of this application also provide an electronic device comprising: one or more processors; and a memory storing computer program instructions, which, when executed, cause the processor to perform steps of a method for screening differential lipid metabolism biomarkers in human tissues.

[0021] Fourthly, some embodiments of this application also provide a computer-readable medium having computer program instructions stored thereon, which can be executed by a processor to implement the method for screening differential lipid metabolism biomarkers in human tissues.

[0022] Compared with related technologies, the solution provided in this application reconstructs the network by combining the diagnostic error back gradient and the perturbation boundary iterative update of the penalty factor, thereby achieving closed-loop optimization to prevent breakage. This improves the problem that traditional biomarker screening mostly uses static networks for unidirectional extraction, which lacks a bottom-level topology correction mechanism for clinical error feedback, resulting in an extremely high false positive rate in clinical validation. Attached Figure Description

[0023] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0024] Figure 1 An exemplary flowchart of a method for screening differential lipid metabolism biomarkers in human tissues, provided for some embodiments;

[0025] Figure 2 A schematic diagram of the architecture of a screening system for differential lipid metabolism biomarkers in human tissues, provided for some embodiments;

[0026] Figure 3 A flowchart of data preprocessing and baseline extraction for a method of screening differential lipid metabolism biomarkers in human tissues, provided for some embodiments;

[0027] Figure 4 A flowchart illustrating the constrained optimization and iterative closed-loop mechanism of a method for screening differential lipid metabolism biomarkers in human tissues, provided for some embodiments;

[0028] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] First Embodiment

[0031] The first embodiment relates to a method for screening differential lipid metabolism biomarkers in human tissues. The core of this embodiment is:

[0032] S1. Calculate the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the standardized data matrix;

[0033] S2. The absolute value of the rank correlation coefficient is fused with the node diagnostic penalty factor to construct a lipid cooperative adjacency matrix, generating a graph Laplacian matrix;

[0034] S3. Extract the Fidler value parameters of the graph Laplacian matrix and the eigenvector centrality parameters of the lipid co-adjacency matrix, and extract the core hub marker clusters;

[0035] S4. Establish a diagnostic model based on the cluster of core hub markers to obtain false positive rate parameters and diagnostic evaluation indicators. If the preset qualified threshold is not reached, calculate the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor to extract the back gradient of the diagnostic error.

[0036] S5. Combine the eigenvalue offset radius, Fiedler value parameter and preset minimum value to define the absolute safe perturbation step size threshold;

[0037] S6. Construct an optimization model with the goal of minimizing the false positive rate parameter and constrained by the absolute safety perturbation step size threshold and the preset minimum value. Solve the incremental update node diagnostic penalty factor, reconstruct the lipid collaborative adjacency matrix and iterate until the diagnostic evaluation index reaches the preset qualified threshold, and output the core hub biomarker cluster.

[0038] By combining the diagnostic error back gradient with the perturbation boundary to iteratively update the penalty factor and reconstruct the network, a closed-loop optimization to prevent breakage is achieved. This improves the problem that traditional biomarker screening mostly uses static networks for unidirectional extraction, which lacks a bottom-level topology correction mechanism for clinical error feedback, resulting in an extremely high false positive rate in clinical validation.

[0039] The following is a detailed description of the implementation details of a method for screening differential lipid metabolism biomarkers in human tissues according to this embodiment. The following content is only for the convenience of understanding and is not necessary for implementing this solution.

[0040] The following combination Figures 1 to 4 Each of the above steps will be explained in detail.

[0041] Step S1 in this embodiment specifically includes:

[0042] In S1, the absolute values ​​of the rank correlation coefficients of differentially expressed lipid sequences in the normalized data matrix are calculated, including:

[0043] Extract raw metabolomics data from human tissue samples;

[0044] Missing value reconstruction and normalization were performed on the raw metabolomics data to generate a standardized data matrix;

[0045] Extract pairwise differentially expressed lipid molecule sequences from the standardized data matrix;

[0046] Calculate the rank correlation coefficients of pairwise differentially expressed lipid sequences, and take the absolute value of the rank correlation coefficients to generate the absolute value of the rank correlation coefficients.

[0047] In practice, it can receive raw metabolomics data of human tissue samples from medical testing terminals. After the data preprocessing module executes missing value reconstruction and normalization instructions, it outputs a standardized data matrix to the feature mapping module.

[0048] After obtaining the raw metabolomics data of human tissue samples, the blank matrix interference response caused by the instrument detection limit can be removed, the missing sites can be filled by the local minimum detection value, and the total ion current can be normalized by combining the response intensity of the internal standard to generate a standardized data matrix that eliminates batch effects.

[0049] The underlying feature mapping module iterates through the standardized data matrix and extracts the relative concentration expression sequences of different differentially expressed lipid molecules. The Spearman rank correlation algorithm engine can be invoked to analyze the co-oscillation intensity of each pair of expressed sequences and calculate the initial rank correlation coefficient. After obtaining the initial rank correlation index, a sign erasure command can be triggered to perform an absolute value transformation operation on all the initial rank correlation indices, thereby generating non-negative absolute values ​​of the rank correlation coefficients.

[0050] It should be noted that the specific calculation formula for the Spearman rank correlation coefficient is common knowledge in the field, and those skilled in the art can implement it based on existing technology. This embodiment does not impose any specific limitations on it.

