Osteoarthritis AI prediction method based on NETs related key genes

Abstract

The invention relates to an osteoarthritis AI (osteoarthritis) prediction method based on NETs (neutrophile granulocyte extracellular traps) related key genes, and aims to develop a model universally applicable to gene diagnosis or molecular diagnosis by analyzing and disclosing a large sample data set through an advanced algorithm by utilizing ML (makeup language) to reveal genetic regulation of neutrophile granulocyte extracellular traps (NETs) in joint tissues. Therefore, the optimal algorithm needs to be compared and screened by using multiple algorithms during modeling, the gene (molecule) diagnosis efficiency is improved by combining the algorithms on the basis, key genes which can best reflect OA characteristics and are driven by NETs (key gene characteristics related to the NETs) are selected, and the disease-related genes and the relationship between the disease-related genes and disease phenotypes are identified, so that the disease-related genes are identified. Therefore, understanding of interaction of biological components is deepened, and diagnosis and treatment strategies are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent diagnosis and treatment technology, and in particular to an AI prediction method, system, and electronic device for osteoarthritis based on NETs-related key genes. Background Art

[0002] Osteoarthritis (OA) is a common chronic joint disease. Its incidence is increasing annually with the aging population, making it a significant global public health concern. Recent discoveries have revealed that neutrophils play a key role in the pathogenesis of OA. Neutrophils are the most abundant bone marrow leukocytes in humans, accounting for 50–70% of all circulating leukocytes. They play a crucial role in the innate immune response, with effector mechanisms including phagocytosis, degranulation, and the formation of neutrophil extracellular traps (NETs). NETs are a network-like ultrastructure composed of a nucleus and granular contents, including chromatin, DNA fibers, and granule proteins. NETs protect the host from infection by trapping and killing microorganisms. However, disordered NETs can also damage surrounding tissues through autoreactivity or by increasing inflammatory responses, impairing immune system development and leading to organ and tissue dysfunction. Neutrophils modulate inflammatory responses by releasing cytokines, chemokines, and extracellular traps (NETs), processes that are particularly prominent in the pathology of OA. Chronic low-grade inflammation is a core factor in the progression of OA. A large number of inflammatory cells and factors (such as IL-6 and TNF-α) in the joint fluid promote the recruitment and activation of neutrophils, exacerbating joint damage and pain.

[0003] Furthermore, functional heterogeneity of neutrophils is evident in OA, with different subsets exhibiting distinct functions in the inflammatory response, potentially influencing disease progression and clinical manifestations. Understanding the mechanisms of action of neutrophils in OA not only helps to shed light on disease pathogenesis but also provides important clues for developing therapeutic strategies targeting neutrophils, such as inhibiting their activation or regulating their function.

[0004] As an important tool for uncovering disease mechanisms, machine learning (ML) has been widely used in the biomedical field, showing great potential in analyzing complex data sets and identifying patterns that are difficult to detect using traditional methods. ML has demonstrated outstanding performance in predicting disease progression, treatment, and drug discovery. By analyzing clinical data to identify biomarkers, it supports personalized treatment strategies. Combined with traditional pharmacological methods, it accelerates the identification of therapeutic targets and drug design optimization, reducing development costs. However, there are no technical solutions in traditional clinical practice that utilize ML to reveal the genetic regulation of neutrophil extracellular traps (NETs) in joint tissues and to help identify disease-related genes and their relationship to disease phenotypes, thus requiring further development. Summary of the Invention

[0005] In order to solve the above problems, this application proposes an AI prediction method, system and electronic device for osteoarthritis based on NETs-related key genes.

[0006] On the one hand, this application proposes an AI prediction method for osteoarthritis based on NETs-related key genes, comprising the following steps: Collect the condition data of OA patients; Importing the disease data into a pre-deployed OA prediction model, identifying OA features in the disease data through the OA prediction model and outputting corresponding NETs-related key genes; According to the OA characteristics and the NETs-related key genes, the corresponding osteoarthritis degree is identified, and a corresponding osteoarthritis prediction report is generated.

