Osteosarcoma non-invasive prognosis prediction method and kit based on ctDNA methylation characteristics

Abstract

The application provides a method and kit for non-invasive prognosis prediction of osteosarcoma based on ctDNA methylation characteristics. The method comprises: sampling and pretreating a peripheral blood sample of a subject to obtain plasma ctDNA; performing methylation detection on the plasma ctDNA to obtain methylation site information of a predetermined CpG site related to osteosarcoma prognosis; weighting and fusing the methylation site and a risk coefficient to obtain a risk score of the methylation site; weighting and fusing the prognosis risk score and a risk classification weight to obtain a probability value of each classification result, and taking the classification result with the maximum probability value as the risk assessment result. Through non-invasive liquid biopsy, the application realizes early prognosis risk stratification before treatment by using specific methylation sites in the peripheral blood ctDNA of osteosarcoma patients. The accuracy of the application reaches 73.3%, the AUC reaches 0.73, the AP reaches 0.76, and high-risk patients can be identified several months in advance, thereby providing a reliable decision basis for clinical individualized treatment.

Description

Technical Field

[0001] This application relates to the field of biological and medical testing, and in particular to a non-invasive prognostic prediction method and kit for osteosarcoma based on ctDNA methylation characteristics. Background Technology

[0002] Osteosarcoma is a rare but highly aggressive primary malignant bone tumor characterized by rapid progression and a high rate of distant metastasis. Although multimodal treatment, primarily neoadjuvant chemotherapy and extensive surgical resection combined with adjuvant chemotherapy, has improved the survival rate of patients with locally advanced lesions, maintaining a 5-year overall survival rate of 65-75%, the long-term survival rate for patients with metastatic disease or postoperative recurrence remains below 20%. This significant prognostic disparity underscores the urgent need for more precise risk stratification tools to optimize individualized treatment strategies for high-risk subgroups and improve their survival outcomes.

[0003] Currently, the prognostic assessment of osteosarcoma mainly relies on clinicopathological features such as tumor size, anatomical location, and Enneking stage, combined with imaging analysis based on MRI texture features or CT three-dimensional reconstruction.

[0004] However, these methods are fundamentally limited by the heterogeneity of tumor biology and cannot elucidate the molecular determinants driving disease progression, resulting in insufficient specificity and robustness of risk stratification. Despite advances in molecular oncology, to date, no prognostic framework integrating key driver genes or validated biomarkers has been successfully translated into clinical practice for osteosarcoma. This gap in biomarker-based risk modeling continues to hinder the achievement of precision oncology goals for this highly malignant disease.

[0005] Epigenetic dysregulation, particularly aberrant DNA methylation, is increasingly recognized as a key mechanism driving osteosarcoma development and progression, acting by mediating tumor suppressor gene silencing and oncogenic pathway activation. Genome-wide methylation studies have revealed significant epigenetic heterogeneity in osteosarcoma, with specific methylation signatures closely associated with clinical progression and prognosis. Furthermore, recent research has highlighted the potential of circulating tumor DNA (ctDNA) methylation patterns as a minimally invasive liquid biopsy tool. These epigenetic features in plasma can reflect information about tumor origin, enabling real-time assessment of disease evolution and treatment efficacy.

[0006] However, fundamental obstacles remain in bringing methylation-based predictive models to clinical practice: current methods struggle to reliably extract effective signals from complex biological noise, while advanced algorithms capable of handling complex data relationships often lack the transparency and interpretability required for clinical decision-making. Furthermore, effectively integrating multi-source testing data and ensuring its clinical consistency remains a key challenge in achieving reliable risk stratification. Summary of the Invention

[0007] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a non-invasive prognostic prediction method and kit for osteosarcoma based on ctDNA methylation characteristics, in order to solve the technical problem that the prior art cannot perform early and accurate prognostic prediction of osteosarcoma through non-invasive liquid biopsy.

[0008] To achieve the above and other related objectives, a first aspect of this application provides a non-invasive prognostic prediction method for osteosarcoma based on ctDNA methylation characteristics, comprising: sampling and preprocessing a peripheral blood sample from a subject to obtain plasma ctDNA; performing methylation detection on the plasma ctDNA to obtain methylation site information of predetermined CpG sites related to osteosarcoma prognosis; weighting and fusing the methylation sites with risk coefficients to obtain a risk score for the methylation sites; weighting and fusing the prognostic risk score with risk classification weights to obtain a probability value for each classification result, and using the classification result with the highest probability value as the risk assessment result.

