A screening method for biomarkers for aiding in the diagnosis of small cell lung cancer

Abstract

The application discloses a screening method of biomarkers for assisting in diagnosing small cell lung cancer, and relates to the technical field of biomedicine. The method acquires multiple groups of samples; the multiple groups of samples include exosome RNA transcriptome sequencing data of multiple SCLC patients and multiple healthy controls; each exosome RNA transcriptome sequencing data includes multiple RNA features; technical quality filtering is performed on the RNA features in all exosome RNA transcriptome sequencing data, and differential expression analysis is performed on the filtered features to determine a candidate RNA set of the SCLC patients and the healthy controls; feature selection is performed on the candidate RNA set through three complementary feature selection methods, and through 20 iterations and 10-fold nested cross-validation, optimal exosome RNA marker combinations are screened from different combinations of RNA features; the optimal exosome RNA marker combinations include LINC00989, CXCL5, MAP3K7CL and TUBB1. The optimal exosome RNA marker combinations screened by the method are beneficial to the diagnosis of small cell lung cancer.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, and in particular to a method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer. Background Technology

[0002] Small cell lung cancer (SCLC) is a highly aggressive subtype of lung cancer with an extremely high mortality rate. It is characterized by round or spindle-shaped cancer cells with little cytoplasm and small size, exhibiting rapid proliferation, a high tendency for early metastasis, and a poor prognosis. The five-year survival rate for SCLC patients remains consistently low at 14-15%, with a median survival of less than two years for early-stage SCLC patients and only about one year for metastatic patients. Due to the lack of specific early diagnostic markers, SCLC patients are often diagnosed at an extensive stage. Therefore, early screening and diagnosis of SCLC are of significant research importance for reducing patient mortality.

[0003] Currently, the main methods for detecting SCLC include medical imaging examinations, pathological biopsies, and liquid biopsies. While low-dose spiral CT is the standard method for early screening and diagnosis of lung cancer, it has a high false-positive rate and requires repeated monitoring, leading to excessive radiation exposure for patients. Pathological biopsies, which involve surgically analyzing tissue sections from the lesion, are the gold standard for SCLC diagnosis, but they are invasive and unsuitable for early screening and diagnosis. Liquid biopsies are a new, non-invasive method that analyzes tumor markers in bodily fluids such as blood, saliva, and urine to achieve early screening and diagnosis of cancer. They are non-invasive, cause minimal harm to patients, and allow for continuous monitoring of high-risk patients. Although traditional serological markers such as neuron-specific enolase (NSE) and progastrin-releasing peptide (ProGRP) are used in SCLC diagnosis, their specificity and sensitivity still fall short of clinical needs. Therefore, the discovery of highly sensitive and specific diagnostic markers is an urgent clinical need. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for screening biomarkers for assisting in the diagnosis of small cell lung cancer, addressing the aforementioned technical problems.

[0005] The present invention adopts the following technical solution: This invention provides a method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer, comprising: Multiple sets of samples were obtained; these samples included exosomal RNA transcriptome sequencing data from multiple SCLC patients and multiple healthy controls; each exosomal RNA transcriptome sequencing data included multiple RNA features. We performed technical quality filtering on RNA features in all exosomal RNA transcriptome sequencing data, and then performed differential expression analysis on the filtered features to determine candidate RNA sets from SCLC patients and healthy controls. The candidate RNA set was selected using three complementary feature selection methods. The optimal combination of exosomal RNA biomarkers was screened from different combinations of RNA features through 20 iterations and 10-fold nested cross-validation. The optimal combination of exosomal RNA biomarkers included LINC00989, CXCL5, MAP3K7CL and TUBB1.

[0006] Optionally, feature selection is performed on the candidate RNA set using three complementary feature selection methods. The optimal combination of exosomal RNA biomarkers is screened from different combinations of RNA features through 20 iterations and 10-fold nested cross-validation, including: LASSO regression, random forest, and support vector machine recursive feature elimination were used to select features from the candidate RNA set. The optimal combination of exosomal RNA biomarkers was screened from 2 to 10 different combinations of RNA features through 20 iterations and 10-fold nested cross-validation.

[0007] Optionally, the criteria for technical quality filtering include: retaining RNA that is expressed in at least 80% of the samples and has a reliable expression level in at least 10% of the samples; a reliable expression level is defined as an RNA count greater than or equal to 5.

[0008] Optionally, differential expression analysis is performed on the filtered features to identify candidate RNA sets from SCLC patients and healthy controls, including: The filtered features were normalized by log2(CPM+1), and the ComBat method was used to eliminate the batch effect introduced by RNA type. Differential expression analysis of exosomal RNA was performed on the characteristics after eliminating batch effects, and expression differences were evaluated by Wald test. A multiple screening criterion was used to identify significantly differentially expressed RNAs and determine the candidate RNA set. The multiple screening criteria included: corrected p value < 0.05, |log2 fold change| > 1.2, and average expression abundance > 50.

[0009] Optionally, the method further includes: Gene Ontology analysis was used to explore the functional enrichment of RNAs in the candidate RNA set in terms of biological processes, molecular functions and cellular components. Pathway enrichment analysis of RNAs in the candidate RNA set was performed using the KEGG database to identify molecular pathways associated with SCLC.

[0010] This invention provides a device for screening biomarkers for the auxiliary diagnosis of small cell lung cancer, comprising: The acquisition module is used to acquire multiple sets of samples; these multiple sets of samples include exosomal RNA transcriptome sequencing data from multiple SCLC patients and multiple healthy controls; each exosomal RNA transcriptome sequencing data includes multiple RNA features. The analysis module is used to perform technical quality filtering on RNA features in all exosomal RNA transcriptome sequencing data, and to perform differential expression analysis on the filtered features to determine the candidate RNA set for SCLC patients and healthy controls. The selection module is used to select features from the candidate RNA set through three complementary feature selection methods. Through 20 iterations and 10-fold nested cross-validation, the optimal combination of exosomal RNA biomarkers is screened from different combinations of RNA features. The optimal combination of exosomal RNA biomarkers includes LINC00989, CXCL5, MAP3K7CL and TUBB1.

[0011] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for screening biomarkers for assisting in the diagnosis of small cell lung cancer.