[0007] As an optional embodiment of the present application, optionally, the method for constructing the OA prediction model includes: The mRNA expression sequencing dataset of historical OA knee tissues was collected and preprocessed to obtain the internal feature set; The differentially expressed genes (DEGs) in the joint tissues of the control group and the OA group in the internal feature set were analyzed, and the obtained genes were compared according to the criteria of p value < 0.05 and |log2FC| > 0.5 to obtain the DEGs of the internal feature set; The obtained DEGs were intersected with the pre-configured neutrophil extracellular trap (NETs)-related key genes, and the intersection was used to obtain the NETs-related key genes; Based on a multimodal machine learning algorithm, the OA characteristics of the NETs-related key genes in different degrees of osteoarthritis were analyzed, and the GSE51588 feature set consisting of the NETs-related key genes and their corresponding OA characteristics was formed; Inputting the GSE51588 feature set into a preset machine learning model, performing corresponding feature training on the machine learning model, and generating the initial OA prediction model; Prepare an external validation set to verify the prediction performance of the initial OA prediction model: If the verification is successful, the OA prediction model is deployed and applied; Otherwise, the above steps are executed again to rebuild the OA prediction model.

[0008] As an optional embodiment of the present application, optionally, the mRNA expression sequencing dataset of historical OA knee joint tissues is collected and preprocessed to obtain an internal feature set, including: Download the following mRNA expression sequencing datasets related to sarcopenia from the GEO database: GSE25941, GSE38718, and GSE9103 gene chip data; The GSE25941, GSE38718, and GSE9103 gene chip data sets were merged and batch effects were removed to obtain the internal feature set: GSE51588.

[0009] As an optional embodiment of the present application, optionally, the external validation set is one or more of the following datasets: GSE19060, GSE46750, GSE57218, or GSE98918 datasets.

[0010] As an optional implementation scheme of the present application, optionally, the machine learning model is a combination model of glmBoost and Enet[alpha=0.8].

[0011] As an optional implementation scheme of the present application, optionally, the method for selecting the machine learning model includes: Several machine learning algorithms were fitted by leave-one-out cross-validation (LOOCV), and the average AUC value of each algorithm on the GSE51588 feature set and the GSE19060, GSE46750, GSE57218, and GSE98918 external validation sets was calculated. The machine learning algorithm with the largest average AUC value was selected as the algorithm model used for model training.

[0012] In another aspect, the present application proposes an AI prediction system for osteoarthritis based on NETs-related key genes, which is used to implement the above-mentioned AI prediction method for osteoarthritis based on NETs-related key genes, comprising: Medical record unit, used to collect the condition data of OA patients; An OA prediction unit, configured to import the disease data into a pre-deployed OA prediction model, identify OA features in the disease data through the OA prediction model, and output corresponding NETs-related key genes; A report generating unit is used to identify the corresponding osteoarthritis degree according to the OA characteristics and the NETs-related key genes, and generate a corresponding osteoarthritis prediction report.

[0013] In another aspect, the present application further provides an electronic device, comprising: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the osteoarthritis AI prediction method based on NETs-related key genes when executing the executable instructions.

[0014] Technical effects of the present invention: This application uses ML to analyze public large-sample data sets through advanced algorithms to reveal the genetic regulation of neutrophil extracellular traps (NETs) in joint tissues, aiming to develop a model that is generally applicable to genetic diagnosis or molecular diagnosis. Therefore, it is necessary not only to compare and screen the optimal algorithm using multiple algorithms during modeling, but also to combine algorithms on this basis to improve the efficacy of genetic (molecular) diagnosis, select key genes that drive NETs (key gene characteristics related to NETs) that best reflect OA characteristics, and help identify disease-related genes and their relationship with disease phenotypes, thereby deepening the understanding of the interaction between biological components and improving diagnostic and treatment strategies.

[0015] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0017] Figure 1-4 Shown is a diagram of data processing and differential gene identification according to the present invention. Figure 1 For the cross validation of GSE51588 and NETs, Figure 2 This is the gradient volcano map related to NETs. Figure 3 is the normalized box plot of the data set, Figure 4 This is a gene expression heat map from the GSE51588 dataset; Figure 5-6 Shown is the GO, KEGG, GSEA functional enrichment analysis data of the present invention, Figure 5 This is the GO and KEGG enrichment analysis diagram. Figure 6 NETs-based GSEA enrichment analysis; Figure 7 Shown is a schematic diagram of the distribution of AUC values ​​of several algorithm models of the present invention for a data set; Figure 8 Shown is a schematic diagram of immune infiltration analysis of high and low groups of the present invention; Figure 9 It is shown as a diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0018] Various exemplary embodiments, features, and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0019] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0020] In addition, numerous specific details are provided in the following detailed description to better illustrate the present disclosure. Those skilled in the art will appreciate that the present disclosure can be practiced without certain specific details. In some instances, methods, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main points of the present disclosure. Example 1

