Plasma lipid isomer biomarker for predicting survival benefit of neoadjuvant therapy of breast cancer and application of plasma lipid isomer biomarker

Abstract

The invention relates to a plasma lipid isomer biomarker for predicting survival benefit of neoadjuvant therapy of breast cancer and application of the plasma lipid isomer biomarker. The plasma lipid isomer biomarker is PC 17: 1 (n-8) 18: 1 (n-7), PC 18: 018: 1 (n-7), PC 14: 018: 1 (n-10), PE 16: 018: 1 (n-7), PE 14: 018: 1 (n-7) and PC 16: 1 (n-7) 18: 1 (n-9). The invention also constructs a breast cancer NAT survival benefit prediction model. According to the method, the survival benefit probability of the patient can be accurately evaluated, and a scientific basis for personalized treatment decision is provided for clinic, so that a more accurate treatment scheme is realized, the treatment effect of the patient is optimized, and excessive treatment or insufficient treatment is avoided.

Description

Technical Field

[0001] The present invention relates to a plasma lipid isomer biomarker for predicting survival benefit of neoadjuvant therapy for breast cancer and application thereof, belonging to the technical field of biomedicine. Background Art

[0002] Breast cancer ranks first in both morbidity and mortality among female malignancies, representing a significant cancer burden for Chinese women. Tumor heterogeneity is a key challenge in breast cancer diagnosis and treatment, increasing the complexity and uncertainty of treatment options. Early diagnosis, efficacy prediction, and improved prognosis of breast cancer have long been key research priorities for clinical researchers.

[0003] Neoadjuvant therapy (NAT) refers to systemic drug therapy administered before surgery. It has become a crucial component of comprehensive breast cancer treatment strategies. Its advantages include downstaging the primary tumor, improving the rate of breast-conserving surgery, predicting tumor sensitivity, and assessing patient benefit. The pathological complete response (pCR) rate is a powerful indicator of NAT response and an important basis for adjusting adjuvant therapy regimens. Previous studies have shown that patients who achieve a pCR with neoadjuvant therapy have better disease-free survival (EFS) and overall survival (OS) than those with residual disease. Therefore, pCR is used as a surrogate endpoint for EFS in prospective clinical studies of neoadjuvant chemotherapy. However, 40-60% of patients still have residual disease after NAT, and some may be overtreated, experiencing treatment-related adverse reactions and mortality. Therefore, there is an urgent need to explore and develop new biomarkers that can accurately predict survival benefit in breast cancer patients. This will provide a scientific basis for personalized treatment and avoid overtreatment or undertreatment.

[0004] Survival benefit, particularly progression-free survival (PFS) and overall survival (OS), is an important clinical indicator for evaluating the efficacy of NAT. The pathological complete response (pCR) rate is often used as a marker for evaluating NAT response, but it does not fully predict a patient's long-term survival benefit. Existing prediction tools often focus on short-term responses and neglect the prediction of long-term survival benefit. Therefore, developing a new model that can accurately predict the survival benefit of NAT is of great significance for the personalized treatment and long-term management of breast cancer patients.

[0005] Lipids are important energy donors for life activities, the structural basis of biological membranes, and participate in the transmission of biological information. They play an important role in the growth, invasion, and metastasis of tumors. Lipid metabolism plays a key role in the occurrence and development of breast cancer. Lipids are not only an important component of cell membranes, but also important molecules for the energy supply and biological information transmission of cancer cells. Tumor cells meet the energy needs required for their proliferation by remodeling lipid metabolism. Changes in lipid metabolism affect tumor growth, invasion, and metastasis, and play an important role in the tumor microenvironment. In recent years, lipidomics has become an important direction in cancer research. It helps identify biomarkers related to tumor response and survival prognosis by systematically analyzing the types of lipids, metabolic pathways, and their changes in tumors.