[0012] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described method for screening biomarkers for assisting in the diagnosis of small cell lung cancer.

[0013] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects: In this invention, RNA features in all exosomal RNA transcriptome sequencing data are technically filtered, and differential expression analysis is performed on the filtered features to determine candidate RNA sets for SCLC patients and healthy controls. Finally, using machine learning methods, multiple feature selection strategies and nested cross-validation are employed to analyze exosomal RNA expression profiles, successfully screening out four exosomal RNA biomarker combinations with potential diagnostic value: LINC00989, CXCL5, MAP3K7CL, and TUBB. The four exosomal RNA combinations screened by machine learning algorithms can integrate complementary molecular information, effectively overcoming the limitations of single biomarkers, and providing a new technical approach for the accurate diagnosis of SCLC and screening of high-risk populations. Attached Figure Description

[0014] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0015] Figure 1 A schematic flowchart of a method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer provided by the present invention; Figure 2 A clustering heatmap of the top 50 differentially expressed RNAs provided by this invention; Figure 3 This invention provides a volcano diagram of differential expression of exosomal RNA between SCLC patients and healthy individuals. Figure 4 A graph illustrating the results of a biological process analysis provided by this invention; Figure 5 A cellular component analysis result diagram provided by the present invention; Figure 6 A molecular function analysis result diagram provided by the present invention; Figure 7 A KEGG pathway enrichment analysis result diagram provided by the present invention; Figure 8 This is a schematic diagram of the enrichment analysis results of GO biological processes provided by the present invention; Figure 9 This is a schematic diagram of the enrichment analysis results of GO cell components provided by the present invention; Figure 10 This is a schematic diagram of the functional enrichment analysis results of GO molecules provided by the present invention; Figure 11 A schematic diagram of KEGG pathway enrichment analysis results provided by this invention; Figure 12 A schematic diagram illustrating the comparative results of diagnostic performance for different combinations of feature quantities provided by the present invention; Figure 13 A schematic diagram comparing the ROC curves of the four exosomal RNA combinations and the single exosomal RNA model provided by this invention; Figure 14 A schematic diagram of a method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer provided by the present invention; Figure 15 A schematic diagram of principal component analysis based on 15,981 RNA features provided by the present invention; Figure 16 This is a schematic diagram comparing the ROC curves of the three models provided by this invention; Figure 17 ROC curves for the diagnosis of SCLC using different machine learning algorithms for three exosomal RNA combinations (CXCL5, TUBB1, and MAP3K7CL) in the GEO database. Figure 18 ROC curves of the four exosomal RNA combinations provided by this invention in different cancer types. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0017] Exosomes are membrane-bound vesicles secreted by cells, containing various biomolecules such as proteins, nucleic acids, and lipids. Their microscopic size ranges from 40 to 160 nm, and they possess a lipid bilayer structure that effectively protects these biomolecules. They are widely distributed in bodily fluids such as blood and saliva. Exosomes enter the bodily fluid circulation from cancer cells via exocytosis, are taken up by recipient cells, and facilitate intercellular communication, closely participating in various stages of tumorigenesis and development. Studies have shown that exosomes have significant value in the early diagnosis, treatment, and drug resistance research of cancer, making them an excellent novel cancer diagnostic biomarker. Therefore, exosomes can better preserve the biological information related to cancer cells, making them an excellent potential biomarker for lung cancer diagnosis. However, research on the correlation and pathways between exosomes and their contents and lung cancer is limited, and corresponding research methods are lacking, severely restricting the research and application of exosomes as lung cancer biomarkers in early diagnosis, precision treatment, and prognosis. In recent years, with the rapid development of high-throughput sequencing and machine learning technologies, exosomes have become important tools in modern bioinformatics research due to their powerful data output and analysis capabilities. Compared to traditional differential expression analysis, machine learning methods can not only capture complex interactions between genes, but also effectively avoid overfitting through cross-validation and feature stability assessment, significantly improving the reliability and generalization ability of biomarkers. Therefore, tumor biomarker screening strategies based on exosome sequencing data using machine learning provide a new research direction for the early diagnosis of malignant tumors such as lung cancer.

[0018] This invention proposes a method for screening biomarkers to aid in the diagnosis of small cell lung cancer (SCLC). This method provides a novel machine learning strategy for rapid and efficient screening of new lung cancer biomarkers and for investigating their pathways of action. The aim is to screen combinations of highly effective diagnostic biomarkers from the exosomal RNA expression profiles of SCLC patients using machine learning. First, using machine learning, a multiple feature selection strategy and nested cross-validation were employed to analyze the exosomal RNA expression profiles, successfully screening four combinations of exosomal RNA biomarkers with potential diagnostic value: LINC00989, CXCL5, MAP3K7CL, and TUBB. These combinations exhibit excellent diagnostic performance, with an area under the receiver operator characteristic (ROC) curve (AUC) value of 0.93, a sensitivity of 0.92, and a specificity of 0.83. Furthermore, the four candidate RNA biomarkers were validated using independent RNA expression profile data from SCLC and other cancers, demonstrating that the screened RNA biomarkers possess high specificity, sensitivity, and good generalization ability. Therefore, the RNA biomarkers screened in this invention are of great research significance for the diagnosis, treatment and prognosis of SCLC. They not only enrich the types of diagnostic biomarkers for SCLC, but also provide potential new targets for liquid biopsy and personalized treatment of SCLC. They are of great research significance for the accurate diagnosis of SCLC and the reduction of mortality in SCLC patients.

[0019] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of a method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer according to the present invention, which specifically includes the following steps: S101, acquire multiple sets of samples; the multiple sets of samples include exosomal RNA transcriptome sequencing data from multiple SCLC patients and multiple healthy controls; each exosomal RNA transcriptome sequencing data includes multiple RNA features.

[0021] Sample Acquisition: Blood exosomal RNA expression profile data were obtained from the exoRBase 2.0 database (http: / / www.exorbase.org / ) for screening diagnostic biomarkers and model building for SCLC. This dataset contains exosomal RNA transcriptome sequencing data from 36 SCLC patients and 118 healthy controls, covering 114,601 RNA features, of which 35,517 are mRNA / lncRNA and 79,084 are circRNA. Expression matrices are provided in normalized count form for subsequent differential expression analysis and machine learning modeling.

