Biomarker combination for evaluating occurrence risk of liver cancer microvascular invasion and application of biomarker combination

Abstract

The invention discloses a biomarker combination for evaluating the occurrence risk of liver cancer microvascular invasion and application of the biomarker combination. The biomarker combination is prepared from V2FC, H2FC, N, N-dimethylglycine, H1FC, lactic acid, creatine and pyruvic acid. The biomarker combination provided by the invention can be applied to preparation of a product for evaluating the risk of occurrence of liver cancer microvascular invasion, a liver cancer-MVI prediction model constructed based on the biomarker combination provided by the invention can rapidly and accurately predict the risk of occurrence of MVI, preoperative accurate evaluation of MVI is facilitated, and the risk of occurrence of liver cancer microvascular invasion is reduced. The personalized treatment of the liver cancer patient is realized, and the prognosis of the liver cancer patient is improved.

Description

A combination of biomarkers for assessing the risk of microvascular invasion in liver cancer and their applications Technical Field

[0001] This invention belongs to the field of biomedical technology and relates to a combination of biomarkers for assessing the risk of microvascular invasion in liver cancer and their application. Background Technology

[0002] Primary liver cancer (HCC) is a malignant tumor with high incidence and mortality rates worldwide. Hepatocellular carcinoma (HCC) is the main pathological type of HCC. Surgical resection is the primary treatment option for patients with early-stage HCC, but the postoperative recurrence rate remains high, seriously affecting patient prognosis.

[0003] Microvascular invasion (MVI) is a key factor in postoperative recurrence and poor prognosis in HCC patients. Its presence indicates a more aggressive biological behavior of HCC and is a crucial reference factor for assessing the likelihood of liver cancer recurrence and assisting in the formulation of treatment strategies. Studies have shown that for high-risk patients with MVI, surgical resection as first-line treatment provides better tumor control and a lower tumor recurrence rate than radiofrequency ablation (P<0.05). MVI is closely related to poor prognosis after liver transplantation. Transarterial ACE combined with sorafenib can improve the prognosis of HCC patients with MVI, while MVI-negative patients do not benefit from the above treatments (Xu Lingcong. Construction of a preoperative prediction model and prognostic analysis of microvascular invasion in hepatocellular carcinoma based on lipid metabolism markers [D]. Lanzhou University). Therefore, accurate preoperative prediction of the risk of MVI will help guide the selection of clinical treatment strategies and prognostic assessment for HCC. However, currently, MVI results can only be obtained through histopathological examination of the resected surgical specimen, lacking practical guiding significance for treatment strategies. Although tumor volume, serum alpha-fetoprotein (AFP), and serum hepatitis B DNA load have been reported to be associated with the risk of MVI in HCC, their predictive efficacy for predicting the risk of MVI still needs to be improved. Summary of the Invention

[0004] The main objective of this invention is to provide a combination of biomarkers for assessing the risk of microvascular invasion in liver cancer and its application, in order to solve at least one of the above-mentioned technical problems.

[0005] According to a first aspect of the invention, a combination of biomarkers for assessing the risk of microvascular invasion in liver cancer is provided, comprising very low-density lipoprotein-free cholesterol-2 component (VLDL-2 free cholesterol, V2FC), high-density lipoprotein-free cholesterol-2 component (HDL-2 free cholesterol, H2FC), N,N-dimethylglycine, high-density lipoprotein-free cholesterol-1 component (HDL-1 free cholesterol, H1FC), lactic acid, creatine, and pyruvic acid.

[0006] The biomarker combination of seven metabolites or blood lipid indicators provided by this invention can be used to quickly and accurately predict the risk of MVI, which is beneficial for accurate preoperative assessment of MVI, and thus facilitates personalized treatment for liver cancer patients and improves their prognosis.

[0007] According to a second aspect of the invention, the use of the biomarker combination of the invention in the preparation of products for assessing the risk of microvascular invasion in liver cancer is provided.

[0008] Based on the biomarker combination provided by this invention, a liver cancer-MVI prediction model can be constructed using various algorithms such as Extreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Machine (SVM), and Naive Bayes (NB). Among these, the HCC-MVI prediction model constructed based on SVM achieved an AUC of 0.957 for predicting MVI in a complete cohort containing 86 HCC patients, outperforming the predictive efficacy of other clinical features. Furthermore, the model exhibits high robustness and clinical applicability.

[0009] In some implementations, the liver cancer-MVI prediction model can be deployed as an application, mini-program, APP, etc. Only the metabolic levels of the seven biomarkers in the biomarker combination provided by this invention in plasma need to be input to assess the risk of MVI in liver cancer patients.

[0010] According to a third aspect of the invention, the use of a product for detecting the metabolic levels of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine and pyruvate in a sample in the preparation of a product for assessing the risk of microvascular invasion in liver cancer is provided.