[0051] Conventional biological network construction methods directly use the original peak area to calculate Euclidean distance, which is easily biased by high-frequency noise caused by the heterogeneity of human tissue matrix. This step uses a preprocessing method combining baseline filling and normalization to filter out instrumental physical noise. By using the rank correlation coefficient, a nonparametric statistical property, instead of linear regression, the nonlinear cascade response relationship of lipid molecules in complex metabolic pathways can be quantified.

[0052] In nature, lipid molecules exhibit not only positive correlations of co-expression and co-accumulation, but also negative correlations of substrate competition and depletion. By forcibly applying an absolute value transformation, the positive co-expression metabolic intensity and the negative competitive metabolic intensity are uniformly mapped to unrestricted scalar edge weights. This eliminates sign interference in metabolic directions, transforming all objectively existing strong biological correlations into non-negative values. This directly satisfies the underlying theorem premise that subsequent steps for solving eigenvectors must rely on non-negative matrices, avoiding the problem of negative weights introduced by the underlying matrix data causing imaginary number anomalies or computational crashes in graph-based network eigenvalue decomposition.

[0053] Step S2 in this embodiment specifically includes:

[0054] In S2, the absolute value of the rank correlation coefficient and the node diagnostic penalty factor are fused to construct the lipid cooperative adjacency matrix, generating the graph Laplacian matrix, which includes:

[0055] The absolute value of the rank correlation coefficient and the corresponding node diagnosis penalty factor are input into a preset non-negative mapping function to calculate the dynamic connection edge weights.

[0056] A lipid cooperative adjacency matrix is ​​constructed based on the dynamic connection edge weights of all differential lipid molecules;

[0057] Calculate the sum of the generation degree matrix of the elements in the corresponding rows of a single differential lipid molecule within the lipid cooperative adjacency matrix;

[0058] The degree matrix is ​​subtracted from the lipid co-adjacency matrix to generate the graph Laplacian matrix.

[0059] Specifically, it can receive the absolute value of the rank correlation coefficient output from step S1 and retrieve the initially assigned node diagnostic penalty factor from memory as input data. The topology network initialization module performs edge weight mapping and matrix difference operations, and finally outputs the lipid cooperative adjacency matrix and the graph Laplacian matrix representing the network connectivity properties to the centrality feature extraction module, which serve as the data base for the next-level centrality evaluation algorithm.

[0060] The system receives the absolute value of the rank correlation coefficient output in step S1 and retrieves the initially assigned node diagnostic penalty factor from memory as input data. The dynamic connection edge weights can be calculated using a non-negative mapping function, the mathematical model of which is:

[0061] ;

[0062] in, Indicates the weight of dynamically connected edges; This represents the absolute value of the rank correlation coefficient between any two differentially expressed lipid molecules. This represents the base of the natural logarithm; and These represent the node diagnosis penalty factors for each pair of corresponding nodes. The underlying layer can restrict the initial value of these node diagnosis penalty factors to strictly 0 in the first loop state, and the range of these values ​​is strictly limited to a non-negative real number interval at the computer's underlying level. To prevent excessive updates during iterations from exceeding limits, boundary truncation can be performed after each parameter update: if the updated penalty factor is less than zero, the underlying layer will forcibly set the value to zero to ensure the physical meaning of the topology network.

[0063] The complete lipid cooperative adjacency matrix is ​​constructed based on all extracted dynamic connection edge weights. A tensor operation unit can be invoked to extract the row vectors corresponding to individual moleculees from the lipid cooperative adjacency matrix and perform numerical accumulation to generate the degree matrix, with off-diagonal elements of the degree matrix set to zero. Next, a matrix subtraction parser can be invoked to subtract the lipid cooperative adjacency matrix from the degree matrix, outputting the graph Laplacian matrix for subsequent topological feature extraction.

[0064] By embedding the node diagnostic penalty factor into a non-negative mapping function with a natural constant as the base, a lipid cooperative adjacency matrix with dynamic boundaries is constructed and a graph Laplacian matrix is ​​generated. This can improve the technical defects of traditional network construction algorithms that rely solely on static concentration to execute hard-coded connections, resulting in a lack of elastic anti-interference mechanisms at the algorithm's underlying layer. This can avoid the problem of abnormal calculation of negative edge weights or the occurrence of fault-like topological breaks in the overall pathological metabolic pathway when the topological network receives subsequent clinical error gradients to execute node penalties.

[0065] Step S3 in this embodiment specifically includes:

[0066] In S3, the Fiedler value parameter of the graph Laplacian matrix and the eigenvector centrality parameter of the lipid co-adjacency matrix are extracted, and the core hub marker clusters are extracted, including:

[0067] Perform eigenvalue decomposition on the graph Laplacian matrix to extract the eigenvalue sequence. The eigenvalue sequence with the second smallest value is used as the Fiedler value parameter.

[0068] Solve for the strictly positive eigenvector corresponding to the largest real eigenvalue of the lipid co-adjacency matrix;

[0069] Each component of the strictly positive eigenvector is extracted as the eigenvector centrality parameter of the corresponding differential lipid molecule;

[0070] The eigenvector centrality parameters of all differential lipid molecules are sorted in descending order, and the molecular combinations that rank in the preset highest interval are extracted to form a core hub marker cluster.