[0006] However, lipid structures are complex, with over 40,000 distinct structures. Variations in multiple dimensions, including head groups, carbon chains, C=C positions, sn-positions, on-chain modifications, and stereochemistry, lead to a rich diversity of structural details and corresponding biological functions, which have significant impacts on life. A major limitation of current conventional lipidomics approaches is the inability to resolve detailed structural features of lipid species, such as the positional specificity of carbon-carbon double bonds (C=C) and the sn-positions of fatty acyl or alkyl chains, which are ubiquitous in biological systems. These differences play a crucial role in regulating the biological behavior of tumor cells. Analytical techniques suitable for scalable resolution of lipid molecular structure and spatial distribution are needed to enable comprehensive qualitative and quantitative analysis of lipids in cells, tissues, or organisms. Recent advances in tandem mass spectrometry (MS), including UV-light photodissociation, ozone-induced dissociation, and photochemical derivatization, have overcome this limitation, enabling high-resolution structural characterization of lipid isomers. These innovations have revealed the unique functional roles of specific lipid isomers. For example, positional isomers of the C=C bond have been shown to affect key biophysical properties such as membrane permeability, thickness, and cholesterol-phospholipid interactions, all of which influence cellular flexibility and may contribute to cancer cell adaptability. Despite these advances, the diagnostic potential of lipid isomers remains largely untapped due to the lack of robust analytical workflows and scalable large-scale biomarker validation tools. Developing comprehensive platforms to identify and characterize lipid isomer biomarkers is crucial to advance their application in precision diagnosis and treatment of breast cancer.

[0007] Despite this, the application of lipid isomers in predicting survival benefit in breast cancer is still in its infancy. Existing technologies primarily focus on lipid profiling in tumor tissue, and most studies have focused on predicting short-term treatment responses. The role of lipid isomers in long-term survival benefit has not been fully explored. Therefore, developing a high-precision lipidomics-based model for predicting survival benefit from neoadjuvant therapy for breast cancer will provide a more comprehensive and effective basis for personalized breast cancer treatment. Summary of the Invention

[0008] In response to the above-mentioned problems, the present invention provides a plasma lipid isomer biomarker for predicting the survival benefit of neoadjuvant therapy (NAT) for breast cancer and its application. The lipid isomer biomarker of the present invention is a lipid fine structure metabolite in the plasma of breast cancer patients. It is extracted and identified with high precision by high-throughput mass spectrometry technology to screen out lipid features that are closely related to the survival benefit of neoadjuvant therapy for breast cancer. At the same time, the present invention uses high-precision lipidomics technology to analyze the lipid isomer features in the plasma of breast cancer patients, and combines the machine learning algorithm XGBoost to construct a model for predicting the survival benefit of breast cancer patients after receiving neoadjuvant therapy. This model can accurately assess the probability of survival benefit of patients and provide a scientific basis for personalized treatment decision-making in the clinic, thereby achieving more accurate treatment plans, optimizing patient treatment effects, and avoiding overtreatment or undertreatment. The technical solution of the present invention is as follows: A plasma lipid isomer biomarker for predicting survival benefit from neoadjuvant therapy for breast cancer, wherein the plasma lipid isomer biomarker is as follows: PC 17:1 (n-8) _18:1(n-7), PC 18:0_18:1(n-7), PC 14:0_18:1(n-10), PE 16:0_18:1(n-7), PE 14:0_18:1(n-7) and PC 16:1 (n-7) _18:1(n-9).

[0009] Furthermore, the present invention also provides a model for predicting the survival benefit of breast cancer patients after receiving neoadjuvant therapy, wherein the modeling algorithm of the model is the Extreme Gradient Boosting Regression (XGBoost) model, and the plasma lipid isomer biomarker as described in claim 1 is used to predict the survival benefit of breast cancer neoadjuvant therapy.

[0010] Furthermore, the model construction method comprises the following steps: (1) All datasets are randomly divided into training sets and test sets to ensure the generalization ability and stability of the model; the training set is used for model training and hyperparameter tuning, and the test set is used to evaluate the final performance of the model; Preferably, 70% of the data is used as the training set and the remaining 30% of the data is used as the test set.

