Method and device for predicting hepatitis B related liver cancer risk based on intestinal flora by using machine learning model

Abstract

The invention discloses a method for predicting hepatitis B related liver cancer risk based on intestinal flora by using a machine learning model, which is a non-invasive, convenient and feasible means by acquiring clinical and excrement samples of a patient for analysis. An excrement sample is sequenced to obtain an intestinal flora biomarker, screening is performed by setting a detection standard, intestinal flora characteristics, including bacteroides, lactobacillus, actinomycetes, erysipelothrix and Rosella, closely associated with the hepatitis B-associated liver cancer are found, a hepatitis B-associated liver cancer risk prediction model is constructed through a machine learning model, and the risk prediction model is used for predicting the risk of the hepatitis B-associated liver cancer. The defects of a traditional statistical analysis method and artificial analysis bias errors are avoided, the intestinal flora characteristics have higher prediction value and accuracy compared with traditional clinical characteristics, the method can be used for assisting early screening and early diagnosis of hepatitis B related liver cancer patients, and the overall survival prognosis of the patients is improved. The invention also provides a prediction device which is used for realizing hepatitis B related liver cancer risk prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of models for predicting the risk of liver cancer in patients with hepatitis B, and in particular to a method and device for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model. Background Art

[0002] Hepatocellular carcinoma is one of the most common malignant tumors in the world. The prevention and treatment of liver cancer has always been a top priority of the medical system. The risk factors of liver cancer mainly include hepatitis virus infection, aflatoxin and alcohol, and the main risk factor in my country is hepatitis B virus infection. Therefore, hepatitis B-related liver cancer is also the most common type in clinical practice. Surgical resection is the most common radical treatment for liver cancer. However, about 70% of patients are already in the middle and late stages at the time of initial diagnosis and have lost the opportunity for radical surgical treatment. These patients have limited treatment options and extremely poor survival prognosis. Therefore, how to predict the occurrence of hepatitis B-related liver cancer early, so as to achieve early diagnosis and early treatment of liver cancer and improve the survival prognosis of patients, is an important problem that needs to be solved in clinical practice.

[0003] There is a close intrinsic connection between the intestinal flora and a variety of hepatitis B-related liver diseases, which we call the "gut-liver" axis. In patients with hepatitis B virus infection, the composition of intestinal microorganisms has changed significantly, characterized by the increase of some pathogenic bacteria, such as Firmicutes, Prevotella, Proteobacteria and Streptococcus. Hepatitis B virus infection is associated with impaired intestinal barrier and translocation of intestinal flora in patients, which further leads to excessive immune response and inflammatory response, thereby causing the occurrence and development of liver cancer. Changes in the intestinal flora can regulate the occurrence of hepatitis B-related liver cancer through various mechanisms, such as microbial-host genetic modification, microbial-derived metabolites (such as short-chain fatty acids, bile acids and choline), immune microenvironment regulation and inflammation-related signal transduction. However, there is currently no study that clearly confirms the potential role and specific mechanism of action of the intestinal flora in hepatitis B-related liver cancer. Therefore, it is of great clinical value to clarify the role of the intestinal flora in the pathogenesis of hepatitis B-related liver cancer, thereby discovering important new intestinal flora biomarkers and constructing relevant prediction models for the early diagnosis of hepatitis B-related liver cancer.

[0004] In recent years, the application of artificial intelligence technology in the medical field has developed rapidly. Multi-omics feature models based on artificial intelligence technology have been gradually applied in the clinical diagnosis and treatment of various cancers, including thyroid cancer, breast cancer, gastric cancer, etc. In addition, machine learning, as one of the key technologies in the field of artificial intelligence, has been increasingly applied to medical research. Compared with traditional statistical analysis methods, machine learning has obvious advantages in the analysis and processing of medical data. Machine learning analysis can cover many aspects of medical research, including the distinction between benign and malignant tumors, the identification of potential biomarkers, the screening of patients, the prediction and diagnosis of diseases, etc. By extracting important clinical features to construct auxiliary diagnosis and treatment models, the constructed models can avoid the bias of traditional models and greatly improve the accuracy, effectiveness and stability of the models. For example, the invention application with publication number CN118538422A discloses a method, device, equipment and storage medium for constructing a risk prediction model for liver cancer in hepatitis B patients. By obtaining clinical data of hepatitis B and liver cancer patients, the collected data are analyzed by multi-factor Logistic regression using traditional statistical methods and the S3ABC nomogram prediction model is established to predict the risk of liver cancer in hepatitis B patients. However, the clinical data obtained by the method provided in this application are limited and are all common clinical indicators, which lack important clinical significance. At the same time, traditional analysis methods have poor processing capabilities for censored data and unbalanced data. The constructed model has poor predictive performance and stability and cannot be verified in other population cohorts. In addition, the human bias in the analysis process and results makes it impossible to achieve good clinical translational applications. Summary of the invention

