Multicomponent malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device

Abstract

This invention discloses a multi-omics device for analyzing the spatiotemporal heterogeneity of immunometabolic reprogramming in malignant pleural effusion. The device includes an integration analysis module for performing multimodal integration and consensus clustering analysis on multi-omics datasets to obtain integration analysis results. Based on the integration analysis results, the training samples of malignant pleural effusion from patients are classified into three subtypes: immunometabolic activation, immunometabolic transition, and immunometabolic inhibition. A screening and construction module is used to screen immunometabolic biomarkers from the features of the immunometabolic inhibition subtype, construct a candidate target set based on these biomarkers, and build a prediction model. An assessment and suggestion module is used to calculate treatment response scores and integrate clinical staging information to construct a nomogram, obtaining prognostic risk assessment results and personalized treatment guidance suggestions. This improves the accuracy of prognostic assessment and the reliability of treatment response prediction, enabling precise risk stratification and personalized treatment guidance for patients.

Description

Technical Field

[0001] This invention relates to the field of prognostic assessment and treatment technology for malignant pleural effusion, and in particular to a multi-omics device for analyzing the spatiotemporal heterogeneity of immunometabolic reprogramming in malignant pleural effusion. Background Technology

[0002] Malignant pleural effusion is a common complication in patients with advanced cancer, with a very poor prognosis and a median survival of only 3–12 months.

[0003] Its unique immunosuppressive and metabolically abnormal microenvironment significantly weakens the efficacy of chemotherapy, targeted therapy, and immunotherapy, and there is currently a lack of precise prognostic stratification tools.

[0004] In the microenvironment of malignant pleural effusion (MPE), immune cell functional remodeling and metabolic reprogramming jointly promote immune escape; for example, lactate can inhibit natural killer cells (NK cells) and CD8-positive T lymphocytes (CD8+). + It enhances T cell function and strengthens the immunosuppressive activity of regulatory T cells (Tregs); abnormal lipid metabolism also contributes to the formation of an immunosuppressive phenotype.

[0005] Currently, clinical prognostic assessment of MPE patients mainly relies on traditional models such as lung function, Eastern Cooperative Oncology Group performance status, sodium ion levels, and tumor type, as well as the Lung (Eastern Cooperative Oncology Group performance status, Na+, and Tumor type, LENT) prognostic scoring system and the Prognostic Model for Survival in Effusions (PROMISE) prognostic scoring system. However, these models are based on macroscopic indicators and have limited predictive accuracy (Area Under the Curve, AUC, approximately 0.78–0.79), making it difficult to reveal the heterogeneity of molecular mechanisms.

[0006] With the development of multi-omics technologies, some studies have attempted to use single-omics data to build prognostic models. However, these models often have high feature dimensions, poor generalization ability, and lack systematic subtyping from the perspective of immune-metabolic interactions, thus failing to guide personalized treatment.

[0007] More importantly, there are currently no reports of immunometabolic molecular subtyping studies for MPE, resulting in limited treatment strategies and poor efficacy. Summary of the Invention

[0008] The main objective of this invention is to provide a multi-omics device for analyzing the spatiotemporal heterogeneity of immunometabolic reprogramming in malignant pleural effusion. This device aims to address the technical problems in existing malignant pleural effusion prognostic assessment models, which are based solely on macroscopic clinical indicators, have limited predictive accuracy, lack systematic molecular subtyping studies on immunosuppression and metabolic abnormalities in the MPE microenvironment, and cannot accurately identify patient subgroups or provide targeted treatment guidance. Consequently, these models result in inaccurate clinical prognostic assessments, highly arbitrary selection of treatment plans, and difficulty in achieving precise stratification and individualized treatment for patients with malignant pleural effusion.

[0009] This invention provides a multi-omics device for analyzing the spatiotemporal heterogeneity of malignant pleural effusion immunometabolic reprogramming. The device includes: an integration analysis module, a screening and construction module, and an evaluation and recommendation module.

[0010] The integrated analysis module is used to acquire a multi-omics dataset consisting of transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine chip data from patients with malignant pleural effusion. It performs multimodal integration and consensus clustering analysis on the multi-omics dataset to obtain integrated analysis results. Based on these results, the training samples of malignant pleural effusion from patients with malignant pleural effusion are classified into three subtypes: immunometabolic activation, immunometabolic transition, and immunometabolic inhibition.

[0011] The screening and construction module is used to screen out immunometabolic biomarkers from the features of the immunometabolic inhibitory subtype, form a candidate target set based on the immunometabolic biomarkers, and construct a prediction model based on the candidate target set.

[0012] The assessment and recommendation module is used to input the current integrated analysis results of the test samples of the patients with malignant pleural effusion into the prediction model, calculate the treatment response score, and integrate clinical staging information to construct a nomogram to obtain prognostic risk assessment results and personalized treatment guidance recommendations.

[0013] Optionally, the integration analysis module is further configured to use a graph neural network algorithm to perform multimodal integration and consensus clustering analysis on different omics data in the multi-omics dataset to obtain integration analysis results;

[0014] The integrated analysis module is also used to classify the malignant pleural effusion training samples of the patients with malignant pleural effusion into three subtypes based on the integrated analysis results: an immunometabolic activation subtype with enhanced immune cell activity and active metabolic pathways, an intermediate immunometabolic transition subtype, and an immunometabolic suppression subtype with an immunosuppressive microenvironment and abnormal metabolism.

[0015] Optionally, the integrated analysis module is further configured to construct a heterogeneous network model using a graph neural network algorithm, using the malignant pleural effusion training samples of the patients with malignant pleural effusion as network nodes, and using the transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine chip data of the patients with malignant pleural effusion as multidimensional features of the nodes of the heterogeneous network model.

[0016] The integrated analysis module is also used to learn the similarity relationship between nodes through the message passing mechanism of graph neural networks, integrate multi-omics information, obtain a low-dimensional embedding representation of each malignant pleural effusion training sample, and form an integrated feature matrix at the sample level.

[0017] The integration analysis module is also used to perform dimensionality reduction on the integration feature matrix, perform multiple rounds of K-means clustering in the dimensionality reduction space, evaluate the stability of clustering under different K values ​​by calculating the consensus matrix, determine the optimal number of clusters by the cumulative distribution function curve as a plateau characteristic and the relative change area, obtain the final clustering result based on the optimal number of clusters, and use the final clustering result as the integration analysis result.