[0011] In some embodiments, the product used to detect the metabolic levels of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine, and pyruvate in a sample can be any reagent, kit, chip, and / or instrument known in the art that can quantitatively detect the metabolic levels of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine, and pyruvate in a sample.

[0012] In some implementations, products for detecting the metabolic levels of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine, and pyruvate in a sample can be reagents, kits, chips, and / or instruments suitable for detecting the metabolic levels of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine, and pyruvate in a sample by nuclear magnetic resonance spectroscopy (NMR).

[0013] In some embodiments, the sample can be plasma. Therefore, using the biomarker combination provided by this invention to assess the risk of MVI in liver cancer offers advantages such as convenient and non-invasive sample acquisition, making it easy to promote and apply clinically.

[0014] In some implementations, liver cancer is referred to as HCC. Attached Figure Description

[0015] Figure 1 shows a forest plot of the AUC values ​​of 10 HCC-MVI prediction models predicting MVI in the training set; Figure 2 shows a forest plot of the AUC values ​​of 10 HCC-MVI prediction models predicting MVI in the validation set; Figure 3 shows a line chart of the performance metrics of 10 HCC-MVI prediction models in the training set; Figure 4 shows a line chart of the performance metrics of 10 HCC-MVI prediction models in the validation set; Figure 5 shows a radar chart of the performance metrics of the SVM model in the training set; Figure 6 shows a radar chart of the performance metrics of the SVM model in the validation set; Figure 7 shows the confusion matrix of the SVM model in the training set; Figure 8 shows the confusion matrix of the SVM model in the validation set; Figure 9 shows the calibration curve of the SVM model in the training set; Figure 10 shows the calibration curve of the SVM model in the validation set; Figure 11 shows the decision curve analysis of the SVM model in the complete queue; Figure 12 shows the SVM model's decision curve in the full queue. ROC curves of M-model and clinical features predicting MVI in the complete cohort; Figure 13 is a summary SHAP plot of global feature importance sorted by average |SHAP| value during the Shap method to quantify the contribution of each variable to the prediction; Figure 14 is a beeswarm plot of the direction and distribution of feature influence; Figure 15 is a SHAP dependency plot of V2FC in the SVM model; Figure 16 is a SHAP dependency plot of creatine in the SVM model; Figure 17 is a SHAP dependency plot of H1FC in the SVM model; Figure 18 is a SHAP dependency plot of H2FC in the SVM model; Figure 19 is a SHAP dependency plot of lactate in the SVM model; Figure 20 is a SHAP dependency plot of pyruvate in the SVM model; Figure 21 is a SHAP dependency plot of N,N-dimethylglycine in the SVM model. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the embodiments. The embodiments are for illustrative purposes only and do not limit the invention in any way. Unless otherwise specified, the raw materials and reagents used in the embodiments are conventional products that can be obtained commercially; experimental methods that do not specify specific conditions in the embodiments are generally performed under conventional conditions in the art or according to the conditions recommended by the manufacturer.

[0017] Example 1: Validation of the diagnostic efficacy of a combination of biomarkers for assessing the risk of microvascular invasion in hepatocellular carcinoma (HCC). 1. Clinical information of participants: Preoperative blood samples were collected from 86 patients with pathologically confirmed HCC who underwent radical resection. Clinical data were retrospectively collected, including age, sex, cirrhosis status, alpha-fetoprotein (AFP) level, abnormal prothrombin level, tumor size, tumor multifocality, vascular invasion status, and tumor stage. Among them, 37 patients were confirmed to have microvascular invasion (MVI) by histopathological examination of the resected surgical specimens.

[0018] Eighty-six patients were randomly assigned to a training set and a validation set in a 6:4 ratio. Differences in continuous clinical characteristics between the groups were compared using t-tests or Mann-Whitney U tests, while differences in categorical clinical characteristics between the groups were compared using chi-square tests or Fisher's exact tests. The results are shown in Table 1. There were no significant differences in baseline clinical characteristics between the two groups.

[0019] Table 1. Baseline characteristics of training and validation set patients used to predict microvascular invasion in hepatocellular carcinoma patients.

[0020] 2. Quantitative detection of the metabolic level of biomarkers in plasma (1) Sample processing: Blood samples were immediately centrifuged at 1500~2000×g for 10 minutes at 2~8℃ after collection. The supernatant was then aliquoted into cryovials and plasma samples were stored at -80℃.

[0021] (2) Nuclear magnetic resonance metabolomics analysis: Nightingale proton nuclear magnetic resonance (NMR) was used. 1 Plasma samples were analyzed using the 1H-NMR platform (ProteinT, Tianjin, China). After thawing, 340 μL of plasma was mixed with an equal volume of NMR lipid buffer (Bruker Plasma Buffer), and 600 μL of the mixture was transferred to a 5 mm NMR tube for automated analysis.