[0005] The purpose of the present invention is to provide a method and device for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model. By collecting clinical and fecal sample data sets from patients in different cohorts, the clinical characteristics and intestinal flora characteristics of the patients are obtained, and a hepatitis B-related liver cancer risk prediction model is constructed based on a machine learning method. The shortcomings of traditional statistical analysis methods and human analysis bias errors are avoided, and non-invasive early screening and diagnosis of hepatitis B-related liver cancer patients are achieved, thereby improving the overall survival prognosis of patients.

[0006] To achieve the above-mentioned purpose of the invention, a method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model comprises the following steps:

[0007] Step 1: Construct clinical cohorts of patients, collect clinical characteristics and stool samples of patients in different clinical cohorts, and perform 16s RNA sequencing on stool samples to obtain statistical information on the abundance of intestinal flora;

[0008] Step 2: Based on the abundance statistics of intestinal flora, the distribution differences of intestinal flora in different cohorts of patients were found through principal coordinate analysis, and the differential flora between different patients was screened using linear discriminant analysis effect to obtain intestinal flora markers related to hepatitis B-related liver cancer;

[0009] Step 3: Based on the intestinal flora biomarkers, combined with the mediating regulatory role of intestinal flora in hepatitis B-related liver cancer, with hepatitis B infection as the independent variable and the occurrence of liver cancer as the dependent variable, the intestinal flora characteristics were obtained based on the 95% confidence interval of the mediation effect and the odds ratio of the regulation effect P < 0.05 as the detection standard;

[0010] Step 4: Use a machine learning model to construct a prediction model for hepatitis B-related liver cancer risk based on the acquired intestinal flora characteristics combined with clinical characteristics, and predict and verify the risk of hepatitis B-related liver cancer.

[0011] In one embodiment, the clinical cohort patients described in step 1 include healthy physical examination population, hepatitis B patients and liver cancer patients, and inclusion criteria and exclusion criteria are set for the clinical cohort patients, wherein the inclusion criteria include: (1) age 18-80 years old; (2) no antibiotic treatment 2 months before admission; (3) complete clinical data; (4) signed informed consent;

[0012] Exclusion criteria included: (1) patients with other malignancies other than liver cancer; (2) patients whose stool samples were not qualified; and (3) patients with other autoimmune diseases, metabolic diseases, or other types of hepatitis.

[0013] In one embodiment, the clinical characteristics of patients in different clinical cohorts described in step 1 include: epidemiological data, laboratory test data, and oncology-related data, wherein the epidemiological data include: gender, age, height, weight, body mass index, diabetes, hypertension, and PS score;

[0014] Laboratory examination data included: hepatitis B, liver cirrhosis, liver function indexes, tumor indexes, and coagulation function indexes;

[0015] Oncology-related data included: tumor size, tumor number, tumor stage, lymph node metastasis, and vascular invasion.

[0016] In one embodiment, the 16s RNA sequencing described in step 1 includes: constructing a library by PCR amplifying the microbial 16s RNA gene, sequencing the microbial RNA gene, and analyzing the relative abundance information of the intestinal flora in the feces.

[0017] In one embodiment, the intestinal flora characteristics in step 3 include: Bacteroides, Lactobacillus, Actinomycetes, Erysipelothrix and Roseburia;

[0018] The Bacteroidetes had a mediating effect, with a 95% confidence interval of [0.0002, 0.0400], P = 0.042;

[0019] The lactobacillus had a negative regulatory effect, with a 95% confidence interval of [-0.046, -0.002], P = 0.030;

[0020] The actinomycetes had a negative regulatory effect, with a 95% confidence interval of [-5.392, -0.035], P = 0.047;

[0021] The Erysipelothrix had a positive regulatory effect, with a 95% confidence interval of [0.053, 0.466], P = 0.014;

[0022] The genus Roseburia had a positive regulatory effect, with a 95% confidence interval of [0.111, 1.158], P = 0.017.

[0023] In one embodiment, the hepatitis B-related liver cancer risk prediction model is constructed by classification and regression tree, gradient boosting machine, logistic regression, neural network, random forest, support vector machine and extreme gradient boosting.