[0018] Optionally, the screening module is further configured to analyze and screen for immunometabolic biomarkers related to overall survival of patients with malignant pleural effusion from the characteristics of the immunometabolic suppressive subtype using univariate Cox proportional hazards regression.

[0019] The screening and construction module is also used to screen out molecules with clear targeted therapeutic potential based on the biological function and therapeutic feasibility of the immune metabolic markers, and to form a candidate target set.

[0020] The screening and construction module is also used to construct a prediction model based on the molecular characteristics of the candidate target set.

[0021] Optionally, the screening module is further configured to analyze the association network of key metabolites related to the overall survival of patients with malignant pleural effusion from the features of the immunometabolic suppression subtype using univariate Cox proportional hazards regression, and to identify immunometabolic biomarkers with prognostic value from the association network.

[0022] Optionally, the screening and construction module is further configured to use the molecular features in the candidate target set as input variables to construct a machine learning algorithm pool composed of various machine learning algorithms;

[0023] The screening module is also used to evaluate the performance of each model in the machine learning algorithm pool through five-fold cross-validation, calculate the average AUC value of each model in five validations, and select the model with the highest AUC value as the prediction model.

[0024] Optionally, the assessment recommendation module is further used to input the current integrated analysis results of the test samples of the patients with malignant pleural effusion into the prediction model, calculate the treatment response score and integrate clinical staging information to construct a nomogram, and obtain prognostic risk assessment results and personalized treatment guidance recommendations.

[0025] Optionally, the candidate target set includes C1q, LDHA, and MCT1, with C1q serving as the core target.

[0026] Optionally, the assessment recommendation module is also used to provide personalized treatment guidance recommendations for target patients who belong to the immunometastatic subtype and whose treatment response score is less than a preset score, using F4 / 80 modified lipid nanoparticles to deliver si-C1qa in combination with anti-PD-1 antibody.

[0027] Optionally, the assessment recommendation module is also used to construct organoid models using malignant pleural effusion cells derived from patients, and to verify the impact of key targets in the candidate target set on the treatment response.

[0028] This invention proposes a multi-omics device for analyzing the spatiotemporal heterogeneity of malignant pleural effusion immunometabolism reprogramming. This device includes: an integration analysis module for acquiring a multi-omics dataset composed of transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine microarray data from patients with malignant pleural effusion; performing multimodal integration and consensus clustering analysis on the multi-omics dataset to obtain integration analysis results; and classifying the malignant pleural effusion training samples from patients with malignant pleural effusion into immunometabolism activation subtypes, immunometabolism transitional subtypes, and immunometabolism suppression subtypes based on the integration analysis results; and a screening and construction module for screening immunometabolism biomarkers from the features of the immunometabolism suppression subtype, and based on the immunometabolism biomarkers... A candidate target set is constructed, and a prediction model is built based on the candidate target set. An evaluation and suggestion module is used to input the current integrated analysis results of the test samples from patients with malignant pleural effusion into the prediction model, calculate the treatment response score, and integrate clinical staging information to construct a nomogram, obtaining prognostic risk assessment results and personalized treatment guidance suggestions. This enables precise analysis of the spatiotemporal heterogeneity of immunometabolic reprogramming in malignant pleural effusion, significantly improving the accuracy of prognostic assessment and the reliability of treatment response prediction. It achieves precise risk stratification and personalized treatment guidance for patients with malignant pleural effusion, thereby improving the clinical prognosis of these patients. This provides a scientific basis and technical support for the precise diagnosis and treatment of malignant pleural effusion, fundamentally improving the biological rationality of the model, enhancing predictive performance, and offering multi-dimensional functionality and strong system usability. Attached Figure Description

[0029] Figure 1 This is a functional block diagram of the first embodiment of the multi-omics malignant pleural effusion immune metabolism reprogramming spatiotemporal heterogeneity analysis device of the present invention;

[0030] Figure 2 A schematic diagram of an IMT model for predicting the prognosis of MPE constructed by integrating multi-omics data from a spatiotemporal heterogeneity analysis device for multi-omics immunometabolic reprogramming of malignant pleural effusion.

[0031] Figure 3 A schematic diagram of the IMT model for molecular typing and prognostic prediction in a spatiotemporal heterogeneity analysis device for multi-omics malignant pleural effusion immunometabolism reprogramming.

[0032] Figure 4 A schematic diagram illustrating the efficacy of intrapleural delivery of si-C1qa@F4 / 80 LNP to enhance anti-PD-1 therapy using a multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device.

[0033] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0034] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0035] The solution of this invention mainly includes: the multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device comprising: an integration analysis module, used to acquire a multi-omics dataset composed of transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine chip data from patients with malignant pleural effusion; performing multimodal integration and consensus clustering analysis on the multi-omics dataset to obtain integration analysis results; and classifying the malignant pleural effusion training samples from patients with malignant pleural effusion into immunometabolic activation subtypes, immunometabolic transitional subtypes, and immunometabolic suppression subtypes based on the integration analysis results; a screening and construction module, used to screen immunometabolic biomarkers from the features of the immunometabolic suppression subtype; forming a candidate target set based on the immunometabolic biomarkers; and constructing a prediction model based on the candidate target set; and an evaluation and suggestion module, used to input the current integration analysis results of the test samples from patients with malignant pleural effusion into the prediction model, calculate the treatment response score, and integrate the results. By constructing nomograms based on clinical staging information, prognostic risk assessment results and personalized treatment guidance can be obtained. This enables precise analysis of the spatiotemporal heterogeneity of immunometabolic reprogramming in malignant pleural effusion, significantly improving the accuracy of prognostic assessment and the reliability of treatment response prediction. It achieves precise risk stratification and personalized treatment guidance for patients with malignant pleural effusion, thereby improving the clinical prognosis of these patients. This provides a scientific basis and technical support for the precise diagnosis and treatment of malignant pleural effusion, fundamentally improving the biological rationality of the model, enhancing predictive performance, and offering multi-dimensional functionality and strong system usability. It solves the technical problems of existing malignant pleural effusion prognostic assessment models that are based solely on macroscopic clinical indicators, have limited predictive accuracy, lack systematic molecular subtyping studies of immunosuppression and metabolic abnormalities in the MPE microenvironment, cannot accurately identify patient subgroups and provide targeted treatment guidance, resulting in inaccurate clinical prognostic assessment, blind selection of treatment plans, and difficulty in achieving precise stratification and individualized treatment for patients with malignant pleural effusion.