[0011] (2) Optimize the hyperparameters of the XGBoost model using grid search and 5-fold cross-validation to determine the optimal hyperparameter configurations such as the number of trees, learning rate, and maximum depth; (3) Evaluate the final model using the test set; (4) The characteristic values ​​of the patient's six plasma lipid isomer biomarkers are substituted into the model to obtain a risk score. If the risk score is greater than the critical value, the patient is judged to be at high risk; if the risk score is lower than the critical value, the patient is judged to be at low risk.

[0012] Furthermore, the present invention also provides a risk scoring system based on the above model, wherein the critical value of the risk score is 0.720.

[0013] Compared with the prior art, the present invention has the following advantages: 1. This invention overcomes the challenges of tumor heterogeneity and the difficulty of obtaining biopsy samples. It provides a novel liquid-based biopsy, namely, stable lipid fine structure metabolism biomarkers in breast cancer NAT plasma. It also constructs a model for predicting the survival benefit of breast cancer patients after neoadjuvant therapy.

[0014] 2. The prediction model for NAT survival benefit in breast cancer constructed by the present invention has AUC values ​​of 0.90, 0.92 and 0.96 for 1-year, 3-year and 5-year progression-free survival benefit, respectively.

[0015] 3. The optimal critical value of 0.720 obtained from the ROC curve can well distinguish the recurrence probability between the high-risk group and the low-risk group (P<0.001). BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Forest plots showing the hazard ratios (HRs) and statistical significance (P values) of specific lipid isoforms associated with survival.

[0017] Figure 2 Figure 2. Feature importance evaluation graph for XGBoost model.

[0018] Figure 3 This is the time-dependent ROC curve.

[0019] Figure 4 This is the risk score distribution diagram between the high-risk group and the low-risk group in the test set.

[0020] Figure 5 Figure 3 is the Kaplan-Meier (KM) curve of disease-free survival (DFS) of breast cancer patients based on lipid profiles. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is described in detail below through specific embodiments. It should be understood that the following specific embodiments are only exemplary, and any changes or modifications that do not depart from the technical solution design of the present invention should be within the scope of protection of the present invention. The present invention is described in detail below with reference to examples.

[0022] Example: A plasma lipid isomer biomarker for predicting survival benefit of neoadjuvant therapy for breast cancer, and construction of a prediction model for survival benefit of NAT in breast cancer (1) Clinical sample collection: Based on the Capital Clinical Characteristics Applied Research Project, the Optimization of Preoperative Chemotherapy Regimens for Different Breast Cancer Subtypes and the Exploration of Related Biomarkers (Trial Registration: ClinicalTrials.gov NCT02041338) and the Screening and Application of Lipid Biomarkers in Body Fluids (Project Number: 20220026) were conducted. Between January 2016 and January 2021, patients with biopsy-confirmed primary invasive breast cancer without distant metastases, diagnosed at clinical stages IIa to IIIc, were enrolled. All patients received 4-6 cycles of NAT before breast cancer surgery, with the treatment regimen and schedule guided by the National Comprehensive Cancer Network (NCCN) guidelines. NAT regimens primarily consisted of anthracyclines and paclitaxels. In addition, HER2-positive patients received targeted therapy with trastuzumab (8 mg / kg loading dose, followed by 6 mg / kg continuous doses) and pertuzumab (840 mg loading dose, followed by 420 mg continuous doses).

[0023] During neoadjuvant therapy, patients underwent physical examination and imaging examination (breast MRI or ultrasound) every two cycles to assess clinical efficacy. Efficacy was assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Clinical efficacy was categorized as complete remission (CR), partial remission (PR), stable disease (SD), and progressive disease (PD).