[0071] Specifically, a linear algebra parser can be invoked to perform characteristic polynomial decomposition on the graph Laplacian matrix, extract all eigenvalues ​​and arrange them in ascending order of numerical value to establish an eigenvalue sequence. The underlying instruction can directly extract the second smallest eigenvalue in the sequence and assign it as the Fiedler value parameter to establish the connectivity benchmark of the global network.

[0072] Simultaneously, for the extracted lipid co-adjacency matrix, the largest real eigenvalue and its corresponding strictly positive eigenvector can be obtained by solving the characteristic equation. The strictly positive eigenvector can be decomposed along the node dimension to obtain the independent component value of each differential lipid molecule mapping, and this component is defined as the eigenvector centrality parameter.

[0073] It should be noted that the specific numerical algorithms for solving the eigenvalues ​​and eigenvectors of a matrix (such as the power iteration method, QR algorithm, etc.) are common knowledge in the field and can be implemented by those skilled in the art based on existing technology. This embodiment does not make specific limitations on this.

[0074] The eigenvector centrality parameters of all differentially expressed lipid molecules are obtained and sorted in descending order. A preset cutoff range for the highest interval can be defined, limiting the range to 10% to 20% of the top values ​​in the sorting. Differentially expressed lipid molecules falling within the preset highest interval are extracted and assembled to generate the final core pivot marker cluster.

[0075] Conventional bioinformatics biomarker screening algorithms often assess node importance by statistically counting the absolute number of connections of isolated nodes—a superficial approach that can easily misidentify free lipids exhibiting nonspecific high-frequency fluctuations at the edges of metabolic networks as core pathogenic targets. This step, by pre-extracting the Fiedler value parameter of the graph Laplacian matrix in the underlying algorithm, can establish a baseline for preventing disruptions in the global algebraic connectivity of the entire biochemical metabolic cascade pathway.

[0076] For the calculation of strictly positive eigenvectors in lipid co-adjacency matrices, the centrality transitivity mapping mechanism can be used to identify pivotal metabolite nodes deeply intertwined with core pathogenic pathways. This can eliminate statistical concentration artifacts caused by isolated free lipids, cut off the data channel from high-false-positive edge nodes to downstream clinical validation cohorts, and solve the problem of poor clinical cohort validation results caused by traditional bioinformatics dimensionality reduction algorithms screening targets that are detached from the underlying biochemical cascade properties.

[0077] Furthermore, in this embodiment, due to the biological matrix heterogeneity of human tissue samples, some differentially expressed lipid molecules may exhibit outlier high concentrations, causing a global topological baseline shift during the characteristic polynomial decomposition process of the standard graph Laplacian matrix. To eliminate the interference of outlier lipid molecules on characteristic decomposition, this step can introduce a non-homogeneous mean operator on the graph Laplacian matrix before performing eigenvalue decomposition. Implement centralized benchmark alignment correction and construct a non-homogeneous graph Laplacian matrix. .

[0078] For example, the mathematical correction model for the non-homogeneous mean operator is as follows:

[0079] ;

[0080] in, This is the corrected Laplace matrix for the non-homogeneous graph. The graph is a Laplace matrix; for The identity matrix; for A column vector of all 1s; This represents the total number of differentially expressed lipid molecules involved in the calculation. The adaptive drift correction coefficient, at the computer's underlying level, can be the reciprocal of the global sample variance of the original metabolomics standardized data matrix.

[0081] A linear algebraic analyzer solves for all eigenvalues ​​of the Laplacian matrix of the non-homogeneous graph and sorts them in ascending order of numerical value, then truncates the second smallest eigenvalue as the Fiedler value parameter. Non-homogeneous centering alignment correction, by constraining the fluctuation baseline of all lipid nodes on the manifold space with non-homogeneous projection, can eliminate anomalous edge weight perturbations of outliers, enabling the calculated Fiedler value parameter to characterize the global connectivity of the pathological metabolic system.

[0082] Step S4 in this embodiment specifically includes:

[0083] In S4, a diagnostic model is established using a cluster of core hub markers to obtain false positive rate parameters and diagnostic evaluation indicators. If the preset qualified threshold is not reached, the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor is calculated to extract the back gradient of the diagnostic error, including:

[0084] A regression classification diagnostic model was established for the validation set using the molecular expression levels within the core hub biomarker cluster as input features.

[0085] Receiver operating characteristic curves are generated based on a regression classification diagnostic model.

[0086] The false positive rate parameter and the area under the curve parameter, which are used as diagnostic assessment indicators, were extracted from the receiver operating characteristic curve.

[0087] If the area under the curve parameter is lower than the preset qualified threshold, the partial derivative equation of the false positive rate parameter with respect to the node diagnostic penalty factor of each differential lipid molecule is established to calculate and output the back gradient of the diagnostic error.

[0088] Specifically, validation set sample data can be extracted, and the regression classification diagnostic model can be established based on the core pivot marker cluster. The classifier algorithm can be invoked to obtain the judgment probability threshold of the model output, and a subject operating characteristic curve in a two-dimensional coordinate system can be generated according to the probability gradient. The parsing engine reads the coordinate boundary data of this curve, extracts the false positive rate parameter, and calls the numerical integration microservice to perform area calculation on the region below the curve, extracting the obtained area under the curve value as the diagnostic evaluation index.