[0036] Reference Figure 1 , Figure 1 This is a functional block diagram of the first embodiment of the multi-omics malignant pleural effusion immune metabolism reprogramming spatiotemporal heterogeneity analysis device of the present invention.

[0037] In the first embodiment of the multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device of the present invention, the multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device includes:

[0038] The system integrates the analysis module 10, the screening and construction module 20, and the evaluation and recommendation module 30; among which,

[0039] The integrated analysis module 10 is used to acquire a multi-omics dataset consisting of transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine chip data from patients with malignant pleural effusion. The module performs multimodal integration and consensus clustering analysis on the multi-omics dataset to obtain integrated analysis results. Based on the integrated analysis results, the training samples of malignant pleural effusion from patients with malignant pleural effusion are divided into three subtypes: immunometabolic activation type, immunometabolic transition type, and immunometabolic inhibition type.

[0040] The screening and construction module 20 is used to screen out immunometabolic biomarkers from the features of the immunometabolic inhibitory subtype, form a candidate target set based on the immunometabolic biomarkers, and construct a prediction model based on the candidate target set.

[0041] The assessment and recommendation module 30 is used to input the current integrated analysis results of the test samples of the patients with malignant pleural effusion into the prediction model, calculate the treatment response score and integrate clinical staging information to construct a nomogram, and obtain prognostic risk assessment results and personalized treatment guidance recommendations.

[0042] It should be noted that four key biomedical data types were collected from patients with malignant pleural effusion: transcriptome sequencing data reflecting gene expression levels, single-cell transcriptome sequencing data revealing cellular heterogeneity, non-targeted metabolome data showing metabolite changes, and cytokine microarray data reflecting the state of the immune microenvironment, forming a comprehensive multi-omics dataset. Then, advanced multimodal integration technology was used to effectively fuse these heterogeneous data, and consensus clustering analysis was employed to uncover hidden biological patterns within the data, avoiding the biases that may arise from single clustering methods. Finally, based on the integrated analysis results, patients with malignant pleural effusion were scientifically classified into three subtypes with different immunometabolic characteristics: the immunometabolic activation subtype (characterized by enhanced immune cell activity and active metabolic pathways), the immunometabolic transition subtype (exhibiting intermediate state characteristics), and the immunometabolic suppression subtype (characterized by an immunosuppressive microenvironment and abnormal metabolic state). This classification method overcomes the limitations of traditional methods that rely solely on macroscopic clinical indicators and can more accurately reflect the intrinsic biological heterogeneity of malignant pleural effusion.

[0043] Understandably, by using statistical methods to screen out immunometabolic biomarkers that are significantly associated with patient prognosis from the characteristics of immunometabolic suppressive subtypes, and then using these biomarkers to form a candidate target set, a multi-algorithm integrated prediction model can be constructed. This enables the transformation from basic molecular discovery to clinical application, simplifying complex immunometabolic characteristics into a quantifiable tool for predicting treatment responses.

[0044] It should be understood that by using different training samples from patients with malignant pleural effusion as the test samples, and employing the same multi-omics analysis process, the integrated analysis results are input into a validated predictive model. The model calculates a quantitative treatment response score based on the patient's unique immune and metabolic characteristics, reflecting the patient's potential response to a specific treatment plan. Simultaneously, this molecular-level prediction result is organically integrated with traditional clinical staging information (e.g., tumor stage, patient performance status) to construct an intuitive nomogram prediction tool, enabling physicians to obtain comprehensive assessment results through simple operation. Finally, the system outputs assessment results including prognostic risk stratification (e.g., low risk, intermediate risk, high risk) and provides differentiated, personalized treatment recommendations for patients at different risk levels. This achieves a complete closed loop of "molecular subtyping - prognostic assessment - treatment guidance," overcoming the limitations of traditional malignant pleural effusion diagnosis and treatment that relies solely on macroscopic clinical indicators. It transforms complex multi-omics data into decision support information that clinicians can understand and act upon, significantly improving the accuracy and effectiveness of treatment plan selection.

[0045] Furthermore, the integration analysis module 10 is also used to perform multimodal integration and consensus clustering analysis on different omics data in the multi-omics dataset using a graph neural network algorithm to obtain integration analysis results.

[0046] The integrated analysis module 10 is also used to classify the malignant pleural effusion training samples of the patients with malignant pleural effusion into an immunometabolic activation subtype with enhanced immune cell activity and active metabolic pathways, an intermediate immunometabolic transition subtype, and an immunometabolic suppression subtype with an immunosuppressive microenvironment and abnormal metabolism, based on the integrated analysis results.

[0047] It should be noted that the integration and analysis module employs graph neural network algorithms to deeply integrate multi-omics data (including transcriptomics, single-cell transcriptomics, non-targeted metabolomics, and cytokine microarray data) from patients with malignant pleural effusion. This algorithm effectively captures the nonlinear relationships and complex interactions between different omics data, mapping high-dimensional heterogeneous data to a unified low-dimensional representation space. Based on this, consensus clustering analysis, a robust statistical method, avoids the random bias of single clustering results, ensuring the reliability and reproducibility of the classification results. Finally, based on the integration and analysis results, the system scientifically classifies patients with malignant pleural effusion into three categories with clearly defined biological characteristics. Subtypes based on biological characteristics: the immunometabolic activation subtype (characterized by significantly enhanced immune cell activity and highly active metabolic pathways, indicating a better treatment response), the immunometabolic transitional subtype (exhibiting intermediate state characteristics and possessing a certain degree of therapeutic plasticity), and the immunometabolic suppression subtype (characterized by the formation of an immunosuppressive microenvironment and metabolic abnormalities, usually associated with poor prognosis). This molecular subtyping method based on multi-omics integration breaks through the limitations of traditional methods that rely solely on clinical phenotypes. It can more accurately reflect the intrinsic biological heterogeneity of malignant pleural effusion, providing a scientific basis for the subsequent development of individualized treatment strategies. It is a key technological link in achieving precision medicine for malignant pleural effusion.