[0024] (2) Immunohistochemistry and pathological response evaluation Estrogen receptor (ER), progesterone receptor (PR), HER2, and Ki67 status were determined by immunohistochemistry (IHC). Tumors were classified as ER / PR-negative if nuclear staining was present in less than 1% of tumor cells, and as ER / PR-positive if nuclear staining was present in 1% or more of tumor cells. The Ki67 cutoff was set at 14%. HER2 negativity was defined as an IHC score of 0 or 1, and HER2 positivity was defined as an IHC score of 3. Tumors with an IHC score of 2 were subjected to fluorescence in situ hybridization (FISH) to confirm HER2 status by assessing the HER2 / CEP17 ratio and HER2 copy number. Standard histopathological analysis was performed by the National Cancer Center to assess pCR for NAT. Specimens were fixed in 10% neutral buffered formalin, processed overnight using a standard tissue processor, cut into 5-mm sections, and stained using an automated staining system. A pathologic complete response (pCR) was defined as the absence of residual invasive cancer, with residual cancer in situ allowed, and the absence of invasive disease in the ipsilateral sentinel node or axillary resected lymph nodes (yPT0 / isN0).

[0025] (3) Peripheral blood sample collection and processing Serial peripheral blood samples were collected from breast cancer patients before NACT treatment (C0) and after two cycles of treatment (C2). At baseline (C0) and after two cycles of treatment (C2), a total of 4 mL of fresh peripheral blood was collected from the patients and placed in EDTA-anticoagulated tubes. Plasma was gently mixed immediately after collection and initially separated within two hours. The plasma separation process involved centrifuging the blood collection tubes at 1600 g for 10 minutes at 4°C. After centrifugation, the upper plasma layer was aliquoted into multiple 2.0 mL centrifuge tubes and centrifuged at 16,000 g for 10 minutes at 4°C. The resulting supernatant was transferred to a new centrifuge tube as the final plasma sample and stored at -80°C. For plasma samples, 50 μL of plasma was diluted with 1 mL of deionized water in a 10 mL centrifuge tube. Methanol (1 mL) and chloroform (2 mL) were added, and the mixture was vortexed for 10 minutes and then centrifuged at 13,000 rpm for 15 minutes. The extraction process was repeated, and the combined chloroform layers were dried under a stream of nitrogen. The lipid extracts were reconstituted in 1 mL of methanol and stored at −20°C until analysis.

[0026] (4) High-precision lipidomics (LC-PB-MS / MS) analysis including isomer analysis A total of 200 individuals were enrolled in this study, and structural lipidomics analysis was performed on plasma samples collected before and after NAT treatment. Multidimensional structure analysis included MS1, fatty acyl chain levels, sn-position, and C=C position levels. The LC-PB-MS platform consisted of a Nexera SIL-30ACFV UPLC system (Shimadzu, Tokyo, Japan), a 9030Q-TOF mass spectrometer (Shimadzu, Tokyo, Japan), and a flow microreactor OMEGA (PURSPEC Technologies). The liquid chromatograph was equipped with a degasser, two pumps, an autosampler, and a column oven. Separation was performed on a BEH UPLC HILIC column (100 mm × 2.1 mm, 1.6 µm, Waters, Milford, MA, USA). The column temperature was set at 30°C. The mobile phases consisted of ACN / acetone isopropanol (50 / 48 / 2, v / v / v) and B: ammonium acetate in water (10 mM). Gradient elution was performed at a flow rate of 0.3 mL / min (A starting at 90%, decreasing to 85% at 2.4 min, decreasing to 80% at 3.2 min, maintaining at 80% from 3.2 to 5 min, decreasing to 70% at 5.1 min and maintaining this ratio until 6 min, increasing to 90% at 6.1 min and maintaining this ratio until 10 min). Mass spectrometric analysis was performed on a 9030 Q-TOF mass spectrometer (Shimadzu, Tokyo, Japan), recording mass spectra in the m / z 100–900 range in positive electrospray mode. Other parameters were as follows: ESI voltage, 4500 V; curtain gas, 25 psi; CAD gas, 8; interface heater temperature, 350°C; nebulizer gas 1 and gas 2, 55 psi; and dust removal potential, 80 V. Post-column PB reactions were performed using an OMEGA flow microreactor (PURSPEC Technologies). The inlet and outlet of the microreactor were connected to the LC column and ESI source via PEEK tubing (0.01 inch inner diameter, 1 / 16 inch outer diameter, 80 cm long). The instrument was equipped with a low-pressure mercury lamp (model 80-1057-01, BHK) emitting at 254 nm.