[0089] The preset pass threshold can be set according to the accuracy requirements of the clinical diagnostic kit. In this embodiment, the preset pass threshold can be set to 0.90 (i.e., the area under the curve parameter, which serves as the diagnostic evaluation indicator, must reach 0.90 or higher before iteration stops). When the comparison finds that the diagnostic evaluation indicator has not reached the preset pass threshold, a partial derivative equation of the false positive rate parameter with respect to the node diagnostic penalty factor for each differential lipid molecule is established to solve for the reverse error. The underlying differential mathematical model can be:

[0090] ;

[0091] in, Indicates that for the first The diagnostic error reverse gradient of differential lipid molecule extraction. This represents partial derivative operations. This represents the false positive rate parameter. Indicates the relationship with the first A node diagnostic penalty factor variable that is differentially bound to lipid molecules.

[0092] It should be noted that, due to the discrete and non-differentiable characteristics of the preceding node truncation operation and the training process of the diagnostic model in a strictly analytical geometric sense, the underlying layer does not directly perform analytical differentiation. Instead, it calls the numerical differentiation engine and uses the finite difference method to approximate the partial derivative. Specifically, it can be used for the... The penalty factor for each differential lipid molecule Apply a very small positive perturbation (as set) Rerun the forward pass process to obtain the false positive rate after the disturbance. Then through the underlying formula The discrete diagnostic error back gradient is calculated. The specific implementation of this finite difference method is a conventional technique in this field, and this embodiment does not impose any specific limitations on it.

[0093] Modern metabolomics bioinformatics analysis tools generally suffer from a disconnect between front-end feature screening and back-end clinical testing. Conventional static network analysis focuses solely on graph theory scores, easily misidentifying highly active but pathologically specific basic substrates in metabolic networks as biomarkers. Introducing such molecules into clinical cohorts for validation directly leads to false positives and misdiagnosis.

[0094] This step adds a negative feedback mechanism based on clinical validation efficacy to the pure graph theory screening. Utilizing the error calculation mechanism of the receiver operating characteristic (ROC) curve, the macroscopic clinical false positive rate is transformed into a penalty gradient targeting specific lipid molecules through partial derivative calculation. This avoids the inefficient method of repeatedly adjusting parameters manually, providing pathologically oriented error navigation parameters for the subsequent adaptive reshaping of the topology network.

[0095] Step S5 in this embodiment specifically includes:

[0096] In S5, the absolute safe perturbation step size threshold is defined by combining the eigenvalue offset radius, the Fiedler value parameter, and the preset minimum value, including:

[0097] Extract the off-diagonal elements (excluding diagonal elements) within a single target row of the graph Laplacian matrix;

[0098] Calculate the sum of the absolute values ​​of all off-diagonal elements to generate the eigenvalue offset radius;

[0099] The connectivity margin parameter is obtained by calculating the difference between the Fiedler value parameter and the preset minimum value;

[0100] The connectivity margin parameter and the eigenvalue offset radius are input into the boundary mapping function to calculate the absolute safe perturbation step size threshold that defines the constraint boundary.

[0101] Specifically, the underlying instructions traverse and read all off-diagonal elements within the graph Laplacian matrix, excluding the main diagonal positions. Absolute value transformations and accumulation operations are sequentially performed on these extracted off-diagonal elements to generate eigenvalue offset radii. Simultaneously, the Fiedler value parameter is invoked, and a subtraction operation is performed using the preset minimum value extracted from memory to ensure the network's fundamental irreducibility, obtaining the connectivity margin parameter representing the dimensionality reduction tolerance limit.

[0102] The connectivity margin parameter and the eigenvalue offset radius are simultaneously incorporated into the boundary mapping function to calculate the absolute safe perturbation step size threshold for preventing network breakage. The mathematical model of this boundary mapping function can be expressed as:

[0103]

[0104] in, This represents the defined absolute safety perturbation step size threshold. This parameter represents the allowable connectivity margin before the network topology degrades to a critical point of failure. The underlying layer can obtain this parameter in real time by calculating the difference between the Fiedler value parameter and the preset minimum value. Indicates the first The eigenvalue offset radius of each differential lipid molecule; This represents a smoothing zero-base constant, which can take the value of 10 to the power of negative six, used to avoid division-by-zero overflow caused by the eigenvalue offset radius being calculated as zero.

[0105] Conventional backpropagation algorithms face the problem of topological breaks in the original graph network due to excessive penalty magnitude when imposing weight penalties on high-frequency false positive nodes. Once the network is discretely broken, the biochemical synergistic relationships between lipid molecules are destroyed at the computational level. This step uses the mathematical property of matrix eigenvalue offset to define the connectivity extremum limit, transforming the requirement for preventing biological network breaks into a connectivity margin value. Based on this, the underlying instructions inversely calculate the limit tolerance boundary for applying dimensionality reduction penalties to specific lipid nodes. This establishes a rigid physical defense for optimization iteration, avoids the computational risk of structural collapse of the lipid cascade network caused by issuing excessive error penalties, and ensures that the final delivered biomarker cluster has a complete dimension for interpreting pathological metabolic networks.