[0048] Furthermore, the integrated analysis module 10 is also used to construct a heterogeneous network model using a graph neural network algorithm, using the malignant pleural effusion training samples of the patients with malignant pleural effusion as network nodes, and using the transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data and cytokine chip data of the patients with malignant pleural effusion as multidimensional features of the nodes of the heterogeneous network model.

[0049] The integrated analysis module 10 is also used to learn the similarity relationship between nodes through the message passing mechanism of graph neural network, integrate multi-omics information, obtain the low-dimensional embedding representation of each malignant pleural effusion training sample, and form an integrated feature matrix at the sample level.

[0050] The integration analysis module 10 is also used to perform dimensionality reduction processing on the integration feature matrix, perform multiple rounds of K-means clustering in the dimensionality reduction space, evaluate the stability of clustering under different K values ​​by calculating the consensus matrix, determine the optimal number of clusters by the cumulative distribution function curve as a plateau period feature and the relative change area, obtain the final clustering result based on the optimal number of clusters, and use the final clustering result as the integration analysis result.

[0051] It should be understood that the integration analysis module first constructs a heterogeneous network model, using training samples from patients with malignant pleural effusion as network nodes, and transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine microarray data as multidimensional features of these nodes. This design accurately reflects the biological reality that samples are the basic unit of analysis. Subsequently, through the message passing mechanism of graph neural networks, the model can learn the biological similarity relationships between sample nodes, effectively integrating heterogeneous information from different omics levels to generate a low-dimensional embedding representation (usually 32-128 dimensions) for each training sample, forming an integrated feature matrix that reflects the comprehensive characteristics of the samples. This matrix itself has already completed the necessary dimensionality reduction, avoiding the "curse of dimensionality" problem caused by high-dimensional data. On this basis, the module directly performs consensus clustering analysis in the low-dimensional embedding space formed by the integrated feature matrix. By repeating the iteration multiple times (usually 500 times, each time randomly sampling 80% of the samples and 90% of the features, but other values ​​can also be set, which are not limited in this embodiment), the consensus matrix is ​​calculated to evaluate the stability under different numbers of clusters (K value), and then the cumulative distribution function is analyzed. By scientifically determining the optimal number of clusters based on the plateau characteristics of the Distribution Function (CDF) curve and calculating the relative area of ​​change, clustering results with biological significance and clinical value can be obtained.

[0052] Furthermore, the screening construction module 20 is also used to analyze and screen out immunometabolic biomarkers related to the overall survival of patients with malignant pleural effusion from the characteristics of the immunometabolic suppression subtype through univariate Cox proportional hazards regression.

[0053] The screening and construction module 20 is also used to screen molecules with clear targeted therapeutic potential based on the biological function and therapeutic feasibility of the immune metabolic markers, and to form a candidate target set.

[0054] The screening and construction module 20 is also used to construct a prediction model based on the molecular characteristics of the candidate target set.

[0055] Understandably, the screening module first employs a univariate Cox proportional hazards regression statistical method for survival analysis to systematically screen for immunometabolic biomarkers significantly associated with overall survival in patients with malignant pleural effusion from the characteristics of the immunometabolic suppressive subtype. This effectively identifies key molecules that truly affect patient prognosis, rather than molecules that are merely phenotype-related but have no prognostic value. Subsequently, the module further evaluates the biological functions of these immunometabolic biomarkers (e.g., whether they participate in key immunometabolic pathways, whether they play a regulatory role in the tumor microenvironment) and their therapeutic feasibility (e.g., whether they are druggable, whether there are corresponding targeted drugs under development). The module identifies molecules with potential for targeted therapy, constructs a candidate target set, and realizes the transformation from diagnostic biomarkers to therapeutic targets. Finally, the module uses the molecular characteristics in the candidate target set as input variables to construct a predictive model. This model can transform complex molecular characteristics into quantifiable therapeutic response predictions, providing an intuitive reference for clinical decision-making. It not only solves the problem of lack of targeted therapeutic targets in the traditional diagnosis and treatment of malignant pleural effusion, but also establishes a complete transformation path from molecular discovery to clinical application, enabling the diagnosis and treatment of malignant pleural effusion to move from empirical treatment to a new stage of precision targeted therapy, significantly improving the pertinence and effectiveness of treatment.

[0056] Furthermore, the screening construction module 20 is also used to analyze the association network of key metabolites related to the overall survival of patients with malignant pleural effusion from the characteristics of the immunometabolic suppression subtype through univariate Cox proportional hazards regression, and to identify immunometabolic biomarkers with prognostic value from the association network.

[0057] It should be understood that the screening module first targets the immunometabolism-suppressive subtype, a patient group with a poor prognosis, and uses the classic survival analysis statistical method of univariate Cox proportional hazards regression to systematically assess the association strength between various metabolites and overall survival in this subtype. Based on this, the module further constructs an association network of key metabolites. This network not only includes direct interactions between metabolites but also integrates the interactions between metabolites and immune-related molecules (e.g., cytokines, immune checkpoint molecules), thus comprehensively reflecting the synergistic effects of immune and metabolic pathways in the tumor microenvironment. By analyzing this complex association network, the module can identify molecules that occupy key nodes in the network and are significantly associated with patient survival. These molecules not only have statistically significant prognostic value but also have clear functional significance in biology, serving as potential immunometabolism biomarkers. This biomarker screening method based on network analysis overcomes the limitations of traditional univariate analysis, enabling the discovery of molecules that are not significant in individual analyses but play a key role in the network. This provides a scientific basis for accurate prognostic assessment and targeted therapy of malignant pleural effusion, allowing clinicians to more accurately identify high-risk patients and develop individualized treatment strategies.

[0058] Furthermore, the screening and construction module 20 is also used to construct a machine learning algorithm pool composed of various machine learning algorithms by using the molecular features of the candidate target set as input variables.

[0059] The screening module 20 is also used to evaluate the performance of each model in the machine learning algorithm pool through five-fold cross-validation, calculate the average AUC value of each model in five validations, and select the model with the highest AUC value as the prediction model.