[0009] In one feasible implementation, the risk coefficient is obtained by training using regression. During the training process of the risk coefficient, an L1 penalty term is introduced as a common loss. The L1 penalty term is the product of the sum of the risk coefficients of each methylation site and the penalty intensity.

[0010] In one feasible implementation, the risk classification weights are trained using a regression method. During the training process of the risk classification weights, elastic network regularization is introduced as a common loss. The elastic network regularization includes L1 regularization and L2 regularization. The L1 regularization is used to reduce some risk classification weights to 0, and the L2 regularization is used to reduce the risk classification weights.

[0011] In one feasible embodiment, the pretreatment includes: centrifuging peripheral blood to separate plasma, extracting ctDNA using a magnetic bead method, and converting the ctDNA to bisulfite.

[0012] In one feasible implementation, the methylation detection is performed using multiplex PCR-mass spectrometry or targeted methylation sequencing.

[0013] To achieve the above and other related objectives, a second aspect of this application provides a non-invasive prognostic prediction kit for osteosarcoma based on ctDNA methylation characteristics. The kit includes: a ctDNA extraction reagent for extracting ctDNA from peripheral blood plasma; a methylation detection reagent for converting the extracted ctDNA to bisulfite and detecting the methylation level of a predetermined CpG site combination; and a data analysis module for outputting a prognostic risk score and prognostic classification result for the subject based on the detected methylation level.

[0014] To achieve the above and other related objectives, a third aspect of this application provides a non-invasive prognostic prediction system for osteosarcoma based on ctDNA methylation characteristics. The system includes: a data acquisition module for acquiring methylation level data of predetermined CpG site combinations in the subject's ctDNA; a risk score calculation module with built-in coefficients of the LASSO-Cox model for calculating the subject's prognostic risk score; a prognostic classification module with built-in ENetLR classification model for outputting a prognostic category based on the risk score and methylation level; and a report generation module for outputting the prognostic risk assessment results.

[0015] To achieve the above and other related objectives, a fourth aspect of this application provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above method.

[0016] To achieve the above and other related objectives, a fifth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0017] To achieve the above and other related objectives, a sixth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0018] As described above, the non-invasive prognostic prediction method and kit for osteosarcoma based on ctDNA methylation characteristics provided in this application have the following beneficial effects:

[0019] 1) High accuracy and early warning capability: This application demonstrated prognostic accuracy in a retrospective cohort of osteosarcoma that surpassed clinical expert assessments and various traditional machine learning models. In particular, it was able to identify patients who subsequently developed radiographically confirmed metastases or recurrences months in advance, providing a valuable window for treatment intervention.

[0020] 2) Efficient feature screening: The LASSO-Cox model is used to automatically screen the combination of key CpG sites most relevant to survival prognosis from high-dimensional whole genome methylation data, effectively overcoming the "curse of dimensionality" and overfitting problems.

[0021] 3) Robust and interpretable model: The ENetLR classification model integrates L1 and L2 regularization, which can handle collinearity between features and output feature importance ranking, making the prediction process interpretable and enhancing clinicians' trust.

[0022] 4) Non-invasive and convenient: Only peripheral blood needs to be collected, avoiding invasive tissue biopsy, which facilitates pre-treatment risk assessment and dynamic monitoring.

[0023] 5) Easy to translate into clinical practice: The reagent kit is standardized, the built-in algorithm is embedded in the data analysis module, the operation process is simple, the results are objectively interpreted, and it is suitable for promotion and use in hospitals at all levels. Attached Figure Description

[0024] Figure 1 This is a schematic diagram illustrating the composition and usage process of the reagent kit provided in the embodiments of this application.

[0025] Figure 2 This is a flowchart illustrating the method provided in an embodiment of this application.

[0026] Figure 3 This is a schematic diagram of the preprocessing process in the method provided in the embodiments of this application.

[0027] Figure 4 A schematic block diagram of the system provided in the embodiments of this application.

[0028] Figure 5 A schematic block diagram of a computer device provided in an embodiment of this application.

[0029] Figure 6 The CpG site coefficients (Top20) selected for the LASSO-Cox model provided in the embodiments of this application.

[0030] Figure 7 Kaplan-Meier survival curves (high-risk vs. low-risk groups) provided for embodiments of this application.

[0031] Figure 8 ROC curve of the ENetLR model provided in the embodiments of this application.