[0022] S102 performs technical quality filtering on RNA features in all exosomal RNA transcriptome sequencing data and performs differential expression analysis on the filtered features to determine candidate RNA sets for SCLC patients and healthy controls.

[0023] The criteria for technical quality filtering include: retaining RNA that is expressed in at least 80% of the samples and has a reliable expression level in at least 10% of the samples; a reliable expression level is defined as an RNA count greater than or equal to 5.

[0024] In one embodiment, differential expression analysis is performed on the filtered features to determine the candidate RNA set for SCLC patients and healthy controls. This includes: normalizing the filtered features to log2 (CPM+1) and eliminating batch effects introduced by RNA type using the ComBat method; performing exosomal RNA differential expression analysis on the features after eliminating batch effects, assessing expression differences using the Wald test, and identifying significantly differentially expressed RNAs using multiple screening criteria to determine the candidate RNA set; the multiple screening criteria include: corrected p-value <0.05, |log2 fold change| >1.2, and average expression abundance >50.

[0025] Specifically, all exosomal RNA transcriptome sequencing data from the samples underwent technical quality filtering, retaining RNAs expressed (non-zero) in at least 80% of samples and with reliable expression levels (count ≥ 5) in at least 10% of samples. The filtered RNA feature dataset was normalized to log2 (CPM+1) using the edgeR analysis package in R. Simultaneously, the ComBat method was used to eliminate batch effects caused by RNA type (mRNA / lncRNA and circRNA). Subsequently, differential expression analysis of exosomal RNAs was performed using the DESeq2 analysis package in R, constructing a statistical model that incorporated batch effects based on sample type (SCLC patients and healthy controls) and RNA type. Expression differences were assessed using the Wald test, and multiple screening criteria (corrected p-value < 0.05, |log2 fold change| > 1.2, mean expression abundance > 50) were used to identify significantly differentially expressed RNAs, providing a candidate RNA set for subsequent feature selection.

[0026] In one embodiment, Gene Ontology analysis is used to explore the functional enrichment of RNAs in the candidate RNA set in terms of biological processes, molecular functions, and cellular components, and pathway enrichment analysis of RNAs in the candidate RNA set is performed using the KEGG database to identify molecular pathways associated with SCLC.

[0027] Specifically, gene function annotation and pathway enrichment analysis were performed using the DAVID (Database for Annotation, Visualization and Integrated Discovery) and KOBAS (KEGG Orthology Based Annotation System) online analysis platforms. Gene Ontology (GO) analysis was used to explore the functional enrichment of differentially expressed RNA in biological processes (BP), molecular components (MF), and molecular functions (CC). Pathway enrichment analysis was performed using the KEGG (Kyoto Encyclopedia of Genes and Genomes) database to identify molecular pathways associated with SCLC. Hypergeometric tests were used to assess the significance of enrichment (p-value < 0.05, Benjamini-Hochberg correction).

[0028] In one embodiment, this invention acquired blood exosome RNA sequencing data from 36 SCLC patients and 118 healthy controls, performed a two-layer filtering analysis, and retained 15,981 high-quality RNA features for subsequent analysis. Differential expression analysis was performed using DESeq2, with RNA type as a covariate to control for batch effects. Based on rigorous screening, 149 significantly differentially expressed RNAs were identified from the 15,981 RNAs, forming the candidate RNA set.

[0029] To gain a deeper understanding of the biological functions of 149 differentially expressed RNAs in the candidate RNA pool, this invention employs a multi-level functional enrichment analysis strategy. First, GO analysis was performed using the DAVID online platform, encompassing three dimensions: biological processes, cellular components, and molecular functions. Subsequently, KEGG pathway enrichment analysis was performed using the KOBAS online platform. All enrichment analyses employed hypergeometric tests, with a corrected p-value <0.05 and an enrichment factor >1.5 as significance criteria, and the Benjamini-Hochberg method was used for multiple comparison correction.

[0030] like Figures 2-7 As shown, Figures 2-7 This is a graph showing the results of multi-level functional enrichment analysis using a candidate RNA set. Figure 2 Cluster heatmap of the top 50 differentially expressed RNAs; Figure 3This is a volcano plot showing differential expression of exosomal RNA between SCLC patients and healthy individuals. Red dots represent significantly upregulated genes, blue dots represent significantly downregulated genes, and gray dots represent genes with no significant difference. Screening criteria: |log2FC|>1.2, corrected p-value<0.05, mean expression level>50; Figure 4 A graph showing the results of biological process analysis; Figure 5 This is a graph showing the results of cell component analysis; Figure 6 The results of molecular functional analysis are shown in the figure. Figure 7 The image shows the results of the KEGG pathway enrichment analysis. Based on the above analysis, clustering heatmap analysis of the top 50 differentially expressed RNAs by p-value indicates that SCLC patients and healthy controls exhibit different expression patterns. Figure 3 Furthermore, the expression patterns within the patient group were relatively consistent. To reveal the potential biological functions of differentially expressed RNAs, this invention performed GO and KEGG pathway enrichment analysis. The differentially expressed RNAs were mainly enriched in biological processes such as blood coagulation and platelet activation. Figure 4 ), located in cellular components such as platelet α-granules ( Figure 5 ), participates in molecular functions such as chemokine receptor binding ( Figure 6 KEGG pathway analysis showed that these RNAs are mainly involved in pathways such as focal adhesion, platelet activation, and cytokine receptor interaction. Figure 7 A comprehensive analysis showed that differentially expressed RNAs in the exosomes of SCLC patients were mainly related to platelet function, cell adhesion and migration, and immune regulation, specifically as follows: Figures 8-11 As shown, this provides an important biological basis for subsequent screening of biomarkers, among which, Figure 8 This diagram illustrates the results of the GO biological process enrichment analysis. Figure 9 This diagram illustrates the results of GO cell component enrichment analysis. Figure 10 A schematic diagram showing the results of GO molecule functional enrichment analysis. Figure 11 This diagram illustrates the results of KEGG pathway enrichment analysis. The size of the bubbles represents the number of enriched genes, and the intensity of the color represents the significance level of enrichment (-log10 P value).