[0060] It should be noted that the screening module first uses the features of the candidate target set (e.g., key molecules such as C1q, LDHA, and MCT1) as input variables to construct an algorithm pool containing various machine learning algorithms (e.g., lasso regression, ridge regression, random survival forest, and support vector machine). This design fully leverages the advantages of different algorithms, avoids the limitations that a single algorithm may have, and ensures that the complex relationship between molecular features and clinical prognosis can be explored from multiple perspectives. Subsequently, the module uses a rigorous five-fold cross-validation method to evaluate the performance of each model in the algorithm pool. This method randomly divides the training data into five parts and uses them alternately. The process is repeated five times using four training models and one testing model to eliminate random bias caused by data partitioning. During this process, the module calculates the average area under the ROC curve (AUC) of each model in the five validations. The AUC value, as the gold standard for evaluating the performance of diagnostic or predictive models, can comprehensively reflect the model's classification ability at different thresholds. The closer the value is to 1, the better the model's performance. Finally, the module selects the model with the highest AUC value as the final predictive model. This selection criterion ensures the optimal performance of the predictive model in prognostic assessment and treatment response prediction, significantly improving the reliability and generalization ability of the prediction results.

[0061] Furthermore, the assessment suggestion module 30 is also used to input the current integrated analysis results of the test sample of the patient with malignant pleural effusion into the prediction model, calculate the treatment response score and integrate clinical staging information to construct a nomogram, and obtain prognostic risk assessment results and personalized treatment guidance suggestions.

[0062] Understandably, the assessment and recommendation module integrates the results of standardized multi-omics analysis of the test samples from patients with malignant pleural effusion, including patient-specific immune and metabolic characteristics. Subsequently, the module inputs these characteristics into a validated prediction model to calculate a quantified treatment response score (typically 0-100 points), which intuitively reflects the patient's likely response to a specific treatment regimen (e.g., immunotherapy). Building upon this, the module innovatively integrates molecular-level prediction results with traditional clinical staging information (e.g., TNM staging, PS score, pleural effusion volume) to construct an intuitive nomogram prediction tool. This allows physicians to quickly assess a patient's one-year survival probability by simply summing the scores of each factor. Finally, the system outputs assessment results including prognostic risk stratification (e.g., low risk, intermediate risk, high risk) and provides differentiated, personalized treatment recommendations for patients at different risk levels.

[0063] Furthermore, the candidate target set includes C1q, LDHA, and MCT1, with C1q serving as the core target.

[0064] It should be noted that the candidate target set includes three key molecules: C1q, LDHA, and MCT1. Together, they constitute the core feature network of the immunosuppressive subtype (IMT3). C1q, as the initiator molecule of the classical complement system pathway, has a significantly higher expression level in the immunosuppressive subtype than in other subtypes. It directly participates in the formation of the immunosuppressive microenvironment and is significantly associated with poor patient prognosis, thus being identified as a core target. LDHA, as a key enzyme in glycolysis, drives the Warburg effect in tumor cells, promoting lactate accumulation and microenvironment acidification. MCT1, as a monocarboxylic acid transporter, mediates the transmembrane transport of metabolites such as lactate, connecting the metabolic interaction between tumor cells and immune cells. These three molecules form a key node in the "immuno-metabolic" interaction network. Among them, C1q was established as a core target due to its core role in immunosuppression, its strong correlation with prognosis, and its clear feasibility for targeted therapy.

[0065] Furthermore, the assessment recommendation module 30 is also used to provide personalized treatment guidance recommendations for target patients who belong to the immunometabolism-suppressive subtype and whose treatment response score is less than a preset score, using F4 / 80 modified lipid nanoparticles to deliver si-C1qa in combination with anti-PD-1 antibody.

[0066] Understandably, the assessment and recommendation module identifies a target patient group that belongs to the immunosuppressive subtype (IMT3) and whose treatment response score is below a preset threshold (e.g., 70 points, though other values ​​can also be used; this embodiment does not impose such limitations). This group typically exhibits a strong immunosuppressive microenvironment and abnormal metabolic characteristics, resulting in poor response to conventional treatments and a poor prognosis. For this high-risk patient group, the module specifically recommends an innovative approach using F4 / 80 modified lipid nanoparticles to deliver si-C1qa in combination with anti-PD-1 antibodies. The scientific principle behind this approach is that F4 / 80 is a specific marker on the surface of macrophages; by modifying lipid nanoparticles, immunosuppressive macrophages in the tumor microenvironment can be targeted. The precise targeted delivery of si-C1qa, as a small interfering RNA, can specifically silence C1q gene expression. C1q, as a key component of the complement system, is abnormally highly expressed in the immunosuppressive subtype and is a core molecule driving the formation of the immunosuppressive microenvironment. At the same time, the combination with anti-PD-1 antibody can relieve T cell function suppression. The synergistic effect of the two can effectively reverse the immunosuppressive state and restore the anti-tumor immune response. This treatment strategy has been verified in organoid models and animal experiments to significantly reduce pleural effusion and prolong patient survival. It breaks through the limitations of the "one-size-fits-all" approach in the treatment of traditional malignant pleural effusion and realizes precise targeted therapy based on molecular subtyping. It provides a new treatment option for patients with malignant pleural effusion with poor prognosis and is a successful example of translational medicine research from basic discovery to clinical application.

[0067] Furthermore, the assessment recommendation module 30 is also used to construct organoid models using malignant pleural effusion cells derived from patients, and to verify the impact of key targets in the candidate target set on the treatment response.

[0068] It should be noted that the assessment and recommendation module utilizes patient-derived malignant pleural effusion cells to construct an organoid model. This model can highly simulate the complex structure and functional characteristics of the tumor microenvironment in vivo, including the three-dimensional spatial arrangement and interactions of tumor cells, immune cells, and stromal cells. Subsequently, the module implements targeted interventions (e.g., gene knockout, RNA interference, or small molecule inhibitor treatment) on key molecules in the candidate target set (e.g., C1q, LDHA, and MCT1), and systematically observes the phenotypic changes of the organoids before and after intervention, including indicators such as cell proliferation, apoptosis, immune cell infiltration, and metabolic characteristics. By quantitatively analyzing these changes, the module can directly verify the impact of key targets on the treatment response, such as whether C1q targeted intervention can reverse the immunosuppressive microenvironment and restore T cell function. This not only confirms the accuracy of the predictive model but also provides experimental evidence for optimizing personalized treatment recommendations, enabling the assessment and recommendation module to recommend treatment plans based on experimental evidence rather than solely on computational predictions, significantly improving the scientific rigor and reliability of precision diagnosis and treatment of malignant pleural effusion.