[0024] In one embodiment, the risk prediction of hepatitis B-related liver cancer described in step 4 is performed through ten-fold cross validation, and the ROC curve, calibration curve and decision curve are drawn to verify the model of hepatitis B-related liver cancer.

[0025] To achieve the above-mentioned purpose of the invention, an embodiment of the present invention also provides a device for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement the method of predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model when executing the computer program.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The method provided by the present invention for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model is a non-invasive and convenient technical means compared to the existing technology, by obtaining clinical and stool sample data of patients; key intestinal flora markers are obtained by sequencing based on stool sample data, and screening is performed by setting detection standards, and intestinal flora characteristics closely related to hepatitis B-related liver cancer are found, including Bacteroides, Lactobacillus, Actinomycetes, Erysipelothrix and Roseburia, and a hepatitis B-related liver cancer risk prediction model is constructed in combination with the machine learning model, which avoids the shortcomings of traditional statistical analysis methods and human analysis bias errors, and the intestinal flora characteristics have higher predictive value than traditional clinical characteristics. The machine learning model based on intestinal flora has higher predictive performance and accuracy, which provides an important reference for predicting the incidence risk of hepatitis B-related liver cancer in patients, assists in early screening and diagnosis of hepatitis B-related liver cancer patients, and improves the overall survival prognosis of patients.

[0028] The present invention also provides a device for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, which is used to achieve the risk prediction of hepatitis B-related liver cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0030] Figure 1 A flow chart of the method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model;

[0031] Figure 2 Alpha diversity analysis of intestinal flora in the liver cancer group and the control group;

[0032] Figure 3 Beta diversity analysis of intestinal flora in the liver cancer group and the control group;

[0033] Figure 4 This is the difference diagram of the dominant intestinal flora between the liver cancer group and the control group;

[0034] Figure 5 ROC curve of the prediction model based on clinical characteristics;

[0035] Figure 6 This is the ROC curve of the prediction model based on intestinal flora constructed using machine learning methods;

[0036] Figure 7 Calibration curve for the intestinal flora-based prediction model constructed using machine learning methods;

[0037] Figure 8Decision curve of the intestinal flora-based prediction model constructed using machine learning methods. DETAILED DESCRIPTION

[0038] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0039] In order to achieve non-invasive early screening and diagnosis of patients with hepatitis B-related liver cancer and improve the overall survival prognosis of patients, the embodiment provides a method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, and predicts the risk of hepatitis B-related liver cancer based on clinical characteristics and intestinal flora characteristics.

[0040] like Figure 1 As shown, the embodiment provides a method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, comprising the following steps:

[0041] Step 1: Construct clinical cohorts of patients, collect clinical characteristics and stool samples of patients in different clinical cohorts, and perform 16s RNA sequencing on stool samples to obtain statistical information on the abundance of intestinal flora.

[0042] In the embodiment, a clinical cohort of patients was constructed based on the set inclusion and exclusion criteria, including healthy people, hepatitis B patients, and liver cancer patients. The inclusion criteria are as follows: (1) age 18-80 years old; (2) no antibiotic treatment in the 2 months before admission; (3) complete clinical data; (4) signed informed consent. The exclusion criteria are as follows: (1) combined with other malignant tumors except liver cancer; (2) no qualified stool samples were collected; (3) combined with other autoimmune diseases, metabolic diseases or other types of hepatitis.

[0043] First, clinical characteristics of patients in different cohorts were collected, including epidemiological data (sex, age, height, weight, BMI, diabetes, hypertension, PS score, etc.), laboratory test data (hepatitis B, liver cirrhosis, liver function indexes, tumor indexes, coagulation function indexes, etc.), and oncology-related data (tumor size, tumor number, tumor stage, lymph node metastasis, vascular invasion, etc.);

[0044] Fecal samples were collected from patients in different clinical cohorts. The feces were freshly collected when the patients visited the hospital and frozen in a -80°C refrigerator within half an hour after sampling.