[0031] GO biological process analysis revealed key biological mechanisms involving differentially expressed RNAs in exosomes of SCLC. The most significantly enriched processes included blood coagulation, hemostasis, wound healing, and platelet activation. The enrichment of these processes has important pathophysiological significance. Abnormalities in blood coagulation and hemostasis are closely associated with tumor-related thromboembolic complications, a common clinical manifestation in SCLC patients. The enrichment of platelet activation is particularly noteworthy because activated platelets not only promote thrombus formation but also promote tumor angiogenesis and distant metastasis by releasing vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF).

[0032] Cellular component analysis revealed that differentially expressed RNA was mainly located in platelet alpha granules, platelet alpha granule lumen, and secretory granule lumen. Platelet alpha granules are the largest and most important secretory granules of platelets, containing abundant proteins, including coagulation factors, growth factors, chemokines, and angiogenesis regulators. These granules release their contents into the extracellular environment upon platelet activation, participating in hemostasis, inflammatory responses, and tissue repair. In the tumor microenvironment, abnormal release of platelet alpha granules may promote tumor cell invasion and metastasis.

[0033] Molecular functional analysis revealed key molecular interaction mechanisms involved in differentially expressed RNAs. Enrichment of CXCR chemokine receptor binding suggests the important role of the chemokine system in the development of SCLC. Chemokines and their receptor systems not only regulate the recruitment and activation of immune cells but also directly affect tumor cell proliferation, migration, and invasion. Enrichment of chemokine activity further supports this view. Enrichment of extracellular matrix binding function reflects the complexity of the interaction between SCLC cells and the matrix microenvironment, an interaction that is a key step in tumor invasion and metastasis.

[0034] KEGG pathway analysis provides a systematic view of molecular pathways involved in differentially expressed RNA. The significant enrichment of the focal adhesion pathway reflects the importance of cell-matrix interactions in SCLC. This pathway involves integrin-mediated cell adhesion and regulates cell morphology, migration, and survival. The enrichment of the platelet activation pathway is consistent with the findings in the GO analysis, highlighting the central role of platelet function in the pathological process of SCLC. The enrichment of the actin cytoskeleton regulation pathway suggests a crucial role of cytoskeleton remodeling in the acquisition of cell motility in SCLC.

[0035] The functional enrichment analysis results were highly consistent with the pathological features of SCLC. SCLC is characterized by rapid proliferation and early widespread metastasis. The abnormalities in platelet activation, cytoskeleton regulation, and the chemokine system discovered in this invention precisely explain the molecular basis of these clinical features. Platelet activation promotes hematogenous dissemination of tumor cells, cytoskeleton remodeling enhances the motility of tumor cells, while dysregulation of the chemokine system creates a favorable migration microenvironment for tumor cells.

[0036] Based on the results of functional enrichment analysis, this invention focuses on genes involved in the aforementioned key biological processes as potential diagnostic biomarker candidates. This provides a biological basis for subsequent machine learning feature screening, improving screening efficiency and biological rationale. The four biomarkers ultimately selected—LINC00989, CXCL5, MAP3K7CL, and TUBB1—are all closely related to these enriched biological functions, validating the effectiveness of the function-guided screening strategy.

[0037] S103 uses three complementary feature selection methods to select features from the candidate RNA set. Through 20 iterations and 10-fold nested cross-validation, the optimal combination of exosomal RNA biomarkers is screened from different combinations of RNA features. The optimal combination of exosomal RNA biomarkers includes LINC00989, CXCL5, MAP3K7CL and TUBB1.

[0038] In one embodiment, feature selection of the candidate RNA set is performed using three complementary feature selection methods. The optimal combination of exosomal RNA biomarkers is screened from different combinations of RNA features through 20 iterations and 10-fold nested cross-validation. This includes: using LASSO regression, random forest and support vector machine recursive feature elimination to perform feature selection of the candidate RNA set, and screening the optimal combination of exosomal RNA biomarkers from 2-10 different combinations of RNA features through 20 iterations and 10-fold nested cross-validation.

[0039] Optionally, three machine learning models were constructed using three feature selection methods: LASSO regression, random forest, and support vector machine recursive feature elimination. These three machine learning models were used to perform feature analysis on the candidate RNA set after preprocessing and differential expression screening. Using a 10-fold nested cross-validation method, after 20 iterations, the optimal combination of exosomal RNA biomarkers with RNA combinations in the range of 2-10 was determined to be specific to small cell lung cancer.

[0040] Specifically, to address the challenges of small sample size and class imbalance, a comprehensive feature selection and model evaluation process based on nested cross-validation was constructed. This process integrates three complementary feature selection methods and ensures the reliability of biomarkers through stability evaluation. Within this framework, the present invention implements three complementary feature selection methods: (1) LASSO regression, which automatically selects the optimal penalty parameter and extracts features with non-zero coefficients; (2) Random forest feature importance analysis, which evaluates feature contribution based on the reduction in average impurity; and (3) Support Vector Machine-Recursive Feature Elimination (SVM-RFE), which sequentially evaluates feature importance. Feature stability scores are calculated through 20 iterations, and the cross-validation consistency of features is evaluated through 10-fold nested cross-validation. Features that show high consistency in cross-validation are selected as candidate biomarkers. At the same time, the present invention evaluates combinations of different numbers of features (2-10) to determine the optimal feature subset that produces the best diagnostic efficacy. Throughout the evaluation process, the outer 10-fold cross-validation is used to evaluate the model's generalization ability, and the inner 5-fold cross-validation is used for feature selection and parameter optimization. The outer layer divides the sample data into a training set (90%) and a test set (10%) to ensure that test set information is not leaked into the feature selection process, thereby avoiding performance evaluation bias. This invention employs a Support Vector Machine (SVM) with a radial basis function (RBF) kernel as the classifier, optimizing model parameters through grid search. Simultaneously, the SMOTE (synthetic minority oversampling technique) is used to address class imbalance and improve the model's sensitivity to SCLC samples. The final model performance is comprehensively evaluated using ROC curves for AUC, sensitivity, specificity, and accuracy.