[0022] (3) Metabolite identification and quantification: Spectroscopic data were processed using Bruker's built-in Amix software, and chemical shift correction was performed using Speaq software. Metabolite identification and quantification were performed based on Bruker's proprietary NMR database, and metabolic level data of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine, and pyruvate were obtained from 86 plasma samples.

[0023] 3. Construction and Validation of HCC-MVI Prediction Models: Based on the metabolic level data of V2FC, H2FC, N,N-dimethylglycine, H1FC, lactate, creatine, and pyruvate in plasma, ten algorithms were used to construct 10 HCC-MVI prediction models: Extreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Machine (SVM), Naïve Bayes (NB), Gradient Boosting Machine (GBM), Generalized Linear Model (GLM), Generalized Linear Model with Elastic-Net (GLMNET), Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), and Recursive Partitioning Tree (RPART). The Shapley Additive ExPlanations (Shap) were used to interpret the models.

[0024] The model employed five-fold cross-validation and ten repeated runs. Model performance was evaluated based on sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), and F1 score. The optimal classification threshold was determined by maximizing the Youden index under the ROC curve. Calibration curve and decision curve analysis were used to assess predictive accuracy and net clinical benefit. The Shap method was used to quantify the contribution of each variable to the prediction.

[0025] The results are shown in Figures 1-21 and Table 2.

[0026] Table 2 Performance metrics of 10 HCC-MVI prediction models on the training and validation sets.

[0027] As shown in Figures 1 and 2, in terms of MVI diagnosis, the AUC values ​​of the 10 HCC-MVI prediction models constructed ranged from 0.723 to 1.000 in the training set and from 0.613 to 0.921 in the validation set.

[0028] The multi-dimensional performance comparison results of 10 HCC-MVI prediction models are shown in Figures 3-4 and Table 2. Compared with other models, the SVM model showed the best diagnostic performance on both the training and validation sets, with AUC values ​​of 0.981 and 0.921, respectively. The training set accuracy was 92.5%, sensitivity was 100%, and specificity was 86.7%; the corresponding values ​​on the validation set were 84.8%, 85.7%, and 84.2% (Figures 5-6).

[0029] The SVM model was selected as the optimal prediction model for further performance evaluation. The confusion matrix showed that the SVM model correctly identified all 23 MVI samples and 26 out of 30 non-MVI samples in the training set; in the validation set, the SVM model correctly classified 12 out of 14 MVI samples and 16 out of 19 non-MVI samples (Figures 7-8). These results indicate that the SVM model has good classification balance. The Brier scores for the training and validation sets were 0.080 (95% CI: 0.055-0.110) and 0.110 (95% CI: 0.066-0.169), respectively (Figures 9-10). No significant calibration differences were observed between the two cohorts, indicating consistent calibration performance. As shown in Figure 11, decision curve analysis showed that the SVM model provided higher net benefits than the "full treatment" and "no treatment" strategies at most threshold probabilities, indicating that the model has good clinical application value. The SVM model achieved an AUC of 0.957 for predicting MVI in the complete cohort, which was superior to other clinical predictors (Figure 12).

[0030] The Shap method was used to interpret the SVM model and quantify the contribution of each feature to the prediction results (Figures 13-14). The Shap dependency plot showed that lactate, creatine, and pyruvate were positively correlated with the Shap value, indicating that they made a positive contribution to MVI prediction. Conversely, V2FC, H1FC, H2FC, and N,N-dimethylglycine were negatively correlated with the Shap value, suggesting that they made a negative contribution to the prediction results (Figures 15-21).

[0031] The above descriptions are merely some embodiments of the present invention. Those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the scope of protection of the present invention.

Claims

1. A combination of biomarkers for assessing the risk of microvascular invasion in liver cancer, characterized in that, The biomarker combination consists of very low-density lipoprotein-free cholesterol-component 2, high-density lipoprotein-free cholesterol-component 2, N,N-dimethylglycine, high-density lipoprotein-free cholesterol-component 1, lactic acid, creatine, and pyruvate.

2. The use of the biomarker combination according to claim 1 in the preparation of products for assessing the risk of microvascular invasion in liver cancer.

3. The use of a product that detects the metabolic levels of each biomarker in the biomarker combination according to claim 1 in a sample in the preparation of a product for assessing the risk of microvascular invasion in liver cancer.

4. The application according to claim 3, characterized in that, The sample was plasma.

5. The application according to claim 3 or 4, characterized in that, The liver cancer mentioned is hepatocellular carcinoma.

6. The application according to claim 5, characterized in that, The product used to detect the metabolic levels of each biomarker in the biomarker combination according to claim 1 in the sample is a reagent, kit, chip, and / or instrument suitable for detecting the metabolic levels of each biomarker in the biomarker combination according to claim 1 in the sample by nuclear magnetic resonance spectroscopy.

Images