[0045] Then, the stool samples were further subjected to 16s RNA sequencing. The genomic DNA of the stool samples was extracted using the EZNA stool DNA kit (D4015, USA). The quality of DNA extraction was tested by agarose gel electrophoresis. The DNA was quantified using an ultraviolet spectrophotometer. Subsequently, different primers were selected for polymerase chain reaction (PCR) amplification. The primers were 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′). PCR products were purified by AMPure XT beads (Beckman Coulter Genomics, USA) and quantified by Qubit (Invitrogen, USA). PCR amplification products were detected by 2% agarose gel electrophoresis. The purified PCR products were evaluated using an Agilent 2100 bioanalyzer and an Illumina library quantification kit. Qualified sequencing libraries were gradient diluted, mixed in the corresponding proportions according to the required sequencing amount, and denatured into single-stranded DNA by NaOH for sequencing. NovaSeq 6000 sequencer was used for 2×250 bp paired-end sequencing, and the corresponding reagent was NovaSeq 6000SP Reagent Kit (500 cycles).

[0046] For the paired-end data obtained by sequencing, the samples were split according to the barcode information, and the ASV (Amplicon Sequence Variants) characteristic sequences and ASV abundance tables were further obtained, based on which the alpha diversity analysis of the bacterial community was performed ( Figure 2 ) and beta diversity analysis ( Figure 3 The constructed clinical cohort patients were divided into two groups: liver cancer group and control group. The alpha diversity analysis mainly evaluated the diversity within the habitat through seven indexes: observed_species, shannon, simpson, chao1, goods_coverage and pielou. Figure 2 The results showed that the Shannon index of liver cancer patients was higher than that of the control group (P = 0.034); Beta diversity analysis was mainly based on calculating four distances (weighted_unifrac, unweighted_unifrac, jaccard, bray_curtis), using six major analyses to evaluate the diversity between habitats (samples / groups), Figure 3 The results showed that the Ssmposon index of liver cancer patients was significantly higher than that of the control group (P=0.018).

[0047] Step 2: Based on the abundance statistical information of intestinal flora, the distribution differences of intestinal flora in different cohorts of patients were found through principal coordinate analysis. The differential flora between different patients was screened using linear discriminant analysis effect to obtain intestinal flora markers related to hepatitis B-related liver cancer.

[0048] Based on the obtained species abundance statistics, the difference analysis between the comparison groups was performed, and the distribution differences of intestinal flora in different patients were found through principal coordinates analysis (PCoA). The differential flora between different patients was screened using linear discriminant analysis effect size (LEfSe). Figure 4 The figure shows the difference in intestinal flora between the liver cancer group and the control group. Using the linear discriminant analysis score (LDA) > 2 as the standard, LEfse was used to evaluate the differences in microbial taxa between the liver cancer population and the normal population, and a new intestinal flora marker closely associated with hepatitis B-related liver cancer was discovered.

[0049] Step 3: Based on the intestinal flora biomarkers, combined with the mediating regulatory role of intestinal flora in hepatitis B-related liver cancer, with hepatitis B infection as the independent variable and the occurrence of liver cancer as the dependent variable, the intestinal flora characteristics were obtained based on the 95% confidence interval of the mediation effect and the odds ratio of the regulation effect P < 0.05 as the detection standard;

[0050] Based on intestinal flora biomarkers, the mediating and moderating role of intestinal flora in hepatitis B-related liver cancer was analyzed: Bootstrap and Johnson-Neyman techniques were used to explore the mediating / moderating role of intestinal flora in the relationship between hepatitis B and liver cancer. The Bootstrap method is a non-parametric statistical inference method used to test the significance of the mediation effect and estimate its confidence interval. The Johnson-Neyman technique aims to solve the specific method of evaluating the relationship between the independent variable and the dependent variable under different values ​​of the moderating variable after the interaction is significant.

[0051] Specifically, the hepatitis B infection status was used as the independent variable, the occurrence of liver cancer was used as the dependent variable, and the two groups of different intestinal microorganisms were used as mediating or moderating variables for analysis. The mediation effect was characterized by a 95% confidence interval or the moderating effect was characterized by an odds ratio of P < 0.05. The key intestinal flora characteristics that play a mediating and moderating role in hepatitis B-related liver cancer were obtained, among which the key intestinal flora characteristics include: Bacteroides, Lactobacillus, Actinomycetes, Erysipelothrix and Roseburia.

[0052] As shown in Table 1, in the embodiment, hepatitis B infection was used as the independent variable, the occurrence of liver cancer was used as the dependent variable, and two groups of different intestinal microorganisms were used as mediating / moderating variables for analysis. The mediation effect was analyzed with a 95% confidence interval, and the moderating effect was analyzed with an odds ratio of P < 0.05 as the detection standard.