[0027] (5) Lipidomics analysis and feature collection LC-MS, LC-MS / MS, and LC-PB-MS / MS data were analyzed using the custom software LipidOA. Raw data were converted to an open file format (.mzmL) using Labsolutions software (Shimadzu, Tokyo, Japan), which LipidOA can read. For relative quantification at the component sum level, MS1 spectra of specific lipid classes and the corresponding internal standard, SPLASH, were imported into LipidOA and corrected for type I isotopes. LipidOA read LC-MS / MS data in negative ion mode and provided lipid identification at the fatty acyl (alkyl) chain and sn position levels. LipidOA read LC-PB-MS / MS data in positive ion mode and provided C=C position identification. The lipid class search included PC, PE, PG, PI, PS, CER, SM, LPC, and LPE. The precursor ion and MS / MS fragment mass tolerances were set to 10 ppm, respectively. Identification at the lipid chain, fatty acyl (alkyl) chain, and sn position levels, as well as the C=C position, was performed using class-specific retention times.

[0028] (6) Establishment of lipidomics profile of breast cancer Using high-precision relative quantitative lipidomics, we dynamically monitor the peripheral fine lipidomics profile at multiple levels, including lipid composition, subclasses, chain length and saturation, isomers, etc., to complete the identification and statistical analysis of differential substances.

[0029] Data preprocessing: Missing values ​​are always represented as "NA". For data without missing values, skip this step. Zero values ​​are processed to remove peaks with a zero value ratio greater than 50% (default setting) in all sample groups, because peaks with a large zero value ratio are mostly noise peaks. For lipidomics data acquisition, especially large-scale lipidomics data acquisition, the drift of metabolite signal intensity over time is a major interference factor. During the data acquisition process (within and between batches), unwanted changes in the measurement results of metabolic peaks are inevitable. The reasons include sample handling and preparation, liquid chromatography column degradation, matrix effects, mass spectrometer contamination, and nonlinear instrument drift during long-term operation.

[0030] (7) Pre-screening using statistical methods for lipid metabolism characteristics This study analyzed lipidomic profiles in 200 breast cancer patients with censored data. Of these 200 patients, 23 experienced recurrence during follow-up, while 177 remained relapse-free at the end of the follow-up period. The lipid profile matrix was calculated as (baseline value - post-treatment value) / post-treatment value, which normalized the lipidomic changes before and after treatment for each patient. By analyzing the baseline and post-treatment changes in all lipid profiles, a preliminary lipidomic profile was constructed.

[0031] In order to further screen lipid features that are significantly associated with disease-free survival (DFS) in breast cancer patients, we used univariate Cox proportional hazards regression analysis (Cox-PH). Through Cox-PH regression analysis, 6 lipid features that are significantly associated with DFS in breast cancer patients were screened out. These features are closely related to the patients' DFS, including the following lipid isomers: PC 17:1 (n-8) _18:1(n-7), PC 18:0_18:1(n-7), PC 14:0_18:1(n-10), PE 16:0_18:1(n-7), PE14:0_18:1(n-7) and PC 16:1 (n-7) _18:1(n-9). These features are closely related to the patients' DFS. The analysis results showed that the regression analysis p values ​​of the lipid features were all less than 0.05, indicating that these lipid isomers are statistically significant in predicting disease-free survival in breast cancer patients. Figure 1 The specific regression analysis results show the regression coefficient, hazard ratio (HR), and significance level (p-value) for each lipid profile. Ultimately, this analysis identified six key lipid isomer profiles, which will serve as input variables for the survival benefit prediction model.