[0032] Figure 9 The PR curve of the ENetLR model provided in the embodiments of this application. Detailed Implementation

[0033] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0034] Before providing a further detailed description of this application, the nouns and terms used in the embodiments of this application are explained, and the nouns and terms used in the embodiments of this application shall be interpreted as follows:

[0035] <1> ctDNA: Circulating tumor DNA, is fragmented DNA released into the peripheral blood by tumor cells through apoptosis, necrosis, or active secretion. It is usually 150-200 bp in length.

[0036] <2> Methylation site information: DNA methylation status at a specific CpG site on the genome, usually expressed as methylation rate (β value, range 0-1), with the value closer to 1 indicating a higher degree of methylation.

[0037] <3> The LASSO-Cox model is a survival analysis method that combines LASSO (Least Absolute Shrinkage and Selection Operator) regularization with the Cox proportional hazards model. By applying an L1 penalty to the regression coefficients, it can simultaneously screen high-dimensional features and perform survival analysis to identify the features most relevant to patient survival time.

[0038] <4> ENetLR model: Elastic Net Logistic Regression is a logistic regression model that combines L1 and L2 regularization. Its loss function includes an L1 term (to generate sparse solutions and perform feature selection) and an L2 term (to handle feature collinearity and enhance stability), making it suitable for high-dimensional classification tasks.

[0039] To address the aforementioned technical problems, this application introduces a non-invasive prognostic prediction method and kit for osteosarcoma based on ctDNA methylation characteristics. It should be understood that the method provided in this application, in addition to its application in osteosarcoma prognostic prediction, can also be applied to the prognostic assessment of other types of tumors in the medical field (such as lung cancer, breast cancer, colorectal cancer, etc.).

[0040] In one embodiment, such as Figure 1 and Figure 2 As shown, a non-invasive prognostic prediction method for osteosarcoma based on ctDNA methylation characteristics is provided, including the following steps:

[0041] Step 1) Sampling and preprocessing peripheral blood samples from the subjects to obtain plasma ctDNA.

[0042] This step included samples from 100 osteosarcoma patients, comprising 89 primary osteosarcoma tumor tissue samples and 11 paired or unpaired peripheral blood ctDNA samples. By fusing the methylation data from both types of samples, a large-sample, multi-source, high-quality methylation matrix was constructed for subsequent feature selection and model training.

[0043] Step 1-1) Collection and processing of tumor tissue samples

[0044] Sample collection: Tumor tissue samples (≥50 mg) obtained from surgical resection or puncture of 89 osteosarcoma patients were immediately frozen in liquid nitrogen.

[0045] DNA extraction: Genomic DNA was extracted using the phenol-chloroform method. DNA purity was detected by Nanodrop (A260 / A280 ratio 1.8-2.0), and DNA integrity was verified by agarose gel electrophoresis (no obvious degradation). After quantification by Qubit, the concentration was adjusted to 50 ng / μL and stored at -20℃ for later use.

[0046] Methylation detection using an Illumina Methylation EPIC 850K chip: Hybridization was detected using an Illumina Methylation EPIC 850K chip. Hybridization conditions included incubation at 48°C for 16 hours. After hybridization, the samples were washed and stained, and raw probe signal intensity data (IDAT format) were generated using an Illumina iScan scanner. This chip covers approximately 853,307 CpG sites, including key functional regions such as promoter regions, CpG islands, and enhancer regions.

[0047] Microarray data preprocessing: IDAT data was read using the R language's minfi package, filtering out low-intensity probes and samples with a detection rate <95%; crossover probes, sex chromosome probes, and probes with unclear annotations were removed; the β value of each CpG site was calculated (β = methylation signal intensity / (methylation signal intensity + non-methylation signal intensity + 100)); CpG sites with a sample missing rate >5% were removed, and the k-nearest neighbor algorithm was used to fill in the missing values, finally generating a preliminary tissue methylation matrix (rows: CpG site ID, columns: tissue sample ID, values: β value).

[0048] Steps 1-2) Collection and processing of plasma ctDNA samples

[0049] Sample collection: Peripheral blood of 8 mL was collected from 11 osteosarcoma patients (which may be a separate group of patients independent of tissue samples, or partially matched patients) before treatment and collected using EDTA anticoagulated ctDNA preservation tubes.

[0050] Plasma separation: Centrifuge at 1500g for 10 min at room temperature to separate the upper plasma layer; take the plasma layer and centrifuge again at 15000g for 10 min (4℃) to remove residual blood cells and cell debris.

[0051] ctDNA extraction: ctDNA was extracted using the magnetic bead method (Qiagen QIAamp circulating nucleic acid kit) and quantified using the Qubit dsDNA HS Assay Kit (concentration ≥10 ng / μL).