[0021] This invention aims to use ML to analyze public large-sample data sets through advanced algorithms to reveal the genetic regulation of neutrophil extracellular traps (NETs) in joint tissues, and aims to develop a model that is universally applicable to genetic diagnosis or molecular diagnosis. Therefore, it is necessary not only to use multiple algorithms to compare and screen the optimal algorithm during modeling, but also to combine algorithms on this basis to improve the efficacy of genetic (molecular) diagnosis, select key genes that best reflect OA characteristics of NETs drivers (OA characteristics of key genes related to NETs, ​​i.e., osteoarthritis characteristics), and help identify disease-related genes and their relationship with disease phenotypes, thereby deepening the understanding of the interaction between biological components and improving diagnostic and treatment strategies.

[0022] The characteristic data of osteoarthritis patients mainly include the following aspects: 1. Clinical manifestations and characteristics 1. Joint pain and tenderness: Initially, it is mild or moderate, intermittent, dull pain that improves with rest and worsens with activity. The pain is often related to weather changes, and in later stages, it may become persistent or nighttime. Weight-bearing joints and the hands are most susceptible.

[0023] 2. Joint stiffness: Joint stiffness and tightness, also known as morning stiffness, occurs when you wake up in the morning or after a long rest. This feeling may ease with activity, but may worsen with lower air pressure or higher humidity. It generally lasts for a short time, typically a few minutes to more than ten minutes, rarely exceeding 30 minutes.

[0024] 3. Joint swelling: This is caused by inflammatory factors stimulating the synovial membrane to thicken and secrete large amounts of fluid. Swelling is usually concentrated around the joints, especially after activity or during changes in weather. Hand joints can be swollen and deformed, and some knee joints can also swell due to the formation of osteophytes.

[0025] 4. Crepitus: Crepitus or a sensation of bone rubbing may occur during joint movement. This is caused by damage to the articular cartilage and uneven joint surfaces. When the articular cartilage is severely worn, friction between the joint surfaces increases, resulting in a creaking sound or a rubbing sensation during joint movement.

[0026] 5. Joint weakness and movement disorders: Symptoms of joint weakness and movement disorders may appear in the late stage. Patients may feel weak legs, joints cannot be straightened or locked when walking, and cannot be fully straightened or perform normal joint movements. It may also cause accidents such as falls.

[0027] The above-mentioned symptom characteristics can be analyzed and identified using technologies such as NLP, and statistics can also be used to count the symptom characteristics of various degrees.

[0028] 2. Imaging Characteristics The X-ray features of osteoarthritis include asymmetric joint space narrowing, subchondral bone sclerosis and cystic changes, bone hyperplasia and osteophyte formation at the joint margins, intra-articular loose bodies, joint deformity, and subluxation. These changes are important for the diagnosis of osteoarthritis.

[0029] These features of different levels can be used to extract imaging features using CNN.

[0030] 3. Laboratory Examination Blood count, protein electrophoresis, immune complexes, and serum complement levels are generally within normal ranges. Patients with synovitis may have mild elevations in C-reactive protein and erythrocyte sedimentation rate. Rheumatoid factor and antinuclear antibodies are negative.

[0031] Data features can be entered directly.

[0032] When implementing this method, please combine the examination of the patient's condition and genetic examination data with clinical examination methods, and no limitation or elaboration will be made here.

[0033] The choice of gene database, such as the GEO database, is up to the user.

[0034] The implementation principle of the present invention will be described in detail below.

[0035] like Figure 9 As shown, this application, on the one hand, proposes an AI prediction method for osteoarthritis based on NETs-related key genes, comprising the following steps: S1. Collect the condition data of OA patients; S2. Importing the disease data into a pre-deployed OA prediction model, identifying OA features in the disease data through the OA prediction model and outputting corresponding NETs-related key genes; S3. According to the OA characteristics and the NETs-related key genes, the corresponding osteoarthritis degree is identified, and a corresponding osteoarthritis prediction report is generated.

[0036] The hospital's HIS system can record the condition of OA patients and their examination data. After the examination, the data can be input into the OA prediction model deployed in the background. The OA prediction model will identify the OA characteristics in the patient's condition data and automatically determine whether there are corresponding NETs-related key genes. In this way, based on a large sample data set, the genetic regulation of neutrophil extracellular traps (NETs) in joint tissues is revealed, and help identify disease-related genes and their relationship with disease phenotypes, thereby deepening the understanding of the interaction between biological components and improving diagnosis and treatment strategies.