[0041] In one embodiment, to further screen exosomal RNA biomarkers associated with small cell lung cancer (SCLC), three complementary feature selection algorithms were applied to 149 differentially expressed RNAs (candidate RNA set): LASSO regression, random forest feature importance analysis, and SVM-RFE. The stability score of each feature was calculated through 20 iterations, and the cross-validation consistency of the features was evaluated using 10-fold nested cross-validation. Several highly stable candidate features were identified, and the optimal biomarker combination consisting of LINC00989, CXCL5, MAP3K7CL, and TUBB1 was finally determined, as shown in Table 1. Table 1 presents the detailed evaluation results of all highly stable RNAs during the feature selection process, with p < 0.05. All four exosomal RNAs were identified as important features in all algorithms, exhibiting the highest feature selection consistency (10 / 10) and the highest p-value, and were significant in the exosomes of SCLC patients, indicating that they have high reliability and stability as diagnostic biomarkers for SCLC. Figure 12 As shown, Figure 12 The comparison of diagnostic performance for different combinations of feature numbers shows the trends in AUC, sensitivity, specificity, and overall score for combinations of 2-10 features. The combination of 4 features exhibited the best overall score, achieving excellent diagnostic performance while maintaining a relatively small number of features. Figure 13 As shown, Figure 13 ROC curves were compared between the four exosomal RNA combinations and the single exosomal RNA model. The ROC curve analysis showed that while individual exosomal RNAs all exhibited good diagnostic ability (AUC 0.89), the combination of the four exosomal RNAs demonstrated superior overall diagnostic performance (AUC 0.93, Sensitivity 0.92, Specificity 0.83). The advantage of multi-marker combinations lies in their ability to integrate complementary information from different molecular markers, more comprehensively reflecting disease characteristics and thus achieving more accurate diagnosis.

[0042] Table 1 Feature selection screening process and stability evaluation results In one embodiment, such as Figure 14As shown, this invention also provides a method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer. Specifically, this embodiment includes: collecting exosomal RNA sequencing data from 36 SCLC patients and 118 healthy controls, covering 114,601 RNA features. All data were filtered, standardized, and batch-corrected to obtain 149 differentially expressed RNAs. Feature selection was performed using LASSO regression, random forest, and support vector machine recursive feature elimination. The optimal combination was selected from 2-10 different combinations of features through 20 iterations and 10-fold nested cross-validation. Finally, the optimal four RNA combinations were selected: LINC00989, CXCL5, MAP3K7CL, and TUBB1. Validation and evaluation of this combination showed excellent performance in distinguishing SCLC patients from healthy controls (areaunder curve (AUC) = 0.93, sensitivity 0.92, specificity 0.83). Pathway enrichment analysis was performed on 149 differentially expressed RNAs using GO and KEGG to assess the significance of enrichment.

[0043] Furthermore, RNA expression profile data from SCLC patient tumor tissues were used to validate the diagnostic performance of a subset of three exosomal RNAs (CXCL5, MAP3K7CL, and TUBB1), demonstrating good diagnostic performance (AUC = 0.81). The specificity of the screened biomarkers was validated using exosomal RNA expression profile data from gastric cancer, liver cancer, and breast cancer, showing that this biomarker combination has high specificity for SCLC (AUC: SCLC, 0.93; gastric cancer, 0.44; liver cancer, 0.57; breast cancer, 0.73). Therefore, exosomal RNA, as a diagnostic biomarker for SCLC, provides a new research approach and direction for the early diagnosis of SCLC.

[0044] In one embodiment, to investigate the differences in exosomal RNA expression data between SCLC patients and healthy individuals, principal component analysis (PCA) was first used to extract principal features from 15,981 preprocessed RNA features for unsupervised cluster analysis. For example... Figure 15 As shown, Figure 15 Principal component analysis based on 15,981 RNA features was performed, where red triangles represent SCLC patients and green dots represent normal healthy controls. The SCLC patient samples and the healthy control samples showed two distinct classes in the principal component space, forming relatively independent clustering regions, indicating that exosome RNA can effectively distinguish between SCLC patients and healthy individuals.

[0045] To evaluate the rationality of each step in the screening process, this invention compared the diagnostic performance of three models: a full-feature diagnostic model based on 15,981 pre-processed RNAs, a diagnostic model based on 149 differentially expressed RNAs after initial screening, and a diagnostic model based on the four exosomal RNA combinations (LINC00989, CXCL5, MAP3K7CL, and TUBB1) selected in the final screening of this invention. Figure 16 As shown, Figure 16 A comparison of ROC curves for three models was conducted, including models based on whole RNA features, 149 differentially expressed RNAs, and four combinations of exosomal RNAs. ROC curve analysis showed that the model based on 15,981 RNAs exhibited good diagnostic efficacy (AUC=0.94, blue curve). The combination of 149 differentially expressed RNAs showed the highest diagnostic performance (AUC=0.99, black curve), significantly improved due to the removal of a large amount of interfering information. The final four exosomal RNA combinations, while significantly reducing the number of features (97.3% compared to the 149 differentially expressed RNAs) still maintained high diagnostic accuracy (AUC=0.93, red curve), only 6.0% lower than the 149 differentially expressed RNA model. This result indicates that the feature selection process effectively extracted the most diagnostically valuable information from the 149 differentially expressed RNAs.

[0046] In one embodiment, to evaluate the generalization ability and cancer specificity of the screened diagnostic biomarkers, the present invention obtained validation datasets from two sources. First, RNA expression profile data from SCLC patient tumor tissues were extracted from the Gene Expression Omnibus (GEO) database (https: / / www.ncbi.nlm.nih.gov / theGEO / , GSE60052). This cohort included 7 normal lung tissue control samples and 79 SCLC tumor tissue samples for independent external validation. Second, blood exosome RNA expression data from other cancer types were obtained from the exoRBase 2.0 database, including 9 patients with gastric cancer, 112 patients with liver cancer, and 140 patients with breast cancer, to evaluate the diagnostic specificity of the screened biomarkers for SCLC.