[0106] Furthermore, the absolute safety perturbation step size threshold calculated in step S5 can be used to introduce a dynamic nonlinear adaptive scaling mechanism for the time step size that evolves with the number of closed-loop iterations. In the middle and later stages of the closed-loop iteration, if the parameter update amount remains static, the value of the penalty factor may oscillate alternately when approaching the constraint boundary, which can easily lead to eigenvalue out-of-bounds errors and cause local structural breaks in the underlying network topology.

[0107] To address the topology breakage problem caused by boundary numerical oscillations, this embodiment calculates the static topology step size threshold using a boundary mapping function. Then, a dynamic nonlinear decay factor in the time dimension can be further introduced to calculate the decay factor at the current time. The dynamic absolute safety perturbation step size threshold output in the closed-loop iteration. The dynamic absolute safety perturbation step size threshold. The time-varying evolution mathematical model can be expressed as:

[0108] ;

[0109] in, This is the static step size threshold; The current closed-loop iteration round number can be passed in real time by an internal counter; The value of the damping convergence control constant is preset and can be limited to the range of [0.05, 0.15]. In this embodiment, it is configured as 0.10.

[0110] Dynamic absolute safety perturbation step size threshold With the number of iterations The increase in exhibits an exponential, non-linear, and smooth decay. In the early stages of the closed-loop iteration, the number of iteration rounds... The value is relatively small, and the dynamic nonlinear decay factor is small. A value close to 1 can maintain a large perturbation window to ensure the convergence rate of the gradient descent of the optimization model; in the later stages of closed-loop iteration, the number of iterations... Gradually increase the dynamic absolute safety perturbation step size threshold. It exhibits exponential adaptive narrowing, and the optimization model approximates the stationary boundary of the Karnofsky Performance Condition (KKT condition) with decreasing fine update increments during the solution process. This can eliminate numerical oscillations at the boundary and limit the network topology reconstruction process to be performed within the safe connectivity threshold.

[0111] Step S6 in this embodiment specifically includes:

[0112] In S6, an optimization model is constructed with the objective of minimizing the false positive rate parameter and constrained by an absolute safety perturbation step size threshold and a preset minimum value. The incremental update node diagnostic penalty factor is calculated, the lipid collaborative adjacency matrix is ​​reconstructed, and the model is iterated until the diagnostic evaluation index reaches a preset qualified threshold. The output core hub biomarker cluster includes:

[0113] A Lagrange-constrained optimization function is constructed as the optimization model by integrating the objective cost function that minimizes the false positive rate parameter, the first constraint condition that does not exceed the absolute safety perturbation step size threshold, and the second constraint condition that is not lower than the preset minimum value.

[0114] The optimal penalty update increment is obtained by numerically solving the stationary equations of partial derivatives and complementary relaxation conditions of the Lagrange-constrained optimization function.

[0115] The optimal penalty update increment is added to the node diagnosis penalty factor to generate the updated node diagnosis penalty factor;

[0116] The lipid collaborative adjacency matrix is ​​reconstructed based on the updated node diagnostic penalty factor, and the graph Laplacian matrix is ​​updated to trigger network topology reconstruction iteration until the diagnostic evaluation index reaches the preset qualified threshold and the iteration stops, outputting the core hub marker cluster.

[0117] In a specific implementation, based on the principle of the Lagrange multiplier method, a dual-constraint optimization model can be constructed with the objective of minimizing the false positive rate parameter and satisfying the following conditions: "the update increment of the node diagnosis penalty factor does not exceed the absolute safety perturbation step size threshold" (first constraint) and "the Fiedler value parameter of the network is not lower than the preset minimum value" (second constraint).

[0118] A numerical analyzer can be invoked to solve the problem by simultaneously solving the stationary partial derivative equations and the complementary relaxation conditions, thereby extracting the optimal penalty update increment that satisfies the aforementioned anti-boundary constraint. It should be noted that the specific numerical methods for solving constrained optimization problems by simultaneously solving the stationary partial derivative equations and the complementary relaxation conditions (such as the interior-point method, sequential quadratic programming, etc.) are common knowledge in the field, and those skilled in the art can implement them based on existing technology. This embodiment does not specifically limit these methods.

[0119] The system executes parameter iteration instructions, summing the optimal penalty update increment with the current node diagnostic penalty factor, inputting it into the topology network initialization module to reconstruct the underlying edge weight structure, and triggering new feature decomposition and centrality evaluation. The iterative monitoring mechanism continuously probes the update process until the diagnostic evaluation indicators reach the preset qualified threshold, solidifies the network topology state, terminates the iteration, and outputs the final core hub marker cluster.

[0120] Traditional bioinformatics algorithms, when performing parameter tuning on high-dimensional lipid graph networks, are prone to getting trapped in local optima due to blind gradient descent or experiencing the collapse of the underlying topology due to uncontrolled step size adjustments. This step establishes a mathematical optimization mechanism with physical boundary constraints at the end. By utilizing the underlying solution of Lagrange multipliers and the Caro requirement-Kuntuak condition, empirical parameter adjustment is transformed into a matrix solution task. This ensures that each adjustment to the network topology evolves in the direction of reducing clinical false positives and is confined within a safe connectivity range. This closed-loop iterative architecture with built-in boundary constraints avoids human intervention, enabling the final delivered biomarkers to achieve high clinical specificity while maintaining the integrity of the deep biological cascade network.