[0052] Bisulfite transformation: Transformation was performed using the EZ DNA Methylation Gold Kit (Zymo Research) under the following conditions: denaturation at 98°C for 10 min and incubation at 64°C for 2.5 h. The transformed product was then purified and recovered.

[0053] WGBS sequencing: The transformed ctDNA was used to construct a library (end repair, A-tailing, and ligation of methylated adapters), with an insert length of 200-500 bp. Paired-end sequencing (PE150) was performed using the Illumina NovaSeq 6000 platform at a sequencing depth ≥30×, generating raw sequencing reads (FASTQ format).

[0054] WGBS data preprocessing: FastP was used to remove adapters and low-quality bases; Bismark was aligned to the hg38 genome; CpG site methylation rates were extracted using Bismark_methylation_extractor; sites with coverage <10×, unclear annotations, and deletion rates >5% were removed; k-NN was used to fill in missing values, and a preliminary ctDNA methylation matrix was generated (rows: CpG site ID, columns: ctDNA sample ID, values: methylation rate).

[0055] Steps 1-3) Fusion of tissue methylation matrix and ctDNA methylation matrix

[0056] CpG site matching: Based on the hg38 genome coordinates, the GenomicRanges package in R language was used to screen for common CpG sites in the two matrices, and sites with completely consistent coordinates and clear annotation information were retained as the core sites for the integrated analysis.

[0057] Batch effect correction: The ComBat tool (sva package) was used to correct the batch effect between the chip (Illumina EPIC) and sequencing (WGBS) platforms, with "sequencing platform" as the batch variable and "sample type" as the covariate, to eliminate technical bias between platforms.

[0058] Data standardization: Tissue β values ​​and ctDNA methylation rates were scaled to the 0-1 range using the min-max normalization method to ensure that the two data types were dimensionally consistent.

[0059] Final filtering: Samples with missing values ​​and CpG sites that still exist after standardization are removed, resulting in a fused methylation matrix (rows: common CpG site IDs, columns: all qualified sample IDs, values: standardized methylation values). In this embodiment, the fused matrix contains 100 samples (89 tissues + 11 ctDNA) and 189,177 common CpG sites.

[0060] Steps 1-4) Defining prognostic labels and dividing data

[0061] Based on clinical follow-up data (overall survival, progression-free survival, and whether metastasis / recurrence occurred), patients were divided into those with good prognosis (GP, label 0) and those with poor prognosis (BP, label 1). In this cohort, approximately 44 patients were in the good prognosis group and approximately 56 patients were in the poor prognosis group.

[0062] Stratified by prognostic labels, the 100 fused samples were randomly divided into a training set (85 cases) and a test set (15 cases) in a ratio of 8.5:1.5. The test set contained 15 valid samples that passed quality control (7 cases with good prognosis and 8 cases with poor prognosis), which were used for subsequent model validation.

[0063] Step 2) Perform methylation detection on the plasma ctDNA to obtain methylation site information of predetermined CpG sites related to osteosarcoma prognosis.

[0064] Specifically, the data analysis module can be a USB flash drive, SD card, chip, or software that provides a cloud interface and has built-in algorithms. The core of this module consists of two models connected in series.

[0065] Step 3) The methylation sites are weighted and fused with the risk coefficients to obtain the risk score of the methylation sites.

[0066] The risk coefficients are obtained through regression training. During the training of the risk probabilities, an L1 penalty term is introduced as a common loss. The L1 penalty term is the product of the sum of the risk coefficients for each methylation site and the penalty strength. For example, the risk probabilities can be obtained using a LASSO-Cox model. The mathematical form of the loss function of the LASSO-Cox model is shown in Equation 1.

[0067]

[0068] in, Indicates the first True risk score for each methylation site Indicates the first Predicted risk score for each methylation site, Indicates L1 penalty term, Indicates the severity of the punishment. Indicates the first Risk coefficient of each methylation site.

[0069] For clarity, these coefficients were determined using training set data during the kit development phase and are integrated into the data analysis module during production. The risk score is a continuous value; a higher score indicates a greater prognostic risk for the patient.

[0070] Step 4) The prognostic risk score and risk classification weight are weighted and fused to obtain the probability value of each classification result. The classification result with the highest probability value is taken as the risk assessment result.