[0053] Table 1 Key intestinal flora that play a mediating regulatory role between hepatitis B and liver cancer

[0054]

[0055]

[0056] The important role of intestinal flora in the development of hepatitis B-related liver cancer was further confirmed, and key intestinal flora markers that had not been confirmed before were found to play an important role in the occurrence and development of hepatitis B-related liver cancer: Bacteroidia (P=0.042, 95% confidence interval: [0.0002, 0.0400]) is an important mediating factor between hepatitis B and liver cancer;

[0057] Lactobacillales (P=0.030, OR=0.976, 95% confidence interval: [-0.046, -0.003]) and Actinomyces (P=0.047, OR=0.066, 95% confidence interval: [-5.392, -0.035]) play a negative regulatory role in HBV-related liver cancer;

[0058] Erysipelotrichaceae (P=0.014, OR=1.296, 95% confidence interval: [0.053, 0.466]) and Roseburia (P=0.017, OR=1.887, 95% confidence interval: [0.111, 1.158]) showed a positive regulatory effect in hepatitis B-related liver cancer.

[0059] In the future, it is hoped that by intervening in key intestinal flora, the risk of hepatitis B-related liver cancer can be reduced and the overall survival prognosis of patients can be improved.

[0060] Step 4: Use a machine learning model to construct a prediction model for hepatitis B-related liver cancer risk based on the acquired intestinal flora characteristics combined with clinical characteristics, and predict and verify the risk of hepatitis B-related liver cancer.

[0061] Machine learning analysis was performed using python software, and seven widely used supervised learning algorithms were used in the embodiment to explore the relationship between variables in the data set, including classification and regression tree (CART), gradient boosting machine (GBM), logistic regression (Logistic), neural network (NN), random forest (RF), support vector machine (SVM) and extreme gradient boosting (XGBoost).

[0062] Classification and Regression Tree (CART) can recursively perform binary splitting on each feature, and then predict the result of the input sample based on the input feature value. It can handle samples with missing values. It can be used for both classification and regression, and can model complex and nonlinear data, and obtain feasible and effective results for large data sources in a short time.

[0063] Gradient Boosting Machine (GBM) is a powerful ensemble learning algorithm that is often used to solve regression and classification problems. It improves the performance of the model by combining multiple weak prediction models (usually decision trees) and is known for its excellent predictive performance and flexibility.

[0064] Logistic regression is a probability-based supervised learning algorithm, which is often used for classification problems. Its advantages are simple model, fast calculation speed and strong interpretability.

[0065] Neural network (NN) is an information processing system based on the structure and function of brain neural network. It has extremely high accuracy in classification problems and has strong fault tolerance for various data problems such as nonlinear or unknown relationships.

[0066] Random Forest (RF) is a relatively new and highly flexible machine learning algorithm that can be used for classification and regression problems. It can alleviate the problem of model overfitting while improving the accuracy of the model. The support vector machine algorithm can solve problems such as small samples, high dimensions and nonlinear features, and the model has strong generalization ability.

[0067] Support vector machine (SVM) can solve the problems of small samples, high latitude and nonlinear features, and the model has strong generalization ability;

[0068] Extreme Gradient Boosting (XGBoost) adopts the integration idea and integrates multiple weak learners into a strong learner through certain methods, which can solve the problem of large-scale data.

[0069] In the embodiment, patients from different cohorts were randomly divided into training set and validation set in a ratio of 7:3. The training set was mainly used to train the model, and the validation set was used to evaluate the prediction performance of the model to prevent the model from overfitting. The model performance was further tested by 10-fold cross validation. First, a prediction model based on traditional clinical indicators was constructed, and the ROC curve (Receiver Operating Characteristic Curve) was drawn. The results showed that the AUC (Area Under Curve) value of the traditional prediction model based on clinical indicators was only 0.661 ( Figure 5 );

[0070] As shown in Table 2 , the HBV-related liver cancer risk prediction model was constructed using 7 machine learning models, and the performance was further verified by 10-fold cross validation.

[0071] Table 2 10-fold cross validation results of 7 machine learning models based on intestinal flora

[0072]

[0073] Combining Table 2 and Figure 6 The ROC curves drawn showed that the AUCs of the hepatitis B-related liver cancer risk prediction models constructed using seven machine learning models all exceeded 0.8, among which the AUC based on support vector machine (SVM) was as high as 0.91.

[0074] In addition, if Figure 7 The calibration curve drawn as shown further reflects the matching degree between the absolute risk predicted by the constructed hepatitis B-related liver cancer risk prediction model and the actually observed risk, confirming the validity and accuracy of the new model; Figure 8 As shown, the decision curve analysis (DCA) of the constructed hepatitis B related liver cancer risk prediction model also further confirmed the practicality and clinical application value of the constructed hepatitis B related liver cancer risk prediction model.