[0121] The steps of the various methods described above are only for clarity. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this application. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but without changing the core design of the algorithm and process, are also within the scope of protection of this application.

[0122] Second Embodiment

[0123] The second embodiment of this application relates to a screening system for differential lipid metabolism biomarkers in human tissues, comprising the following modules:

[0124] The basic correlation calculation module is used to calculate the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the standardized data matrix;

[0125] The topology network initialization module is used to fuse the absolute value of the rank correlation coefficient with the node diagnostic penalty factor to construct the lipid cooperative adjacency matrix and generate the graph Laplacian matrix.

[0126] The centrality feature extraction module is used to extract the Fidler value parameters of the graph Laplacian matrix and the eigenvector centrality parameters of the lipid co-adjacency matrix, and to extract the core hub marker clusters.

[0127] The diagnostic error extraction module is used to establish a diagnostic model based on the core hub marker cluster to obtain the false positive rate parameter and diagnostic evaluation index. If the preset qualified threshold is not reached, the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor is calculated to extract the back gradient of the diagnostic error.

[0128] The disturbance boundary delineation module is used to combine the eigenvalue offset radius, the Fiedler value parameter, and the preset minimum value to define the absolute safe disturbance step size threshold.

[0129] The constrained optimization iteration module is used to construct an optimization model with the objective of minimizing the false positive rate parameter and constrained by an absolute safety perturbation step size threshold and a preset minimum value. It solves the incremental update node diagnostic penalty factor, reconstructs the lipid collaborative adjacency matrix and iterates until the diagnostic evaluation index reaches the preset qualified threshold, and outputs the core hub biomarker cluster.

[0130] It is not difficult to see that this embodiment is a system embodiment corresponding to the first embodiment, and this embodiment can be implemented in conjunction with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the first embodiment.

[0131] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent from this embodiment.

[0132] Furthermore, some embodiments of this application also provide an electronic device. The electronic device can be various forms of digital computer, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, mainframe computers, cellular phones, smartphones, wearable devices, and other similar computing devices.

[0133] The electronic device includes: one or more processors; and a memory storing computer program instructions, which, when executed, cause the processor to perform the steps of a method for screening differential lipid metabolism biomarkers in human tissues provided in any one or more embodiments.

[0134] Figure 5An exemplary structural diagram of the electronic device is disclosed. The electronic device includes one or more processors 1101, a memory 1102, an input device 1103, and an output device 1104. The various components are interconnected via a bus or other means (the diagram shows an example of bus connection). The processor 1101 can be used to execute instructions stored in the memory 1102 to control the overall operation of the electronic device. The memory 1102 may include a program storage area and a data storage area, wherein the program storage area stores the operating system and applications required for at least one function; the data storage area stores data created according to the use of the electronic device, etc. The memory 1102 may include high-speed random access memory and may also include non-transitory memory, such as disk storage devices, flash memory devices, or other non-transitory solid-state storage devices. In some embodiments, the memory 1102 may also include storage resources located remotely to the processor and accessible via a network.

[0135] Input device 1103 can be used to receive input numerical or character information or user operation signals, such as a touch screen, keypad, mouse, trackpad, touchpad, indicator, one or more mouse buttons, trackball, joystick, etc. Output device 1104 may include display devices (such as liquid crystal displays, light-emitting diode displays, plasma displays, and optional touch screens), auxiliary lighting devices (such as LEDs), and haptic feedback devices (such as vibration motors), etc.

[0136] To facilitate user interaction, the electronic device may be configured to include a display device (such as an LCD or CRT monitor) and input devices such as a keyboard and pointing devices (e.g., a mouse or touchpad). Feedback can be any form of sensory feedback (e.g., visual feedback, auditory feedback); input may also be received via voice, touch, or other means.

[0137] This application also relates to a computer-readable medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of a method for screening differential lipid metabolism biomarkers in human tissues provided in any one or more embodiments. The computer-readable medium may be a memory included in an electronic device, or it may be a standalone storage medium not assembled into the device.

[0138] It should be noted that the computer-readable medium described in this application may be a computer-readable signal medium, a computer-readable storage medium, or a combination of both. Examples include, but are not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. Specific examples of storage media may include, but are not limited to, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory, optical fibers, portable CD-ROMs, optical storage devices, magnetic storage devices, etc., or any suitable combination thereof.

[0139] Computer-readable media may store one or more programs that can be used by or in conjunction with an instruction execution system. The media may be permanent or non-permanent, removable or non-removable, and may store information by any method or technology, including computer-readable instructions, data structures, program modules, or other data.

[0140] The computer program code used to implement the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​(such as Java, Smalltalk, and C++) and conventional procedural programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. The remote computer can be connected to the user's computer via any network (including a local area network or a wide area network) or can be connected to an external computer.

[0141] In the above embodiments, the functions can be implemented in whole or in part by software, hardware, firmware, or any combination thereof, for example, by using application-specific integrated circuits, general-purpose computers, or other similar hardware devices. In some embodiments, the software program of this application can be executed by a processor to implement the steps or functions; it can also be implemented by hardware, for example, as a circuit that works in conjunction with the processor to execute the steps or functions.