[0047] Because the GEOs dataset only contains expression data for three mRNAs (CXCL5, TUBB1, and MAP3K7CL), and lacks data for LINC00989, the diagnostic performance of the three exosomal mRNA combinations was evaluated. The results showed good diagnostic efficacy, with the random forest model achieving the best performance (AUC = 0.84), followed by SVM (AUC = 0.82), and KNN showing relatively low performance (AUC = 0.66), as shown in Table 2. Figure 17 As shown, Figure 17ROC curves for SCLC diagnosis of three exosomal RNA combinations (CXCL5, TUBB1, and MAP3K7CL) in the GEO database were obtained using different machine learning algorithms. Although expression data for LINC00989 was lacking in the validation set, all three exosomal mRNA models demonstrated good performance, indirectly indicating the effectiveness of the feature selection strategy provided in this invention. Considering that these three exosomal RNAs showed the same diagnostic value as LINC00989 in the original dataset, it is speculated that the four exosomal RNAs and their combinations also possess excellent diagnostic performance in a broader patient population.

[0048] To further validate the diagnostic specificity of the four exosomal RNA combinations for SCLC, exosomal RNA expression profiles from three other cancer types—gastric cancer, liver cancer, and breast cancer—were used. Figure 18 As shown, Figure 18 The ROC curves for four exosomal RNA combinations in different cancer types are shown. The AUC values ​​were 0.93 for SCLC, 0.44 for gastric cancer, 0.57 for liver cancer, and 0.73 for breast cancer. ROC curve analysis revealed significant performance differences among the four exosomal RNA combinations across different cancer types. In gastric cancer, the AUC value was only 0.44; in liver cancer, it was 0.57; and in breast cancer, it reached 0.73. These values ​​are significantly lower than the performance in SCLC (AUC = 0.93), indicating that the four selected exosomal RNA biomarkers have high diagnostic specificity for SCLC.

[0049] Table 2. Performance of three exosomal RNA combinations (CXCL5, MAP3K7CL, TUBB1) on the external validation set. This invention successfully screened four exosomal RNAs (LINC00989, CXCL5, MAP3K7CL, and TUBB1) from the exosomal RNA expression profiles of SCLC patients' blood as potential diagnostic biomarkers for SCLC through bioinformatics analysis and a rigorous feature selection strategy. The four screened exosomal RNAs and their combinations exhibited excellent diagnostic performance in distinguishing SCLC patients from healthy individuals, with an AUC of 0.93, a sensitivity of 0.92, and a specificity of 0.83. This provides new biomarkers for the accurate diagnosis of SCLC and also offers a new perspective for exploring the molecular pathological mechanisms of SCLC.

[0050] The challenges in SCLC diagnosis mainly lie in two aspects: first, patients are often diagnosed at an advanced stage, and the diagnostic performance of traditional serological markers such as NSE and ProGRP is limited; second, the high molecular heterogeneity of SCLC makes early diagnosis difficult with a single marker. Against this backdrop, the multi-marker combination strategy using exosomal RNA has demonstrated unique advantages. A combination of four exosomal RNAs screened using machine learning algorithms can integrate complementary molecular information, effectively overcoming the limitations of single markers and providing a new technical pathway for the accurate diagnosis of SCLC and screening of high-risk populations.

[0051] Compared with existing studies, this invention has several key features: the combination of four exosomal RNAs exhibits higher sensitivity and specificity than traditional SCLC serological markers (with a sensitivity typically of 60-70%), indicating that exosomal RNAs are excellent diagnostic markers and providing new possibilities for early clinical screening of lung cancer patients. The application of nested cross-validation and SMOTE technology effectively addresses the problems of small sample size and class imbalance, ensuring the reliability of the model evaluation.

[0052] Based on the GEO dataset, the diagnostic performance of three exosomal mRNAs (CXCL5, MAP3K7CL, and TUBB1) was evaluated. Despite the lack of LINC00989 data, the combination of the three exosomal mRNAs maintained good diagnostic efficacy in the independent validation cohort (AUC=0.84 in the random forest model), confirming the effectiveness of the feature selection strategy of this invention. Furthermore, in other cancer datasets, the combination of four exosomal RNA biomarkers showed low diagnostic performance (gastric cancer, AUC=0.44; liver cancer, AUC=0.57; breast cancer, AUC=0.73), showing a significant difference compared to SCLC (AUC=0.93), indicating that these exosomal RNAs may be involved in the pathophysiological processes specific to SCLC, providing important clues for a deeper understanding of its molecular mechanisms.

[0053] Four screened biomarkers (LINC00989, CXCL5, MAP3K7CL, and TUBB1) were significantly downregulated in SCLC exosomes, which has important biological significance. LINC00989 is a long non-coding RNA that acts as a tumor suppressor in various cancers, primarily participating in tumor progression by regulating platelet function and immune responses. CXCL5, a key chemokine, regulates immune responses and cell migration through the CXCR2 receptor; its downregulation in exosomes may reflect the molecular mechanism by which SCLC cells maintain an immunosuppressive state. MAP3K7C, a negative regulator of the p38 MAPK signaling pathway, leads to overactivation of pro-inflammatory pathways and affects the regulation of B cell and T cell function, participating in the formation of immune escape mechanisms. TUBB encodes β-tubulin 1, which is specifically expressed mainly in platelets and megakaryocytes, participating in cytoskeleton regulation and the regulation of immune cell infiltration. The analytical results of this invention are in high agreement with previously reported results, indicating that the screened exosomal RNAs play an important regulatory role in platelet function, immune regulation, and cytoskeleton stability in SCLC.

[0054] Among them, the molecular mechanism analysis of SCLC exosomal RNA diagnostic biomarkers: 1. Molecular mechanism and functional analysis of LINC00989 LINC00989 is a long non-coding RNA that is significantly downregulated in SCLC exosomes (log2FC = -3.30). Studies have shown that this lncRNA, as a methylation-driven prognostic model component, exhibits a downregulation trend in hepatocellular carcinoma and is significantly associated with patient survival. Tumor Education Platelets (TEPs) studies (encompassing 55 healthy controls and 228 cancer patients) found widespread downregulation of LINC00989 in platelets from cancer patients. These studies support the SCLC exosome data presented in this invention. Mechanistic studies indicate that LINC00989 and RGS18 (a negative regulator of G protein receptors that inhibits platelet activation) are synergistically downregulated in cervical cancer exosomes, potentially promoting tumor progression by activating platelets, which is highly consistent with the "platelet activation pathway" in GO enrichment analysis of SCLC. Simultaneously, breast cancer studies have shown that its involvement in lncRNA tags is associated with immune responses and cytokine pathways, suggesting that it may mediate immune escape. In summary, LINC00989 may exert its effects in SCLC through a three-pronged approach: synergistically regulating platelet functional genes (such as RGS18) to promote early hematogenous metastasis; dysregulating immune pathways to facilitate immune escape; and acting as an intercellular communication factor to downregulate and reshape the tumor microenvironment. However, its specific molecular mechanisms in SCLC still require further experimental verification.