[0069] In the specific implementation, see Table 1. Table 1 below is a schematic table of IMT model classification and characteristics of the multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device, showing the malignant pleural effusion immunometabolic classification (ImmuneMetabolic) The Typing (IMT) model, based on multi-omics data integration analysis, classifies patients with malignant pleural effusion into three subtypes with significant biological differences and clinical prognostic value: the immunometabolic activation subtype (IMT1), the immunometabolic transition subtype (IMT2), and the immunometabolic suppression subtype (IMT3). IMT1 (immunometabolically activated subtype) is characterized by an abundance of anti-tumor immune cells, including effector cell subsets such as CD8+ T, NK, Th1, and γδT17, while immunosuppressive cells such as tumor-associated macrophages (TAMs) and regulatory T cells (Tregs) are less abundant. At the molecular pathway level, inflammatory responses, T cell receptor signaling, and antigen presentation pathways are significantly activated. Cytokine profiling shows a significant increase in pro-inflammatory factors such as CXCL9 / 10 / 11, IL-12 / IL-18, and IFN-γ. Metabolite analysis shows… Enrichment of immune activation-related metabolites such as glutamine, α-ketoglutarate, and creatine; in terms of clinical prognosis, IMT1 patients have the longest median survival; IMT2 (immunometabolic transitional subtype) is characterized by a decrease in the number of anti-tumor immune cells and an increase in intermediate-state cells such as exhausted precursor CD8+ T cells; activation of molecular pathways such as TGF-β signaling, glycolysis, and lipid oxidative phosphorylation; a cytokine environment exhibiting a mixed state of pro-inflammatory and inhibitory factors, including IL-6, VEGF-A, and CCL2; metabolites characterized by an increase in energy metabolism intermediates such as pyruvate and acetyl-CoA; moderate clinical prognosis; the most significant feature of IMT3 (immunometabolic suppression subtype) is the significant infiltration of immunosuppressive cells, including C1Q+TAM, TNFR2+Treg, SIRPG+CD8+ T cells, and cancer-associated fibroblasts (C-TAM). Fibroblasts (CAFs), etc.; activation of pathways such as fatty acid synthesis and tryptophan metabolism in molecular pathways; cytokine profiling shows high enrichment of inhibitory factors, such as CXCL16, SIRPG, C1q, TGF-β, PD-L1, and IDO1; metabolite analysis shows enrichment of immunosuppression-related metabolites such as lactate, palmitic acid, fatty acids, and kynurenine; the worst clinical prognosis; this classification model not only reveals the spatiotemporal heterogeneity of immune and metabolic interactions in the microenvironment of malignant pleural effusion, but also provides precise molecular evidence for patient prognosis assessment and individualized treatment.

[0070] Table 1: IMT model classification and characteristics of a multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device:

[0071]

[0072] In the specific implementation, see Figure 2 , Figure 2 A schematic diagram of an IMT model for predicting the prognosis of MPE, constructed by integrating multi-omics data from a spatiotemporal heterogeneity analysis device for multi-omics immunometabolic reprogramming of malignant pleural effusion, is shown below. Figure 2 As shown, (A) is a flowchart of multi-omics sequencing and model construction; (B) volcano plots, as shown by univariate Cox analysis, display immune metabolism-related genes significantly associated with overall survival (OS) in MPE patients; (C) tsne plot of immune cell distribution in MPE single-cell sequencing; (D) volcano plots, as shown by univariate Cox analysis, display metabolites significantly associated with overall survival (OS) in MPE patients; (E) volcano plots, as shown by univariate Cox analysis, display cytokines significantly associated with overall survival (OS) in MPE patients; (F) heatmaps showing activation of different metabolic and immune-related pathways in MPE; (G) consensus clustering to identify the IMT group in the training cohort; and (H) KM survival curves showing the survival of different IMT subgroups in the internal cohort.

[0073] In the specific implementation, see Figure 3 , Figure 3 A schematic diagram of the IMT model for molecular typing and prognostic prediction in a multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device, as shown below. Figure 3 As shown, (A) heatmap shows the differentially expressed immunometabolological genes in the IMT subgroup during MPE transcriptome sequencing; (B) heatmap shows the distribution map of cell subpopulations in the IMT subgroup during MPE single-cell sequencing; (C) heatmap shows the differentially expressed metabolites in the IMT subgroup during MPE metabolomics sequencing; (D) heatmap shows the differentially expressed cytokines in the IMT subgroup during MPE cytokine microarray sequencing; (E) KM survival curve shows the survival of different IMT subgroups in the external validation cohort; (F) ROC curve shows the prediction of MPE patient survival by the IMT model, LENT model, and PROMISE model in the external validation cohort.

[0074] In the specific implementation, see Figure 4 , Figure 4 A schematic diagram illustrating the efficacy of intrapleural delivery of si-C1qa@F4 / 80 LNP to enhance anti-PD-1 therapy using a multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device. Figure 4As shown, (A) Modification strategy for targeting LNPs: Antibodies were conjugated to the surface of LNPs for active targeting using an antibody modification strategy; (B) LNPs modified with F4 / 80, containing si-C1qa and Cy5.5, were imaged by transmission electron microscopy of si-NC@F4 / 80 LNPs and si-C1qa@F4 / 80 LNPs, with a scale bar of 100 nm; (C) Dynamic light scattering (DLS) was used to measure the hydrodynamic diameters of si-NC@F4 / 80 LNPs and si-C1qa@F4 / 80 LNPs; (DE) Validation of F4 / 80-conjugated LNPs for macrophage delivery; MPE mice were intrapleurally administered @Cy5.5-F4 / 80 LNPs, including LNPs (without antibody conjugation), IgG LNPs, and F4 / 80 LNPs, and MPE was collected 24 hours later to detect Cy5.5 expression; (F) Western blotting was used to detect si-C1qa@F4 / 80 LNPs. LNP (200 nM) significantly inhibited C1qa expression; (GI) MPE mice were randomly divided into five groups: Control group, si-C1qa group, si-C1qa@F4 / 80 LNP group, anti-PD-1 antibody (αPD-1) group and si-C1qa@F4 / 80 LNP + αPD-1 combined treatment group; (G) Tumor burden of mice in different treatment groups was monitored by in vivo imaging on days 8, 11, 14, 20 and 25 (n=4); (H) Kaplan-Meier survival analysis of MPE mice in different treatment groups (n=10); (I) Pleural effusion volume of MPE mice on day 14 (n=5).