[0037] Doctors can combine the OA characteristics of the patient's condition and output the corresponding NETs-related key genes, flexibly adjusting the corresponding diagnosis and treatment strategies.

[0038] The OA prediction model of the present invention will be described below.

[0039] As an optional embodiment of the present application, optionally, the method for constructing the OA prediction model includes: The mRNA expression sequencing dataset of historical OA knee tissues was collected and preprocessed to obtain the internal feature set; The differentially expressed genes (DEGs) in the joint tissues of the control group and the OA group in the internal feature set were analyzed, and the obtained genes were compared according to the criteria of p value < 0.05 and |log2FC| > 0.5 to obtain the DEGs of the internal feature set; The obtained DEGs were intersected with the pre-configured neutrophil extracellular trap (NETs)-related key genes, and the intersection was used to obtain the NETs-related key genes; Based on a multimodal machine learning algorithm, the OA characteristics of the NETs-related key genes in different degrees of osteoarthritis were analyzed, and the GSE51588 feature set consisting of the NETs-related key genes and their corresponding OA characteristics was formed; Inputting the GSE51588 feature set into a preset machine learning model, performing corresponding feature training on the machine learning model, and generating the initial OA prediction model; Prepare an external validation set to verify the prediction performance of the initial OA prediction model: If the verification is successful, the OA prediction model is deployed and applied; Otherwise, the above steps are executed again to rebuild the OA prediction model.

[0040] 1. Data source collection and preprocessing Historical OA knee tissue mRNA expression sequencing datasets in the form of microarray and bulk RNA-seq were obtained from the GEO database (https: / / www.ncbi.nlm.nih.gov / geo / ). Probe names were converted to gene symbols using the matching platform annotation files. All microarray datasets were log2 transformed and normalized using the “limma” R package (see Figure 3 Download the raw counts of the batch RNA-seq dataset and perform logarithmic transformation using the LimmaVoom function. Limma-voom is a type of generalized linear model that fits a linear equation Y to the expression level of each gene, uses weighted least squares to estimate the difference in gene expression, and uses the Bayesian method to correct for multiple testing problems. The formula is as follows: X is the gene expression, β is the value that limma will find ( β 0 is the system default initial value), δ is the residual assumed to be normally distributed in the entire data set: .

[0041] This method has the characteristics of high accuracy, small influence of outliers, good false positive control, and fast calculation, and is used as a powerful gold standard for RNA-seq analysis.

[0042] For implementation, we downloaded the sarcopenia-related gene chip data from GSE25941, GSE38718, and GSE9103 from the GEO database. The datasets were merged and batch effects removed to generate the internal feature set (GSE51588). To construct the internal feature set (GSE51588), we used the "limma" R package to analyze differentially expressed genes (DEGs) in joint tissues between the control and OA groups within the GSE51588 dataset using the limma-voom method. The resulting genes were compared using the criteria of p-value < 0.05 and |log2FC| > 0.5 to generate the DEGs for the feature set. Furthermore, based on previous studies, we selected 136 key genes associated with neutrophil extracellular traps (NETs). To identify common gene signatures between these DEGs and NETs, ​​we performed an intersection analysis of the datasets. Specifically, because NET deficiency is common in the progression of osteoarthritis, developing effective strategies for predicting OA based on key genes in the NET signature has become a research hotspot in the arthritis field. Due to the differential expression of NETs characteristic key genes in different degrees of osteoarthritis and the significant differences in OA occurrence stratified by 21 NETs characteristic differential genes (such as those in the attached Figure 1The goal was to construct a robust OA prediction model based on key genes characteristic of NETs. DEGs were compared with NETs, ​​and their intersection was extracted to identify key NET-related genes. This approach effectively reduced noise and focused on results with greater biological significance.

[0043] In the specific implementation, the R language "limma" package was used to screen the DEGs of sarcopenia with a threshold of P < 0.05 and |log2FC| > 0.5 for the processed data. Finally, 965 DEGs were obtained, of which 484 genes were significantly up-regulated and 481 genes were significantly down-regulated. The R language "ggplot2" package was used to draw a volcano plot, see Figure 2 .