[0032] (8) Construction of a survival benefit prediction model Based on the initial lipid feature screening, univariate Cox proportional hazards regression (Cox-PH) analysis was performed on all lipid features to identify lipid features significantly associated with disease-free survival (DFS) in breast cancer patients. This Cox-PH regression analysis ultimately identified six lipid isomer features: PC 17:1 (n-8) _18:1(n-7), PC 18:0_18:1(n-7), PC 14:0_18:1(n-10), PE 16:0_18:1(n-7), PE 14:0_18:1(n-7), and PC 16:1 (n-7) _18:1(n-9). These features were significantly correlated with DFS. Based on these pre-trained feature variables, we selected the extreme gradient boosting (XGBoost) regression model as the final modeling algorithm. XGBoost is a machine learning algorithm based on gradient boosting trees that can handle nonlinear data, has strong generalization capabilities, and is highly efficient in training. During the model building process, we randomly divided all data sets into training sets and test sets to ensure the generalization ability and stability of the model. Among them, 70% of the data was used as the training set, and the remaining 30% of the data was used as the test set. The training set is used for model training and hyperparameter tuning, and the test set is used to evaluate the final performance of the model. The hyperparameters of the XGBoost model were optimized by grid search (Grid Search) and 5-fold cross-validation (5-fold Cross-Validation) to determine the optimal hyperparameter configuration such as the number of trees, learning rate, and maximum depth. The optimized model achieved a C-index of 0.82 in the training set and a C-index of 0.81 in the test set, showing good survival prediction ability. In order to further evaluate the importance of the 6 screened lipid isomer features in survival prediction, the feature importance evaluation method of the XGBoost model was used to quantify the contribution of each lipid feature in the model. The feature importance score is based on the gain value (Gain), which indicates the contribution of each feature to the improvement of model accuracy. The results are as follows. Figure 2 shown.

[0033] To further evaluate the prediction effect of the model at different time points, the present invention constructed a time-dependent ROC curve to evaluate the patient's 1-year, 3-year and 5-year survival probability. In the 1-year, 3-year and 5-year predictions, the model's AUC for 1-year survival probability reached 0.957, the AUC for 3-year survival probability was 0.915, and the AUC for 5-year survival probability was 0.902, indicating that the model's ability to predict survival at different time points is stable and reliable. Figure 3 shown.

[0034] (9) Construction of risk scoring system First, the six lipid feature values ​​of the patients determined in advance were input into the optimized survival prediction model to calculate the risk score for each patient. To better evaluate the survival prognosis performance of the model in different risk groups, we further determined the optimal cut-off value of the risk score by combining the time-dependent ROC curve. Youden's index was used, which is a commonly used indicator for evaluating the performance of diagnostic models. The calculation formula of the Youden index is: J = Sensitivity + Specificity − 1 Among them, Sensitivity refers to the proportion of correct predictions among all true positive samples, and Specificity refers to the proportion of correct predictions among all true negative samples. The Youden Index aims to find a balance point that can maximize the sum of sensitivity and specificity, so as to obtain the optimal critical value of the risk score, so that the model can improve sensitivity while ensuring high specificity when distinguishing high-risk groups from low-risk groups. By analyzing the time-dependent ROC curve of 5-year survival prediction, we determined that the critical value of the risk score is 0.720, and the Youden index corresponding to this value is the maximum value, which optimizes the sensitivity and specificity of risk stratification. When the patient's risk score is greater than 0.720, the patient is judged to be in the high-risk group; when the risk score is lower than this value, the patient is classified as the low-risk group. The risk scores show significant separation between the high-risk group and the low-risk group, such as Figure 4 As shown in Figure 2, patients in the high-risk group have significantly higher risk scores. The Kaplan-Meier (KM) curve was further used to evaluate the difference in survival probability between patients in the high-risk group and the low-risk group in the training set and the test set. Figure 5 The results showed that the survival probability of patients in the high-risk group was significantly lower than that of the low-risk group. In the training set, the log-rank test p-value was less than 0.001; in the test set, the log-rank test p-value was 0.017, both of which were statistically significant. Based on this, the risk scoring system provides a powerful basis for survival prediction in clinical practice, helping doctors make more accurate and personalized treatment decisions for breast cancer patients.