[0071] Specifically, the risk classification weights are trained using a regression method. During the training of the risk classification weights, elastic network regularization is introduced as a common loss. The elastic network regularization includes L1 regularization and L2 regularization. The L1 regularization is used to reduce some risk classification weights to 0, and the L2 regularization is used to reduce the risk classification weights.

[0072] For example, the methylation rate of the selected CpG sites can be used as the input feature vector and fed into the pre-trained ENetLR classification model. ENetLR is a binary logistic regression model, and the mathematical expression of the loss function of the ENetLR classification model is shown in Equation 2:

[0073]

[0074] in, Represented as the inherent loss for classification, This is expressed as the intensity of punishment. Represented as L1 regularization term, This is represented as the control ratio. Represented as L2 regularization term, This is represented as the risk classification weight.

[0075] Therefore, the ENetLR classification model possesses both the feature selection capability of LASSO (L1 regularization term) and the ability to penalize collinearity between features through Ridge (L2 regularization term), thus improving prediction stability. Finally, if the probability of a poor prognosis is >0.5, the subject is classified as high-risk (poor prognosis); otherwise, as low-risk (good prognosis). The probability value is output as the confidence level.

[0076] Step 5) Output the prognostic risk assessment report.

[0077] The report generation module automatically generates electronic or paper reports containing the following information: basic subject information (name, age, sample number), methylation level detection values ​​of each key CpG site, calculated risk score, prognostic classification result (good / poor), probability of poor prognosis (confidence level), and test quality control information (such as internal reference Ct value, conversion efficiency, etc.).

[0078] Example: Construction and validation of a non-invasive prognostic prediction kit for osteosarcoma based on ctDNA methylation characteristics

[0079] 1. Research cohorts and samples

[0080] This study included 100 patients with pathologically confirmed primary osteosarcoma. Tumor tissue samples were obtained from surgical resection or biopsy in 89 of these patients, and peripheral blood samples were collected from 11 patients before treatment. Clinicopathological data and long-term follow-up data, including overall survival (OS) and final outcome (death or osteosarcoma-related events were defined as poor prognosis), were collected. After stratification by prognostic outcome (good / poor), the samples were randomly assigned to a training set (85 cases) and a test set (15 cases) in an 8.5:1.5 ratio. Due to some samples failing quality control during testing, the final effective test set included 15 cases, with 7 cases in the good prognosis group and 8 cases in the poor prognosis group.

[0081] 2. ctDNA extraction, whole-genome methylation sequencing, and data preprocessing

[0082] (The method is the same as before; repeated descriptions are omitted here, and only the core steps are retained.)

[0083] The final methylation matrix (sample × CpG site) containing 189,177 common CpG sites was obtained, and the methylation values ​​were represented by β values ​​(0-1).

[0084] 3. LASSO-Cox Feature Screening and Risk Scoring Model Establishment

[0085] On the training set (85 cases), a 5-fold cross-validation was performed using the LASSO-Cox model (CoxnetSurvivalAnalysis, l1_ratio=1.0) to select the optimal penalty coefficient α with maximizing the C-index as the optimization objective. Finally, 36 CpG sites with non-zero coefficients were selected, and their coefficient distribution is shown below. Figure 6 As shown in the table below, all CpG sites and their coefficients are listed:

[0086] cg12258093 1.926933 Risk factors (positively correlated) cg26993500 1.40841 Risk factors cg25886479 1.206653 Risk factors cg25285794 1.146615 Risk factors cg15166492 0.857445 Risk factors cg23687466 0.832839 Risk factors cg12715266 0.663243 Risk factors cg00862376 0.643374 Risk factors cg01168339 0.600097 Risk factors cg02090742 0.569204 Risk factors cg16863795 0.431265 Risk factors cg13782932 0.365715 Risk factors cg02054513 0.360217 Risk factors cg09531376 0.338678 Risk factors cg13300301 0.313502 Risk factors cg17733616 0.258339 Risk factors cg18303423 0.191126 Risk factors cg06747432 0.159767 Risk factors cg07209034 0.155213 Risk factors cg10241675 0.106164 Risk factors cg03233624 0.076466 Risk factors cg13707623 0.069793 Risk factors cg09936190 0.061074 Risk factors cg01168962 0.054708 Risk factors cg11931731 0.042304 Risk factors cg25716623 0.021646 Risk factors cg11817999 0.0182 Risk factors cg10177238 -0.03283 Protective factors (negative correlation) cg23947326 -0.12412 Protective factors cg24125901 -0.12463 Protective factors cg03352106 -0.20497 Protective factors cg21706229 -0.7254 Protective factors cg22327646 -0.83783 Protective factors cg24319171 -0.94767 Protective factors cg22311200 -1.12818 Protective factors

[0087] Based on the above coefficients, the risk score calculation formula is as follows:

[0088] RiskScore = Σ(β i × Methyl i ), of which Methyl i β represents the methylation rate (β value) of the i-th CpG site. i This represents the corresponding LASSO-Cox regression coefficient. This formula is built into the kit's data analysis module and is used to calculate risk scores for new samples.