[0044] In order to keep the data distribution consistent among the datasets, the density maps before and after removing the batch effect are obtained. Figure 7 .

[0045] The expression analysis of the internal feature set is performed, see Figure 4 Shown is the gene expression heat map of the internal feature set: GSE51588.

[0046] During model training, both a training feature set and a validation set are required. Here, the GSE51588 feature set was used as an internal training set for the machine learning algorithm to screen for optimal OA features for key NETs-related genes. The GSE19060, GSE46750, GSE57218, and GSE98918 datasets were included as external validation sets. (To ensure the widespread use of feature genes in this modeling dataset and other datasets (such as individual patients from different regions and hospitals), in addition to the modeled GSE51588, external validation sets (also OA datasets) from other regions and hospitals were also included.) These external validation sets can verify the robustness and diagnostic efficacy of the previously established model.

[0047] 2. Gene Enrichment Analysis In addition to the above pretreatment methods, gene enrichment analysis can also be performed. Gene set enrichment analysis (GSEA) can be used to compare the significant functional differences between OA and normal samples to explore the role of NETs functional defects in the progression of osteoarthritis leading to osteoarthritis.

[0048] To investigate the biological processes (BPs), cellular components (CCs), molecular functions (MFs), and potential pathways of GSE51588, gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was performed on DE-DT using the "clusterprofiler" R package. Enrichment analysis is a bioinformatics method used to identify biologically significant functions or pathways from a gene list. Its core concept is to use statistical methods to assess the enrichment of a target gene set in specific functional annotations (such as GO terms or KEGG pathways) to reveal the underlying biological functions and mechanisms of genes. Gene set enrichment analysis (GSEA) was used to compare significant functional differences between OA and normal samples. GSEA statistical tests typically use hypergeometric tests, Fisher's exact tests, or gene ranking. For example, the hypergeometric test aims to determine whether the number of genes with a given functional annotation in the target gene set is significantly higher than the expected value from a random distribution. Specifically, the number of genes in a given GO term in the background gene set is N, and the number of genes in the target gene set that belong to the same GO term is k. Enrichment significance is calculated using the hypergeometric distribution formula. In addition, to control the false positive rate, the Benjamini-Hochberg method is usually used to correct the p-value for multiple testing to obtain an adjusted p-value (q-value). Significantly enriched GO terms and KEGG pathways provide important clues for further understanding the functions of differentially expressed genes and their role in NETs characteristic genes.

[0049] GO, KEGG, GSEA functional enrichment analysis Figure 5-6 As shown, based on the gene expression profile, GSEA enrichment analysis based on NETs characteristics was performed on the dataset. GO and KEGG enrichment analysis were performed on 965 DEGs to better understand the biological processes they are involved in. GO and KEGG enrichment analysis results showed that the 965 DEGs mainly activated inflammation-related biological processes. GSEA enrichment analysis results showed that NETs functional defects in the progression of osteoarthritis led to osteoarthritis.

[0050] 3. Feature training and generation and validation of OA prediction models based on NETs-related key genes In order to identify key genes related to NETs and construct a classification prediction model for predicting NETs in OA disease, 12 classic machine learning algorithms were integrated, including random forest (RF), elastic net (Enet), Lasso regression (Lasso), ridge regression (Ridge), support vector machine (SVM), partial least squares regression generalized linear model (plsRglm), stepwise regression (Stepglm), generalized linear model boosting (glmBoost), gradient boosting machine (GBM), linear discriminant analysis (LDA), extreme gradient boosting (XGBoost) and naive Bayes (NaiveBayes).

[0051] The following steps are covered: Feature selection, such as Lasso, Enet, and Ridge, uses regularization methods to screen out features (differential genes) with significant predictive power. For OA features that are most correlated with the expression of key genes related to NETs in predicting OA disease, physicians can prepare corresponding data sets in advance to facilitate the algorithmic screening of the most characteristic gene (molecular) features, that is, the OA features that are most relevant to the corresponding key genes related to NETs. For example, the disease and genetic data of several patients can be extracted to use the algorithm to screen out the OA features that are most relevant to the patient's "key genes related to NETs"; Nonlinear modeling: such as RF, GBM, and XGBoost, which can capture complex nonlinear relationships and improve the predictive performance of the model; Classification capabilities: such as SVM, LDA, and NaiveBayes, suitable for processing high-dimensional data and building classification models; Model improvement: Various modeling strategies, such as glmBoost and GBM, are used to improve the generalization ability of the model through iterative optimization. By integrating these algorithms, we can comprehensively evaluate the performance of different modeling methods in predicting NETs-related key genes in OA disease, screen out the most characteristic gene (molecular) features, and thus generate an osteoarthritis prediction model based on NETs-related key genes to ensure the robustness and reliability of the results.