[0075] In its specific implementation, this embodiment obtains a dataset by acquiring multi-omics and clinical data on malignant pleural effusion, performs multi-omics integrated analysis on the dataset, constructs an immunometabolic typing model, and obtains molecular typing results; specifically including:

[0076] 1.1 Multi-omics sequencing data and complete clinical follow-up information of patients with malignant pleural effusion were screened in a multicenter cohort to obtain a dataset; the multi-omics data included transcriptomics, single-cell transcriptomics, non-targeted metabolomics, and cytokine microarray data. Clinical information included overall survival and treatment response.

[0077] 1.2 Based on the dataset, a graph neural network algorithm was used for multimodal integration, and patients were divided into three subtypes through consensus clustering: IMT1 (immunometabolic activation subtype, with the longest median survival), IMT2 (immunometabolic transition subtype, with a moderate median survival), and IMT3 (immunometabolic suppression subtype, with the shortest median survival). Cluster analysis showed that IMT3 was characterized by infiltration of immunosuppressive cells and enrichment of metabolites (such as lactate and kynurenine).

[0078] 1.3 Survival analysis showed significant differences in prognosis among the three subtypes: IMT1 vs IMT3, p<0.001; the external validation cohort (N=120) confirmed the robustness of the subtypes, and the predicted AUC was significantly better than that of the traditional model.

[0079] Molecular subtyping results are used to identify key biomarkers and targets, resulting in a candidate target set. Multiple algorithms are then used for optimization and validation to derive a treatment strategy. Specifically, this includes:

[0080] 2.1 Based on IMT3, univariate Cox regression analysis was used to screen key biomarkers, such as C1q and lactate-related genes, to obtain a candidate target set.

[0081] 2.2 Construct a machine learning model, optimize prediction accuracy on the training and validation sets, and select the optimal treatment strategy.

[0082] Based on a candidate target set, targeted therapy strategies are designed and optimized to obtain personalized treatment plans, and treatment response scores are calculated. Specifically, this includes:

[0083] 3.1 Using organoid models and high-throughput screening, we validated the efficacy of targets such as LDHA inhibitors and C1q monoclonal antibodies.

[0084] 3.2 Design nano-targeted delivery systems (such as LNP-encapsulated siRNA) to increase local drug concentration and obtain personalized solutions.

[0085] 3.3 Input patient data into the model to calculate treatment response scores for risk stratification.

[0086] By utilizing treatment response scores and clinical information, prognostic predictions and treatment recommendations are obtained, which are then systematically combined to create an intelligent diagnosis and treatment system; specifically including:

[0087] 4.1 By integrating scores and clinical stages, a nomogram was constructed to predict survival probability, and a calibration curve was used to verify its reliability.

[0088] 4.2 Develop a web application to enable online typing and treatment recommendations, supporting rapid clinical decision-making.

[0089] Therefore, this embodiment realizes an efficient and accurate MPE management tool, which is expected to promote the development of precision medicine.

[0090] It should be noted that in this embodiment, multi-omics and clinical data of malignant pleural effusion are obtained to form a dataset. Multi-omics integration analysis is performed on the dataset to construct an immune metabolic typing model and obtain molecular typing results.

[0091] The molecular typing results are used to identify key biomarkers and targets to obtain a candidate target set. Then, the candidate target set is used for multi-algorithm optimization and verification to obtain a treatment strategy.

[0092] Based on the candidate target set, targeted therapy strategies are designed and optimized to obtain personalized treatment plans, and then the treatment response scores of the test samples are calculated using the treatment plans.

[0093] Using the treatment response score and clinical information, prognostic prediction results and treatment recommendations are obtained. The classification model, treatment strategy and prediction result process are then systematized to obtain an intelligent diagnosis and treatment system, which outputs prognostic risk assessment and treatment guidance results for malignant pleural effusion.

[0094] Multi-omics sequencing data and complete clinical follow-up information of patients with malignant pleural effusion were screened in a multicenter cohort to obtain a dataset; wherein, the multi-omics sequencing data included transcriptome, single-cell transcriptome, non-targeted metabolome and cytokine microarray data;

[0095] Based on the aforementioned dataset, a graph neural network algorithm was selected to integrate multimodal data and, according to consensus clustering analysis, patients were divided into three subtypes: the immunometabolic activation subtype (IMT1), the immunometabolic transition subtype (IMT2), and the immunometabolic suppression subtype (IMT3).

[0096] Based on overall survival and treatment response, the prognostic risk differences among the three subtypes were analyzed to complete immunometabolic typing and obtain molecular typing results.

[0097] Based on the aforementioned inhibitory subtype, univariate Cox regression analysis was used to screen for prognostic immunometabolic biomarkers, resulting in a set of candidate targets driven by high-risk subtypes.

[0098] We select machine learning algorithms such as lasso regression, ridge regression, random survival forest, and support vector machine to construct an algorithm pool, and calculate the prediction accuracy on the training set and validation set. We then use the prediction accuracy to select the optimal model and obtain the treatment strategy.

[0099] Using high-throughput screening and organoid models, key targets, including LDHA, MCT1, and C1q, were screened from the candidate target set.

[0100] Based on the aforementioned key targets, a nano-targeted delivery system and a combination therapy strategy were designed, and the efficacy was verified using in vitro and in vivo models to obtain personalized treatment plans.

[0101] The multi-omics data of the sample to be tested are input into the treatment model to calculate the treatment response score, thus obtaining the treatment response score of the sample to be tested.