[0035] Test Example 1: Following the aforementioned analysis of lipidomic data from breast cancer patients, plasma samples from 200 patients were analyzed. By combining baseline and post-treatment changes in lipidomic profiles with univariate Cox proportional hazards regression (Cox-PH) analysis, six significantly correlated lipid profiles were identified: PC 17:1 (n-8) _18:1(n-7), PC 18:0_18:1(n-7), PC 14:0_18:1(n-10), PE 16:0_18:1(n-7), PE 14:0_18:1(n-7), and PC 16:1 (n-7) _18:1(n-9). Using an XGBoost regression model, a disease-free survival (DFS) prediction model for breast cancer was constructed. The optimized model achieved a C-index of 0.82 in the training set and 0.81 in the validation set, demonstrating excellent survival prediction capabilities.

[0036] By further analyzing the predictive effect of the model at different time points, a time-dependent ROC curve was constructed to evaluate the patient's 1-year, 3-year, and 5-year survival probabilities. The model's AUC for 1-year survival was 0.957, the AUC for 3-year survival was 0.915, and the AUC for 5-year survival was 0.902, demonstrating the model's predictive stability and reliability at different time points. The critical value of the risk score determined by the Youden index was 0.720. Patients with a value above this value were classified as high-risk, and those with a value below this value were classified as low-risk. A further log-rank test (log-rank test) had a p-value of 0.017, indicating that the difference in survival probability between the high-risk and low-risk groups was statistically significant (p-value < 0.05).

Claims

1. A plasma lipid isomer biomarker for predicting survival benefit of neoadjuvant therapy for breast cancer, characterized in that: The plasma lipid isomer biomarkers are as follows: PC 17:1 (n-8) _18:1(n-7), PC 18:0_18:1(n-7), PC 14:0_18:1(n-10), PE 16:0_18:1(n-7), PE 14:0_18:1(n-7) and PC 16:1 (n-7) _18:1(n-9).

2. Use of the plasma lipid isomer biomarker as described in claim 1 in predicting the survival benefit of neoadjuvant therapy for breast cancer.

3. A model for predicting the survival benefit of breast cancer patients after neoadjuvant therapy, characterized in that: The modeling algorithm of the model is an extreme gradient boosting regression model, and the plasma lipid isomer biomarker as described in claim 1 is used to predict the survival benefit of neoadjuvant therapy for breast cancer.

4. The method for constructing the model according to claim 3, comprising the following steps: (1) All data sets are randomly divided into training sets and test sets. The training set is used for model training and hyperparameter tuning, and the test set is used to evaluate the final performance of the model. (2) Optimize the hyperparameters of the XGBoost model using grid search and 5-fold cross-validation to determine the optimal hyperparameter configurations such as the number of trees, learning rate, and maximum depth; (3) Evaluate the final model using the test set; (4) Substituting the characteristic values ​​of the six plasma lipid isomer biomarkers of the patient as described in claim 1 into the model to obtain a risk score. If the risk score is greater than the critical value, the patient is judged to be at high risk; when the risk score is lower than the critical value, the patient is judged to be at low risk.

5. The construction method according to claim 4, characterized in that In step (1), 70% of the data is used as the training set, and the remaining 30% of the data is used as the test set.

6. A risk scoring system based on the model obtained by the construction method according to claim 4, characterized in that: The critical value of the risk score is 0.

720.

7. A risk scoring system based on the model obtained by the construction method according to claim 5, characterized in that: The critical value of the risk score is 0.720.

Images