[0089] 4. Training the ENetLR classification model

[0090] The ENetLR classification model was trained using the risk score of each sample in the training set and the methylation values ​​of the aforementioned 36 CpG sites as input features (38 dimensions in total), and the prognostic label (0 = good prognosis, 1 = poor prognosis) as output. The optimal hyperparameters were determined using GridSearchCV combined with 5-fold cross-validation and ROC-AUC as the optimization metric: C = 1.0, L1_ratio = 0.7. The model was then saved and embedded in the data analysis module of the kit.

[0091] 5. Test set validation results

[0092] like Figure 7 As shown, the model's prognostic performance was evaluated on an independent test set (15 cases). The test set sample IDs, ground truth labels, predicted probabilities, and predicted labels are shown in the table below:

[0093] OS-101 0 0.4782 0 OS-002 0 0.4864 0 OS-073 0 0.5128 1 OS-107 1 0.5622 1 OS-039 0 0.4802 0 OS-001 1 0.5358 1 OS-051 1 0.5616 1 ZYC 0 0.3131 0 XZT 0 0.3083 0 OS-057 1 0.4656 0 OS-074 1 0.5463 1 OS-027 0 0.5668 1 OS-050 1 0.3649 0 OS-045 1 0.6309 1 OS-087 1 0.5480 1

[0094] 5.1 Survival Differentiation Ability (LASSO-Cox Risk Score)

[0095] Based on the calculated median risk score (median = 0.65), the 15 patients in the test set were divided into a high-risk group (n=7) and a low-risk group (n=8). Kaplan-Meier survival curves showed that the survival probability of patients in the low-risk group remained at a high level throughout the follow-up period (still >0.80 at 2500 days), while the survival probability of patients in the high-risk group decreased to 0.33 at approximately 500 days and then remained at around 0.18. The log-rank test showed that the difference in overall survival between the two groups was statistically significant (p=0.008). The C-index of the Cox risk score model on the test set was 0.72.

[0096] The Cox risk score uses the median as the classification threshold, and its confusion matrix is ​​as follows:

[0097] The actual prognosis is good. 6 1 Poor true prognosis 4 4

[0098] Accuracy = (6+4) / 15 = 66.7%, Sensitivity = 4 / 8 = 50.0%, Specificity = 6 / 7 = 85.7%.

[0099] 5.2 Binary Classification Performance (ENetLR Classification Model)

[0100] The confusion matrix obtained by the ENetLR classification model on the test set is as follows:

[0101] The actual prognosis is good. 5 2 Poor true prognosis 2 6

[0102] The calculation yielded:

[0103] Accuracy = (5+6) / 15 = 11 / 15 = 73.33%

[0104] Sensitivity (Recall) = 6 / 8 = 75.0%

[0105] Specificity = 5 / 7 = 71.43%

[0106] F1 score = 2 × (0.75 × 0.75) / (0.75 + 0.75) = 0.75

[0107] like Figure 8 As shown, the area under the curve (AUC) is calculated to be 0.73.

[0108] like Figure 9 As shown, the precision-recall (PR) curve data is as follows (some key points):

[0109] The area under the curve (AP) is 0.76, indicating that the model still has excellent classification performance under the condition of imbalance between positive and negative samples.

[0110] 5.3 Feature Importance Analysis

[0111] The ENetLR model's feature importance ranking (Top 20) shows that the risk score and the methylation levels of CpG sites such as cg12258093, cg26993500, and cg09000583 contribute the most to prognostic classification, consistent with the LASSO-Cox screening results, thus verifying the robustness of these sites as biomarkers.

[0112] 5.4 Comparison with traditional methods

[0113] On the same test set, three senior orthopedic oncology experts conducted prognostic assessments based on clinicopathological features (tumor size, Enneking stage, chemotherapy response, etc.), achieving an average accuracy of only 46.7% (7 / 15), significantly lower than the 73.3% accuracy of the ENetLR model in this application's kit (paired t-test, p=0.012). Furthermore, compared to traditional machine learning models such as Support Vector Machine (SVM), Random Forest (RF), and k-Nearest Neighbors (KNN), this application's model demonstrated the best performance in terms of AUC and accuracy, as shown in the table below.