[0075] The constructed intestinal flora-based hepatitis B-related liver cancer risk prediction model was compared with the traditional model constructed by statistical methods based on clinical characteristics. It was found that the intestinal flora has a higher predictive value than traditional clinical characteristics. At the same time, the intestinal flora model constructed with the assistance of machine learning can significantly improve the prediction performance and accuracy of the model, while having good robustness and applicability. The constructed intestinal flora machine learning model will be further prospectively applied to the diagnosis and treatment of clinical patients, assist in the assessment of the risk of liver cancer in hepatitis B patients, assist in the early screening and diagnosis of hepatitis B-related liver cancer patients, and also reduce the risk of hepatitis B-related liver cancer by intervening in key intestinal flora and improve the overall survival prognosis of patients.

[0076] Based on the above method, the embodiment also provides a device for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, which is used to achieve the risk prediction of hepatitis B-related liver cancer.

[0077] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, characterized in that: The following steps are involved: Step 1: Construct clinical cohorts of patients, collect clinical characteristics and stool samples of patients in different clinical cohorts, and perform 16s RNA sequencing on stool samples to obtain statistical information on the abundance of intestinal flora; Step 2: Based on the abundance statistics of intestinal flora, the distribution differences of intestinal flora in different cohorts of patients were found through principal coordinate analysis, and the differential flora between different patients was screened using linear discriminant analysis effect to obtain intestinal flora markers related to hepatitis B-related liver cancer; Step 3: Based on the intestinal flora biomarkers, combined with the mediating regulatory role of intestinal flora in hepatitis B-related liver cancer, with hepatitis B infection as the independent variable and the occurrence of liver cancer as the dependent variable, the intestinal flora characteristics were obtained based on the 95% confidence interval of the mediation effect and the odds ratio of the regulation effect P < 0.05 as the detection standard; Step 4: Use a machine learning model to construct a prediction model for hepatitis B-related liver cancer risk based on the acquired intestinal flora characteristics combined with clinical characteristics, and predict and verify the risk of hepatitis B-related liver cancer.

2. The method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model according to claim 1, characterized in that: The clinical cohort patients described in step 1 include healthy people, hepatitis B patients, and liver cancer patients, and the inclusion and exclusion criteria are set for the clinical cohort patients. The inclusion criteria include: (1) age 18-80 years old; (2) no antibiotic treatment in the 2 months before admission; (3) complete clinical data; (4) signed informed consent; Exclusion criteria included: (1) patients with other malignancies other than liver cancer; (2) patients whose stool samples were not qualified; and (3) patients with other autoimmune diseases, metabolic diseases, or other types of hepatitis.

3. The method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model according to claim 1, characterized in that: Clinical characteristics of patients in different clinical cohorts described in step 1, including epidemiological data, laboratory test data, and oncology-related data. Epidemiological data include gender, age, height, weight, body mass index, diabetes, hypertension, and PS score; Laboratory examination data included: hepatitis B, liver cirrhosis, liver function indexes, tumor indexes, and coagulation function indexes; Oncology-related data included: tumor size, tumor number, tumor stage, lymph node metastasis, and vascular invasion.

4. The method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model according to claim 1, characterized in that: The 16s RNA sequencing described in step 1 includes: constructing a library by PCR amplifying the microbial 16s RNA gene, sequencing the microbial RNA gene, and analyzing the relative abundance information of the intestinal flora in the feces.

5. The method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model according to claim 1, characterized in that: The intestinal flora characteristics described in step 3 include: Bacteroides, Lactobacillus, Actinomycetes, Erysipelothrix and Roseburia.

6. The method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model according to claim 1, characterized in that: The HBV-related liver cancer risk prediction model described in step 4 was constructed using classification and regression trees, gradient boosting machines, logistic regression, neural networks, random forests, support vector machines, and extreme gradient boosting.

7. The method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model according to claim 1, characterized in that: The risk prediction of hepatitis B-related liver cancer described in step 4 is verified by ten-fold cross validation, and the ROC curve, calibration curve and decision curve are drawn to verify the model of hepatitis B-related liver cancer.

8. A device for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model, comprising a memory and a processor, wherein the memory is used to store a computer program, characterized in that: The processor is used to implement the method for predicting the risk of hepatitis B-related liver cancer based on intestinal flora using a machine learning model as described in any one of claims 1 to 8 when executing the computer program.

Images