[0052] On this basis, combined with the leave-one-out cross-validation (LOOCV) framework, 106 algorithm combination models were constructed to optimize predictive performance and ensure the robustness of the results. This multi-algorithm integration and combination strategy enables the exploration of optimal model configurations to improve predictive performance, reduce the potential bias introduced by a single algorithm, ensure the stability of model results, comprehensively evaluate the performance of different modeling methods, and screen for the optimal model, providing a reliable basis for subsequent biological validation.

[0053] During the specific implementation, in the training cohort, 106 machine learning algorithm combinations were fitted by leave-one-out cross-validation (LOOCV) to construct a prediction model, and the average AUC value of each algorithm in the internal dataset and the GSE19060, GSE46750, GSE57218, and GSE98918 external validation sets was calculated.

[0054] like Figure 7 The results showed that among the algorithms with similar AUC values, the NETs-related key genes included in the glmBoost+Enet[alpha=0.8] algorithm were the most able to reflect the characteristics of OA. Therefore, the combination of glmBoost and Enet[alpha=0.8] was selected as the final model. Subsequently, based on the glmBoost+Enet[alpha=0.8] algorithm, the risk score of each sample in the training set was calculated, and all samples were divided into high-risk group and low-risk group according to the median risk score. Figure 8 As shown, immune infiltration analysis in the three algorithms showed that the abundance of neutrophils was higher in the high-risk group than in the low-risk group.

[0055] The present invention aims to develop a universally applicable model for genetic or molecular diagnosis. Therefore, modeling requires not only comparing and selecting the optimal algorithm using multiple algorithms, but also combining these algorithms to enhance the efficacy of genetic (molecular) diagnosis. Key genes driving NETs (key NET-related gene signatures) that best reflect OA characteristics are selected for modeling. Including an external validation set here verifies the robustness and predictive performance of OA characteristic models constructed using multiple machine learning algorithms, aiming for model generalization. The area under the receiver operating characteristic (ROC) curve (AUC) was calculated for each model. Among algorithms with similar AUC values, the NET-related key genes included in the glmBoost+Enet[alpha=0.8] algorithm best reflected OA characteristics. Coefficient values ​​for the key NET-related genes included in the glmBoost+Enet[alpha=0.8] algorithm were extracted, and sample risk was assessed based on expression levels and corresponding coefficient values.

[0056] Therefore, the present invention used leave-one-out cross-validation (LOOCV) to fit several machine learning algorithms. The average AUC value for each algorithm was calculated for the GSE51588 feature set and the external validation sets of GSE19060, GSE46750, GSE57218, and GSE98918. The machine learning algorithm with the highest average AUC value was selected as the algorithm model for model training. After analysis, a combination of glmBoost and Enet [alpha = 0.8] was ultimately selected as the machine learning model.

[0057] 4. Application Analysis of the Model - Immune Infiltration Analysis like Figure 8 As shown in the figure, three immune infiltration algorithms, CIBERSORT, Estimate, and Quanti-seq, were used to calculate the levels of immune cell infiltration in high-risk and low-risk groups based on key NET-related gene signatures. Furthermore, the "ggplot2" R package was used to generate heat maps illustrating the relationship between genes and immune cell infiltration.

[0058] The following describes the principles of each algorithm: 1) Data preparation Obtain gene expression data: Ensure that expression data for key genes related to NETs (neutrophil extracellular traps) have been obtained. For details, see the relevant RNA-Seq or microarray experiments mentioned above.

[0059] Grouping information: Samples were divided into high-risk and low-risk groups based on NETs-related characteristics or clinical information.

[0060] 2) Immune infiltration algorithm uses CIBERSORT algorithm: Input: Gene expression matrix and immune cell signature matrix (LM22 signature matrix provided by CIBERSORT or other relevant signatures).

[0061] Steps: The CIBERSORT algorithm was used to calculate the relative abundance of immune cells in each sample.

[0062] Output: Immune cell infiltration ratio matrix.

[0063] Estimate Algorithm: ‌Input‌: Gene expression matrix.