[0114] Table 1. Performance comparison between the model in this application and traditional methods (test set n=15)

[0115] SVM 0.47 0.48 0.63 0.29 0.56 Random Forest 0.47 0.52 0.50 0.43 0.50 XGboost 0.40 0.38 0.50 0.29 0.47 MLP 0.53 0.57 0.63 0.43 0.59 This application method 0.73 0.73 0.75 0.71 0.75

[0116] 6. Examples of Clinical Use

[0117] A 14-year-old female patient presented with osteosarcoma of the distal right femur. Initial chest CT scan showed no metastasis. Peripheral blood was collected and tested using the kit described in this application. Among the 36 CpG sites tested, risk-related sites (e.g., cg12258093) showed elevated methylation rates (0.82), while protection-related sites (e.g., cg09000583) showed low methylation rates (0.15). The data analysis module calculated a risk score of 1.52 (above the built-in threshold of 0.65), and the ENetLR model output a poor prognosis probability of 0.94. The report indicated "High risk (poor prognosis), recommending enhanced follow-up and consideration of intensive adjuvant therapy." The clinician adjusted the treatment strategy accordingly. Twelve months post-surgery, although no radiographically visible metastasis was observed, the patient remained under high-risk management, and her condition remained stable. This case validates the kit's early warning capability before treatment.

[0118] 7. Summary of Model Interpretability

[0119] The decision-making process of this model can be explained based on biological and statistical principles: 1) All 36 CpG sites selected by LASSO-Cox are located in the functional regions of known osteosarcoma-related genes; 2) The sign of the regression coefficient directly indicates the direction of the risk / protective effect of each methylation site; 3) The importance ranking of ENetLR features quantifies the contribution of each biomarker to the final prediction; 4) Kaplan-Meier curves and log-rank tests validate the effectiveness of risk stratification from a survival perspective. These characteristics enable clinicians to understand and trust the model's predictive results, promoting its clinical adoption.

[0120] Since the model training process, inference process, and the entire implementation process have been described in detail in the above embodiments, they will not be repeated here.

[0121] It should also be noted that in the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect.

[0122] It should be noted that, in the embodiments of this application, the words "exemplary" or "for example" indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0123] Figure 4 This is a schematic block diagram of an osteosarcoma prognostic risk stratification system based on DNA methylation characteristics provided in an embodiment of this application. Figure 4 As shown, the system includes: a data acquisition module, a risk score calculation module, a prognosis classification module, and a report generation module.

[0124] It should be noted that this application used a fusion dataset of osteosarcoma tissue DNA methylation data (89 cases) and plasma ctDNA methylation data (11 cases) during the model training and feature screening stages to expand the sample size and improve the robustness of feature screening. The 36 prognostic-related CpG sites ultimately selected were independently verified to have highly consistent methylation levels in tissue and ctDNA, and therefore can be directly applied to ctDNA-based kit detection. In actual use, the kit only requires peripheral blood collection from patients to extract ctDNA; tissue samples are not required.

[0125] It should be understood that the specific processes by which each module performs the corresponding steps described above have been detailed in the above method embodiments, and will not be repeated here for the sake of brevity. It should also be understood that the module division in the embodiments of this application is illustrative and merely a logical functional division; other division methods may exist in actual implementation. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor, exist as separate physical entities, or have two or more modules integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0126] Figure 5 This is a schematic block diagram of a computer device provided in an embodiment of this application. Figure 5 As shown, the computer device includes at least one processor 601, a memory 602, at least one network interface 603, and a user interface 605. The various components in the device are coupled together via a bus system 604. It is understood that the bus system 604 is used to implement communication between these components. In addition to a data bus, the bus system 604 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 5 The general will label all buses as bus systems.

[0127] The user interface 605 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0128] It is understood that memory 602 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable categories of memory.

[0129] In this embodiment, the memory 602 is used to store various types of data to support the operation of the electronic terminal 600. Examples of this data include any executable program that operates on the electronic terminal 600, such as the operating system 6021 and application programs 6022. The operating system 6021 contains various system programs, such as the framework layer, core library layer, and driver layer, for implementing various basic services and handling hardware-based tasks. The application program 6022 may contain various applications, such as a media player and a browser, for implementing various application services. The method for prognostic risk stratification of osteosarcoma based on DNA methylation characteristics provided in this embodiment can be included in the application program 6022.