[0142] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0143] The scope of this application is defined by the appended claims rather than the foregoing description, and is therefore intended to encompass all variations falling within the meaning and scope of equivalents of the claims. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device in software or hardware. Terms such as "first," "second," etc., are used only for distinguishing descriptions and do not indicate any particular order, nor should they be construed as indicating or implying relative importance.

[0144] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily made by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims, and the above embodiments should be regarded as exemplary and non-limiting.

Claims

1. A method for screening differential lipid metabolism biomarkers in human tissues, characterized in that, Includes the following steps: S1. Calculate the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the standardized data matrix; S2. The absolute value of the rank correlation coefficient is fused with the node diagnostic penalty factor to construct a lipid cooperative adjacency matrix, and a graph Laplacian matrix is ​​generated. S3. Extract the Fiedler value parameter of the graph Laplacian matrix and the eigenvector centrality parameter of the lipid co-adjacency matrix, and extract the core hub marker cluster; S4. Establish a diagnostic model based on the core hub marker cluster to obtain false positive rate parameters and diagnostic evaluation indicators. If the preset qualified threshold is not reached, calculate the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor to extract the back gradient of the diagnostic error. S5. Combine the eigenvalue offset radius, the Fiedler value parameter, and the preset minimum value to define the absolute safety perturbation step size threshold. S6. Construct an optimization model with the objective of minimizing the false positive rate parameter and constrained by the absolute safety perturbation step size threshold and the preset minimum value, solve the incremental update of the node diagnostic penalty factor, reconstruct the lipid collaborative adjacency matrix and iterate until the diagnostic evaluation index reaches the preset qualified threshold, and output the core hub marker cluster. In S3, the extraction of the Fiedler value parameter of the graph Laplacian matrix and the eigenvector centrality parameter of the lipid co-adjacency matrix, and the extraction of the core hub marker cluster, includes: performing eigenvalue decomposition on the graph Laplacian matrix to extract an eigenvalue sequence, and extracting the second smallest eigenvalue in the eigenvalue sequence as the Fiedler value parameter; calculating the strictly positive eigenvector corresponding to the largest real eigenvalue of the lipid co-adjacency matrix; extracting each component of the strictly positive eigenvector as the eigenvector centrality parameter of the corresponding differential lipid molecule; sorting the eigenvector centrality parameters of all differential lipid molecules in descending order, and extracting the molecular combinations that are in the preset highest interval to constitute the core hub marker cluster; In S5, the step of defining the absolute safety perturbation step size threshold by combining the eigenvalue offset radius, the Fiedler value parameter, and the preset minimum value includes: extracting the non-diagonal elements (excluding diagonal elements) within a single target row of the graph Laplacian matrix; calculating the sum of the absolute values ​​of all the non-diagonal elements to generate the eigenvalue offset radius; calculating the difference between the Fiedler value parameter and the preset minimum value to obtain the connectivity margin parameter; and inputting the connectivity margin parameter and the eigenvalue offset radius into a boundary mapping function to calculate the absolute safety perturbation step size threshold that defines the constraint boundary. In S6, the construction of an optimization model aimed at minimizing the false positive rate parameter and constrained by the absolute safety perturbation step size threshold and the preset minimum value, solves the incremental update of the node diagnostic penalty factor, reconstructs the lipid co-adjacency matrix and iterates until the diagnostic evaluation index reaches the preset qualified threshold, and outputs the core hub biomarker cluster including: a first constraint that the update increment of the node diagnostic penalty factor does not exceed the absolute safety perturbation step size threshold, and a second constraint that the Fiedler value parameter is not lower than the preset minimum value. The optimal model is constructed by establishing Lagrange-constrained optimization functions under certain constraints. The optimal penalty update increment is obtained by numerically solving the stationary equations of partial derivatives and complementary relaxation conditions of the Lagrange-constrained optimization function. This optimal penalty update increment is then added to the node diagnostic penalty factor to generate the updated node diagnostic penalty factor. Based on the updated node diagnostic penalty factor, the lipid cooperative adjacency matrix is ​​reconstructed, and the graph Laplacian matrix is ​​updated to trigger network topology reconstruction iterations until the diagnostic evaluation index reaches the preset qualified threshold, at which point the iteration stops, and the core hub marker cluster is output.

2. The method for screening differential lipid metabolism biomarkers in human tissues according to claim 1, characterized in that, In S1, the calculation of the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the normalized data matrix includes: Extract raw metabolomics data from human tissue samples; The original metabolomics data were subjected to missing value reconstruction and normalization to generate the standardized data matrix; Extract the pairwise differential lipid molecule expression sequences within the standardized data matrix; Calculate the rank correlation coefficient between each pair of differentially expressed lipid molecules, and take the absolute value of the rank correlation coefficient to generate the absolute value of the rank correlation coefficient.