[0064] Steps: Run the Estimate algorithm to calculate the immune score (Immune Score), stromal score (StromalScore), and estimated score (Estimate Score).

[0065] Output: Immune score matrix and other related scores.

[0066] ‌QuantiSeq Algorithm‌: Input: Gene expression matrix and immune cell marker gene list provided by QuantiSeq.

[0067] ‌Step‌: Calculate immune cell infiltration levels using the QuantiSeq algorithm.

[0068] Output: Immune cell infiltration ratio matrix.

[0069] 3) Data integration and analysis ‌Integrate immune infiltration results‌: Integrate the immune cell infiltration results obtained by the three algorithms into one data frame to ensure consistent sample IDs.

[0070] Group comparison: Calculate the average infiltration level of each immune cell in the high-risk group and the low-risk group, and perform statistical comparison (such as t-test).

[0071] 4) Generate heat map Data preparation: Select key immune cell and NET-related genes of interest and organize them into a data format suitable for drawing heat maps.

[0072] ‌Draw a heatmap using ggplot2‌: Install and load the ggplot2 package as follows: install.packages("ggplot2") library(ggplot2).

[0073] Data preprocessing: Ensure that the data frame contains sample IDs, immune cell / gene names, and corresponding expression / infiltration values.

[0074] ‌ Draw a heat map, the code is as follows‌: # Assume that data is a sorted data frame, rownames are immune cell / gene names, and colnames contain sample ID and group information data_long<- melt(data, id.vars = "SampleID", variable.name = "Feature", value.name = "Value") data_long$Group<- factor(data_long$SampleID) # Assume that the SampleID column contains grouping information and needs to be adjusted according to actual conditions p<- ggplot(data_long, aes(x = Feature, y = SampleID, fill = Value)) +geom_tile() + scale_fill_gradient2(low = "blue", high = "red", mid = "white", midpoint = median(data_long$Value, na.rm = TRUE), limit = c(min(data_long$Value, na.rm = TRUE), max(data_long$Value, na.rm = TRUE)), space = "Lab", name="Expression / Infiltration Level") + theme_minimal()+theme(axis.text.x = element_text(angle = 45, hjust =1)) +facet_wrap(~Group) # Draw multiple small pictures based on group information print(p).

[0075] Adjust heatmap details: Adjust color gradients, labels, titles, etc. as needed.

[0076] Interpretation and reporting of results ‌Interpretation of the heat map‌: The color depth of the heat map explains the difference in infiltration levels of different immune cells in the high-risk group and the low-risk group.

[0077] ‌Combined with NETs key genes‌: Discuss the potential relationship between the expression of NETs-related key genes and immune cell infiltration.

[0078] Report Writing: Organize the results into a report including methods, results, discussion, and conclusion sections, along with heat maps and other relevant graphics.

[0079] Obviously, those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned control embodiments. Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned control embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD). The storage medium can also include a combination of the above-mentioned types of memory. Example 2

[0080] Based on the implementation principle of Example 1, this application, on the other hand, proposes an AI prediction system for osteoarthritis based on NETs-related key genes, which is used to implement the above-mentioned AI prediction method for osteoarthritis based on NETs-related key genes, comprising: Medical record unit, used to collect the condition data of OA patients; An OA prediction unit, configured to import the disease data into a pre-deployed OA prediction model, identify OA features in the disease data through the OA prediction model, and output corresponding NETs-related key genes; A report generating unit is used to identify the corresponding osteoarthritis degree according to the OA characteristics and the NETs-related key genes, and generate a corresponding osteoarthritis prediction report.

[0081] The functions and interactions of the above-mentioned units can be understood in conjunction with the description of Example 1, and will not be described in detail here.

[0082] The modules or steps of the present invention described above can be implemented using a general-purpose computing system. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Alternatively, they can be implemented using program code executable by the computing system, which can then be stored in a storage system and executed by the computing system. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software. Example 3

[0083] Furthermore, in another aspect, the present application further provides an electronic device, comprising: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the osteoarthritis AI prediction method based on NETs-related key genes described in Example 1 when executing the executable instructions.

[0084] The electronic device of the present disclosure includes a processor and a memory for storing processor-executable instructions. The processor is configured to implement the AI-based osteoarthritis prediction method based on NETs-related key genes described in Example 1 above when executing the executable instructions.

[0085] It should be noted that the number of processors can be one or more. Furthermore, the electronic device according to the embodiments of the present disclosure may also include an input system and an output system. The processor, memory, input system, and output system may be connected via a bus or other means, which are not specifically limited herein.