[0102] By integrating the treatment response score and clinical staging information, an initial nomogram for predicting survival probability and treatment response is obtained. The initial nomogram is then validated using a calibration curve to verify the reliability of the prediction, resulting in prognostic predictions and treatment recommendations for clinical decision-making.

[0103] Web applications for calculating the classification model and the treatment recommendations were developed to obtain an intelligent diagnosis and treatment system that outputs prognostic risk assessment and treatment guidance results for malignant pleural effusion.

[0104] This embodiment achieves the following technical effects:

[0105] It fills a gap in the field: for the first time, it achieves immunometabolic molecular subtyping for MPE, constructs an IMT model, divides patients into three subtypes, and predicts AUC significantly better than traditional models, thereby improving the biological rationality of the model from the source.

[0106] Excellent predictive performance: The model is validated in an external cohort, with accurate prognostic stratification and can guide treatment response.

[0107] The clinical translation pathway is clear: it integrates targeted therapy and intelligent systems to achieve a unified approach of "subtyping-treatment-monitoring" and supports personalized medicine.

[0108] Multifunctional: It can be used not only for prognostic assessment, but also to predict immunotherapy response and screen targets, and has the potential for "diagnosis-treatment" synergy.

[0109] The system is highly user-friendly: it is deployed via a web application, so users do not need professional skills to use it, making it suitable for widespread adoption.

[0110] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0111] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0112] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A multi-omics device for analyzing the spatiotemporal heterogeneity of immunometabolic reprogramming in malignant pleural effusion, characterized in that, The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device includes: an integration analysis module, a screening and construction module, and an evaluation and suggestion module; wherein... The integrated analysis module is used to acquire a multi-omics dataset consisting of transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data, and cytokine chip data from patients with malignant pleural effusion. It performs multimodal integration and consensus clustering analysis on the multi-omics dataset to obtain integrated analysis results. Based on these results, the training samples of malignant pleural effusion from patients with malignant pleural effusion are classified into three subtypes: immunometabolic activation, immunometabolic transition, and immunometabolic inhibition. The screening and construction module is used to screen out immunometabolic biomarkers from the features of the immunometabolic inhibitory subtype, form a candidate target set based on the immunometabolic biomarkers, and construct a prediction model based on the candidate target set. The assessment and recommendation module is used to input the current integrated analysis results of the test samples of the patients with malignant pleural effusion into the prediction model, calculate the treatment response score and integrate clinical staging information to construct a nomogram, and obtain prognostic risk assessment results and personalized treatment guidance recommendations. The integration and analysis module is also used to perform multimodal integration and consensus clustering analysis on different omics data in the multi-omics dataset using graph neural network algorithms to obtain integration and analysis results; The integrated analysis module is also used to construct a heterogeneous network model using a graph neural network algorithm, using the malignant pleural effusion training samples of the patients with malignant pleural effusion as network nodes, and using the transcriptome sequencing data, single-cell transcriptome sequencing data, non-targeted metabolome data and cytokine chip data of the patients with malignant pleural effusion as multidimensional features of the nodes of the heterogeneous network model. The integrated analysis module is also used to learn the similarity relationship between nodes through the message passing mechanism of graph neural networks, integrate multi-omics information, obtain a low-dimensional embedding representation of each malignant pleural effusion training sample, and form an integrated feature matrix at the sample level. The integration analysis module is also used to perform dimensionality reduction on the integration feature matrix, perform multiple rounds of K-means clustering in the dimensionality reduction space, evaluate the stability of clustering under different K values ​​by calculating the consensus matrix, determine the optimal number of clusters by the cumulative distribution function curve as a plateau characteristic and the relative change area, obtain the final clustering result based on the optimal number of clusters, and use the final clustering result as the integration analysis result.

2. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 1, characterized in that, The integrated analysis module is also used to classify the malignant pleural effusion training samples of the patients with malignant pleural effusion into three subtypes based on the integrated analysis results: an immunometabolic activation subtype with enhanced immune cell activity and active metabolic pathways, an intermediate immunometabolic transition subtype, and an immunometabolic suppression subtype with an immunosuppressive microenvironment and abnormal metabolism.

3. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 1, characterized in that, The screening module is also used to analyze and screen out immunometabolic biomarkers related to the overall survival of patients with malignant pleural effusion from the characteristics of the immunometabolic suppression subtype through univariate Cox proportional hazards regression. The screening and construction module is also used to screen out molecules with clear targeted therapeutic potential based on the biological function and therapeutic feasibility of the immune metabolic markers, and to form a candidate target set. The screening and construction module is also used to construct a prediction model based on the molecular characteristics of the candidate target set.

4. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 3, characterized in that, The screening module is also used to analyze the association network of key metabolites related to the overall survival of patients with malignant pleural effusion from the features of the immunometabolism-inhibiting subtype using univariate Cox proportional hazards regression, and to identify immunometabolism biomarkers with prognostic value from the association network.

5. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 3, characterized in that, The screening and construction module is also used to construct a machine learning algorithm pool consisting of various machine learning algorithms by using the molecular features of the candidate target set as input variables. The screening module is also used to evaluate the performance of each model in the machine learning algorithm pool through five-fold cross-validation, calculate the average AUC value of each model in five validations, and select the model with the highest AUC value as the prediction model.

6. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 1, characterized in that, The assessment and recommendation module is also used to input the current integrated analysis results of the test samples of the patients with malignant pleural effusion into the prediction model, calculate the treatment response score and integrate clinical staging information to construct a nomogram, and obtain prognostic risk assessment results and personalized treatment guidance recommendations.

7. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 3, characterized in that, The candidate target set includes C1q, LDHA and MCT1, with C1q as the core target.

8. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 7, characterized in that, The assessment recommendation module is also used to provide personalized treatment guidance for target patients who belong to the immunometabolism-suppressive subtype and whose treatment response score is less than the preset score, using F4 / 80 modified lipid nanoparticles to deliver si-C1qa in combination with anti-PD-1 antibody.

9. The multi-omics malignant pleural effusion immunometabolic reprogramming spatiotemporal heterogeneity analysis device as described in claim 8, characterized in that, The assessment recommendation module is also used to construct organoid models using malignant pleural effusion cells derived from patients, and to verify the impact of key targets in the candidate target set on treatment response.

Images