3. The method for screening differential lipid metabolism biomarkers in human tissues according to claim 1, characterized in that, In S2, the step of fusing the absolute value of the rank correlation coefficient with the node diagnostic penalty factor to construct a lipid cooperative adjacency matrix and generating a graph Laplacian matrix includes: The absolute value of the grade correlation coefficient and the pairwise corresponding node diagnostic penalty factors are input into a preset non-negative mapping function to calculate the dynamic connection edge weights; The lipid cooperative adjacency matrix is ​​constructed based on the dynamic connection edge weights of all differential lipid molecules; Calculate the sum of the generation degree matrix of the elements in the corresponding rows of a single differential lipid molecule within the lipid cooperative adjacency matrix; The graph Laplacian matrix is ​​generated by subtracting the lipid co-adjacency matrix from the degree matrix.

4. The method for screening differential lipid metabolism biomarkers in human tissues according to claim 1, characterized in that, In S4, the step of establishing a diagnostic model based on the core hub marker cluster to obtain false positive rate parameters and diagnostic evaluation indicators, and calculating the partial derivative of the false positive rate parameters with respect to the node diagnostic penalty factor to extract the back gradient of the diagnostic error if the preset qualified threshold is not reached, includes: A regression classification diagnostic model was established for the validation set using the molecular expression levels within the core hub marker cluster as input features. Receiver operating characteristic curves are generated based on the regression classification diagnostic model; The false positive rate parameter and the area under the curve parameter, which serves as the diagnostic assessment indicator, are extracted by analyzing the receiver operating characteristic curve. If the area under the curve parameter is lower than the preset qualified threshold, the partial derivative equation of the false positive rate parameter with respect to the node diagnostic penalty factor of each differential lipid molecule is established, and the diagnostic error inverse gradient is output.

5. A screening system for differential lipid metabolism biomarkers in human tissues, characterized in that, Includes the following modules: The basic correlation calculation module is used to calculate the absolute value of the rank correlation coefficient of differentially expressed lipid molecule sequences in the standardized data matrix; The topology network initialization module is used to fuse the absolute value of the rank correlation coefficient with the node diagnostic penalty factor to construct a lipid cooperative adjacency matrix and generate a graph Laplacian matrix. The centrality feature extraction module is used to extract the Fiedler value parameter of the graph Laplacian matrix and the eigenvector centrality parameter of the lipid co-adjacency matrix, and to extract the core hub marker cluster; The diagnostic error extraction module is used to establish a diagnostic model based on the core hub marker cluster to obtain false positive rate parameters and diagnostic evaluation indicators. If the preset qualified threshold is not reached, the module calculates the partial derivative of the false positive rate parameter with respect to the node diagnostic penalty factor to extract the back gradient of the diagnostic error. The disturbance boundary definition module is used to define an absolute safe disturbance step size threshold by combining the eigenvalue offset radius, the Fiedler value parameter and the preset minimum value. The constrained optimization iteration module is used to construct an optimization model with the objective of minimizing the false positive rate parameter and constrained by the absolute safety perturbation step size threshold and the preset minimum value, solve the incremental update of the node diagnosis penalty factor, reconstruct the lipid collaborative adjacency matrix and iterate until the diagnostic evaluation index reaches the preset qualified threshold, and output the core hub marker cluster. The centrality feature extraction module is specifically used for: performing eigenvalue decomposition on the graph Laplacian matrix to extract an eigenvalue sequence; extracting the second smallest eigenvalue within the eigenvalue sequence as the Fiedler value parameter; calculating the strictly positive eigenvector corresponding to the largest real eigenvalue of the lipid co-adjacency matrix; extracting each component of the strictly positive eigenvector as the eigenvector centrality parameter of the corresponding differential lipid molecules; sorting the eigenvector centrality parameters of all differential lipid molecules in descending order; and extracting the molecular combinations that rank in the preset highest interval to form the core hub marker cluster. The perturbation boundary delineation module is specifically used for: extracting the non-diagonal elements (excluding diagonal elements) within a single target row of the graph Laplacian matrix; calculating the sum of the absolute values ​​of all the non-diagonal elements to generate the eigenvalue offset radius; calculating the difference between the Fiedler value parameter and the preset minimum value to obtain the connectivity margin parameter; and inputting the connectivity margin parameter and the eigenvalue offset radius into the boundary mapping function to calculate the absolute safety perturbation step size threshold for delineating the constraint boundary. The constrained optimization iteration module is specifically used to: integrate the objective cost function that minimizes the false positive rate parameter, the first constraint that the update increment of the node diagnostic penalty factor does not exceed the absolute safety perturbation step size threshold, and the second constraint that the Fiedler value parameter is not lower than the preset minimum value, to construct a Lagrange constrained optimization function as the optimization model; numerically solve the partial derivative stationary equation and complementary relaxation conditions of the Lagrange constrained optimization function to obtain the optimal penalty update increment; add the optimal penalty update increment to the node diagnostic penalty factor to generate the updated node diagnostic penalty factor; reconstruct the lipid cooperative adjacency matrix based on the updated node diagnostic penalty factor and update the graph Laplacian matrix to trigger network topology reconstruction iteration until the diagnostic evaluation index reaches the preset qualified threshold and the iteration stops, and the core hub marker cluster is output.

6. An electronic device, characterized in that, The electronic device includes: one or more processors; and a memory storing computer program instructions, which, when executed, cause the processor to perform the steps of a method for screening differential lipid metabolism biomarkers in human tissues as described in any one of claims 1 to 4.

7. A computer-readable medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method for screening differential lipid metabolism biomarkers in human tissues as described in any one of claims 1 to 4.

Images