[0086] Memory, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and various modules, such as the program or module corresponding to the AI-based osteoarthritis prediction method based on NETs-related key genes in the disclosed embodiments. The processor executes the software programs or modules stored in the memory to perform various functional applications and data processing in the electronic device.

[0087] The input system can be used to receive input numbers or signals. The signals can be key signals related to user settings and function control of the device / terminal / server. The output system can include display devices such as display screens.

[0088] While various embodiments of the present disclosure have been described above, the foregoing description is intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An AI prediction method for osteoarthritis based on NETs-related key genes, characterized by: The steps include: Collect the condition data of OA patients; Importing the disease data into a pre-deployed OA prediction model, identifying OA features in the disease data through the OA prediction model and outputting corresponding NETs-related key genes; According to the OA characteristics and the NETs-related key genes, the corresponding osteoarthritis degree is identified, and a corresponding osteoarthritis prediction report is generated.

2. The AI ​​prediction method for osteoarthritis based on NETs-related key genes according to claim 1, characterized in that: The method for constructing the OA prediction model includes: The mRNA expression sequencing dataset of historical OA knee tissues was collected and preprocessed to obtain the internal feature set; The differentially expressed genes (DEGs) in the joint tissues of the control group and the OA group in the internal feature set were analyzed, and the obtained genes were compared according to the criteria of p value < 0.05 and |log2FC| > 0.5 to obtain the DEGs of the internal feature set; The obtained DEGs were intersected with the pre-configured neutrophil extracellular trap (NETs)-related key genes, and the intersection was used to obtain the NETs-related key genes; Based on a multimodal machine learning algorithm, the OA characteristics of the NETs-related key genes in different degrees of osteoarthritis were analyzed, and the GSE51588 feature set consisting of the NETs-related key genes and their corresponding OA characteristics was formed; Inputting the GSE51588 feature set into a preset machine learning model, performing corresponding feature training on the machine learning model, and generating the initial OA prediction model; Prepare an external validation set to verify the prediction performance of the initial OA prediction model: If the verification is successful, the OA prediction model is deployed and applied; Otherwise, the above steps are executed again to rebuild the OA prediction model.

3. The AI ​​prediction method for osteoarthritis based on NETs-related key genes according to claim 2, characterized in that: The mRNA expression sequencing dataset of historical OA knee joint tissues is collected and preprocessed to obtain an internal feature set, including: Download the following mRNA expression sequencing datasets related to sarcopenia from the GEO database: GSE25941, GSE38718, and GSE9103 gene chip data; The GSE25941, GSE38718, and GSE9103 gene chip data sets were merged and batch effects were removed to obtain the internal feature set: GSE51588.

4. The AI ​​prediction method for osteoarthritis based on NETs-related key genes according to claim 2, characterized in that: The external validation set is one or more of the following datasets: GSE19060, GSE46750, GSE57218, or GSE98918 datasets.

5. The AI ​​prediction method for osteoarthritis based on NETs-related key genes according to claim 2, characterized in that: The machine learning model is a combination model of glmBoost and Enet[alpha=0.8].

6. The AI ​​prediction method for osteoarthritis based on NETs-related key genes according to claim 2, characterized in that: The method for selecting the machine learning model includes: Several machine learning algorithms were fitted by leave-one-out cross-validation (LOOCV), and the average AUC value of each algorithm on the GSE51588 feature set and the GSE19060, GSE46750, GSE57218, and GSE98918 external validation sets was calculated. The machine learning algorithm with the largest average AUC value was selected as the algorithm model used for model training.

7. An AI prediction system for osteoarthritis based on NETs-related key genes, used to implement the AI ​​prediction method for osteoarthritis based on NETs-related key genes according to any one of claims 1 to 6, characterized in that: include: Medical record unit, used to collect the condition data of OA patients; An OA prediction unit, configured to import the disease data into a pre-deployed OA prediction model, identify OA features in the disease data through the OA prediction model, and output corresponding NETs-related key genes; A report generating unit is used to identify the corresponding osteoarthritis degree according to the OA characteristics and the NETs-related key genes, and generate a corresponding osteoarthritis prediction report.

8. An electronic device, characterized in that include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement an AI prediction method for osteoarthritis based on NETs-related key genes as described in any one of claims 1-6 when executing the executable instructions.

Images