[0055] 2. Chemokine network regulatory mechanism of CXCL5 CXCL5 (CXC motif chemokine ligand 5) is an important member of the chemokine family, exerting its biological functions by binding to the CXCR2 receptor and playing a key role in inflammatory responses and tumor microenvironment regulation. In this study, CXCL5 was significantly downregulated in SCLC exosomes (log2FC = -3.75), a finding that interestingly contrasts with its expression pattern in most solid tumor tissues. Numerous studies have confirmed that CXCL5 is significantly upregulated in various malignant tumors, including gastric cancer, colorectal cancer, pancreatic cancer, head and neck squamous cell carcinoma, and non-small cell lung cancer, participating in processes such as epithelial-mesenchymal transition, tumor cell proliferation, metastatic cell migration, and angiogenesis. Lung cancer-related studies have shown that CXCL5 is significantly overexpressed in cancer cells, leading to upregulation of PD-L1 expression through phosphorylation of the PXN / AKT signaling pathway, thereby hindering CD8+ T cell immune function and positively correlated with poor patient prognosis. This study also confirmed that CXCL5 can recruit neutrophils, and these PD-L1-positive neutrophils further exacerbate CD8+ T cell exhaustion. Importantly, the expression pattern of CXCL5 in exosomes differs from its expression in tissues. Studies of cervical cancer plasma exosomes have found that while CXCL5 is upregulated in cervical cancer tissues, it is significantly downregulated in exosomes, highly consistent with the findings of this invention in SCLC exosomes. In SCLC-related studies, circulating tumor cells have been found to secrete ENA-78 / CXCL5, recruiting neutrophils with angiogenic properties and promoting tumor invasion and angiogenesis. The downregulation of CXCL5 in exosomes may reflect that SCLC cells maintain an immunosuppressive state by reducing the long-range propagation of pro-inflammatory signals, consistent with the common immunosuppressive phenotype in SCLC patients, and directly related to the chemokine activity and extracellular matrix interaction pathways highlighted in the enrichment analysis results of this invention.

[0056] 3. Signal path modulation function of MAP3K7CL MAP3K7CL (MAP3K7 C-terminal like), also known as TAK1L (TGF-beta activated kinase like) or C21orf7, is a kinase-related gene. It is significantly downregulated in SCLC exosomes (log2FC = -3.16). This gene exhibits a consistent downregulation pattern in lung diseases. RNA sequencing analysis revealed that MAP3K7CL is significantly downregulated in tumor-educational leukocytes of non-small cell lung cancer (NSCLC) patients. Large-scale gene expression analysis has revealed the unique tissue distribution characteristics of MAP3K7CL. In some studies, systematic analysis of 459 human kinase genes in 5681 tissue samples revealed that MAP3K7CL expression is blood / immune tissue specific (e.g., lymphoma, peripheral leukocytes), highly expressed in lymphoma, mesenchymal stem cells, and peripheral leukocytes, and closely related to multiple biological processes such as B cell and T cell signal transduction, immune responses, signal transduction, metabolism, mesodermal development, and cytoskeleton organization. Functional mechanism studies have shown that MAP3K7CL, as a negative regulator of the p38MAPK signaling pathway, leads to overactivation of pro-inflammatory pathways and promotes the proliferation of tumor-associated fibroblasts, clinically associated with prolonged patient survival, thus establishing its tumor suppressor function. In blood biomarker studies, MAP3K7CL was significantly downregulated in platelets of sepsis patients and in cell-free RNA detection of gastrointestinal cancers, validating its value in liquid biopsy. Based on this evidence, the significant downregulation of MAP3K7CL in SCLC exosomes has important pathological significance. As a negative regulator of the p38MAPK signaling pathway, its downregulation leads to overactivation of pro-inflammatory pathways, promoting a chronic inflammatory state in the tumor microenvironment, which is highly consistent with the inflammation-related pathway alterations found in the enrichment analysis of this invention. Furthermore, considering the important role of MAP3K7CL in the regulation of B cell and T cell function, its downregulation may significantly affect the anti-tumor immune response in SCLC patients, promoting the formation of immune escape mechanisms.

[0057] 4. Cytoskeleton regulatory mechanisms of TUBB1 TUBB1, encoding β-tubulin 1, is an important member of the tubulin family and is specifically expressed primarily in platelets and megakaryocytes. It is significantly downregulated in SCLC exosomes (log2FC = -2.79), consistent with multiple cancer studies. Studies in breast cancer and pancreatic ductal adenocarcinoma have found TUBB1 to be significantly downregulated as a prognostic gene, involved in key processes such as spindle formation and cell division; its aberrant expression may disrupt cell cycle regulation and promote tumor progression. Lung adenocarcinoma studies further confirm that the tubulin β-1 chain encoded by TUBB1 is mainly expressed in platelets and megakaryocytes, and its dysfunction may lead to genomic instability. Notably, the platelet-specific expression of TUBB1 is highly consistent with platelet-related pathways (platelet α-granules, platelet activation) in the enrichment analysis of this invention, suggesting that platelet-derived exosomes can carry molecular information reflecting platelet functional status. Tumor immunology studies have shown that TUBB1 expression is correlated with the abundance of various tumor-infiltrating immune cells, particularly positively correlated with γδ T cells and negatively correlated with follicular helper T cells, and can serve as a potential target for various drugs. Based on these studies, the downregulation of TUBB1 in SCLC exosomes may exert its effects through a dual mechanism: regulating platelet activation and aggregation to promote tumor hematogenous dissemination (consistent with the early widespread metastasis characteristics of SCLC); and influencing immune cell infiltration to alter the tumor microenvironment and promote immune escape.