[0130] The methods disclosed in the embodiments of this application can be applied to or implemented by processor 601. Processor 601 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 601 or by instructions in software form. The processor 601 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 601 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. General-purpose processor 601 may be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiments of this application can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0131] In an exemplary embodiment, the electronic terminal 600 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to perform the aforementioned method.

[0132] According to the method provided in the embodiments of this application, this application also provides a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute... Figures 1 to 3 The illustrated embodiments represent any of the osteosarcoma prognostic risk stratification methods based on DNA methylation characteristics.

[0133] According to the method provided in the embodiments of this application, this application also provides a computer-readable storage medium storing program code, which, when run on a computer, causes the computer to perform the above-described method.

[0134] As used in this specification, the terms "component," "module," "system," etc., are used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. As illustrated, applications running on computing devices and computing devices can both be components. One or more components may reside in a process and / or an execution thread, and components may be located on a single computer and / or distributed among two or more computers. Furthermore, these components can be executed from various computer-readable media on which various data structures are stored. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0135] Those skilled in the art will recognize that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0136] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0137] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0138] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0139] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0140] In the above embodiments, the functions of each functional unit can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. A computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs, DVDs), or semiconductor media (e.g., solid-state disks, SSDs, etc.).

[0141] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0143] In summary, this application provides a non-invasive prognostic prediction method and kit for osteosarcoma based on ctDNA methylation characteristics. The technical solution of this application not only overcomes the limitations of single data sets but also establishes a multi-dimensional, precise diagnostic model through deep integration of in situ tumor tissue and peripheral blood ctDNA methylation data. Furthermore, it demonstrates prognostic accuracy surpassing clinical expert assessments and various traditional machine learning models in a retrospective cohort of osteosarcoma cases. In particular, it can identify patients who subsequently develop radiographically confirmed metastases or recurrences months in advance, providing a valuable window for treatment intervention. Therefore, this application effectively overcomes the various shortcomings of existing technologies and possesses high industrial applicability.

[0144] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A non-invasive prognostic prediction method for osteosarcoma based on ctDNA methylation characteristics, characterized in that, include: Peripheral blood samples from the subjects were collected and preprocessed to obtain plasma ctDNA; The plasma ctDNA was subjected to methylation detection to obtain methylation site information of predetermined CpG sites associated with osteosarcoma prognosis; The methylation sites are weighted and fused with risk coefficients to obtain a risk score for the methylation sites; The prognostic risk score is weighted and fused with the risk classification weight to obtain the probability value of each classification result. The classification result with the highest probability value is taken as the risk assessment result.

2. The method according to claim 1, characterized in that, The risk coefficients are obtained by training using regression. During the training process of the risk coefficients, an L1 penalty term is introduced as a common loss. The L1 penalty term is the product of the sum of the risk coefficients of each methylation site and the penalty intensity.

3. The method according to claim 1, characterized in that, The risk classification weights are trained using a regression method. During the training process of the risk classification weights, elastic network regularization is introduced as a common loss. The elastic network regularization includes L1 regularization and L2 regularization. The L1 regularization is used to reduce some risk classification weights to 0, and the L2 regularization is used to reduce the risk classification weights.

4. The method according to claim 1, characterized in that, The pretreatment includes: centrifuging peripheral blood to separate plasma, extracting ctDNA using magnetic beads, and converting the ctDNA to bisulfite.

5. The method according to claim 1, characterized in that, The methylation detection was performed using multiplex PCR-mass spectrometry or targeted methylation sequencing.

6. A non-invasive prognostic prediction kit for osteosarcoma based on ctDNA methylation characteristics, characterized in that, The kit includes: a ctDNA extraction reagent for extracting ctDNA from peripheral blood plasma; a methylation detection reagent for converting the extracted ctDNA to bisulfite and detecting the methylation level of a predetermined CpG site combination; and a data analysis module for outputting a prognostic risk score and prognostic classification result for the subject based on the detected methylation level.

7. A non-invasive prognostic prediction system for osteosarcoma based on ctDNA methylation characteristics, characterized in that, The system includes: a data acquisition module for acquiring methylation level data of predetermined CpG site combinations in the subject's ctDNA; a risk score calculation module, which has built-in coefficients of the LASSO-Cox model for calculating the subject's prognostic risk score; a prognostic classification module, which has built-in the ENetLR classification model for outputting the prognostic category based on the risk score and methylation level; and a report generation module for outputting the prognostic risk assessment results.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images