[0058] This invention utilizes bioinformatics analysis and machine learning methods, combining multi-algorithm fusion feature selection with nested cross-validation to establish a novel biomarker screening strategy suitable for small-sample, high-dimensional data. Four exosomal RNA diagnostic biomarkers (LINC00989, CXCL5, MAP3K7CL, and TUBB) were screened from the exosomal RNA expression profiles of SCLC patients' blood. The four exosomal RNAs and their combinations demonstrated excellent performance in the diagnosis of SCLC patients and healthy individuals (AUC=0.93, sensitivity=0.92, specificity=0.83), with significantly reduced diagnostic efficacy in other cancer types, exhibiting good SCLC specificity. This provides a research protocol for specific biomarker screening and early diagnosis of SCLC, and has significant research value in the precision diagnosis of lung cancer.

[0059] When applying the biomarker screening method for assisting in the diagnosis of small cell lung cancer provided by this invention, it is not necessary to rely on... Figure 1 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this invention does not impose any restrictions on it.

[0060] The above describes a method for screening biomarkers for assisting in the diagnosis of small cell lung cancer, provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding device for screening biomarkers for assisting in the diagnosis of small cell lung cancer, the device comprising: The acquisition module is used to acquire multiple sets of samples; these multiple sets of samples include exosomal RNA transcriptome sequencing data from multiple SCLC patients and multiple healthy controls; each exosomal RNA transcriptome sequencing data includes multiple RNA features. The analysis module is used to perform technical quality filtering on RNA features in all exosomal RNA transcriptome sequencing data, and to perform differential expression analysis on the filtered features to determine the candidate RNA set for SCLC patients and healthy controls. The selection module is used to select features from the candidate RNA set through three complementary feature selection methods. Through 20 iterations and 10-fold nested cross-validation, the optimal combination of exosomal RNA biomarkers is screened from different combinations of RNA features. The optimal combination of exosomal RNA biomarkers includes LINC00989, CXCL5, MAP3K7CL and TUBB1.

[0061] Specific limitations regarding the screening device for biomarkers used in the auxiliary diagnosis of small cell lung cancer can be found in the limitations of the screening method for biomarkers used in the auxiliary diagnosis of small cell lung cancer mentioned above, and will not be repeated here. Each module in the aforementioned screening device for biomarkers used in the auxiliary diagnosis of small cell lung cancer can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0062] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 The provided method for screening biomarkers to aid in the diagnosis of small cell lung cancer.

[0063] This invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for various operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above-mentioned functions. Figure 1 The provided method for screening biomarkers to aid in the diagnosis of small cell lung cancer.

[0064] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0065] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A method for screening biomarkers for the auxiliary diagnosis of small cell lung cancer, characterized in that, include: Multiple sets of samples were obtained; these samples included exosomal RNA transcriptome sequencing data from multiple SCLC patients and multiple healthy controls; each exosomal RNA transcriptome sequencing data included multiple RNA features. We performed technical quality filtering on RNA features in all exosomal RNA transcriptome sequencing data, and then performed differential expression analysis on the filtered features to determine candidate RNA sets from SCLC patients and healthy controls. Three machine learning models were constructed using three feature selection methods: LASSO regression, random forest, and support vector machine recursive feature elimination. These three machine learning models were used to perform feature analysis on the candidate RNA set. Feature stability scores were calculated through 20 iterations, and the cross-validation consistency of features was evaluated using 10-fold nested cross-validation. The optimal combination of exosomal RNA biomarkers with 2-10 RNA combinations for small cell lung cancer was determined, and the optimal combination of exosomal RNA biomarkers included LINC00989, CXCL5, MAP3K7CL, and TUBB1.

2. The method according to claim 1, characterized in that, The criteria for technical quality filtering include: retaining RNA that is expressed in at least 80% of the samples and has a reliable expression level in at least 10% of the samples; a reliable expression level is defined as an RNA count greater than or equal to 5.

3. The method according to claim 1, characterized in that, Differential expression analysis was performed on the filtered features to identify candidate RNA sets from SCLC patients and healthy controls, including: The filtered features were normalized by log2(CPM+1), and the ComBat method was used to eliminate the batch effect introduced by RNA type. Differential expression analysis of exosomal RNA was performed on the characteristics after eliminating batch effects, and expression differences were evaluated by Wald test. A multiple screening criterion was used to identify significantly differentially expressed RNAs and determine the candidate RNA set. The multiple screening criteria included: corrected p value < 0.05, |log2 fold change| > 1.2, and average expression abundance > 50.

4. The method according to claim 1, characterized in that, The method further includes: Gene Ontology analysis was used to explore the functional enrichment of RNAs in the candidate RNA set in terms of biological processes, molecular functions and cellular components. Pathway enrichment analysis of RNAs in the candidate RNA set was performed using the KEGG database to identify molecular pathways associated with SCLC.

5. A device for screening biomarkers for the auxiliary diagnosis of small cell lung cancer, characterized in that, include: The acquisition module is used to acquire multiple sets of samples; these multiple sets of samples include exosomal RNA transcriptome sequencing data from multiple SCLC patients and multiple healthy controls; each exosomal RNA transcriptome sequencing data includes multiple RNA features. The analysis module is used to perform technical quality filtering on RNA features in all exosomal RNA transcriptome sequencing data, and to perform differential expression analysis on the filtered features to determine the candidate RNA set for SCLC patients and healthy controls. The feature selection module is used to construct three machine learning models using three feature selection methods: LASSO regression, random forest, and support vector machine recursive feature elimination. These three machine learning models are used to perform feature analysis on the candidate RNA set. The feature stability score is calculated through 20 iterations, and the cross-validation consistency of the features is evaluated through 10-fold nested cross-validation. The study investigates the optimal combination of exosomal RNA biomarkers for small cell lung cancer with 2-10 RNA combinations, and determines the optimal combination of exosomal RNA biomarkers. The optimal combination of exosomal RNA biomarkers includes LINC00989, CXCL5, MAP3K7CL, and TUBB1.

6. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 4.

7. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 4.

Images