Gall-stone disease risk prediction method and system based on oral flora

Abstract

The invention discloses a gall-stone disease risk prediction method and system based on oral flora, and belongs to the technical field of gall-stone disease risk prediction. The method comprises the following steps: collecting samples, wherein the samples comprise oral bacterial species and abundance of a gall-stone patient group and a healthy control group; comparing the gall-stone patient group sample with the healthy control group sample, and determining the relevance between the oral flora and the gall-stone disease risk; selecting a plurality of oral bacterial species as input feature species of the prediction model; and inputting the feature value corresponding to the input feature type of the to-be-tested person into the random forest model to predict the gall-stone disease risk of the to-be-tested person. According to the method, the oral bacterial species and abundance in the sample can be obtained through 16S rRNA gene sequencing only by collecting the saliva sample of the to-be-detected person, and compared with an existing prediction model, the sample collection difficulty is greatly reduced, and popularization is more facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of gallstone disease risk prediction, and in particular relates to a gallstone disease risk prediction method and system based on oral flora. Background Art

[0002] Gallstones are a common digestive system disease that not only affects the local area, but also significantly increases the risk of other diseases. Although laparoscopic cholecystectomy has played an important role in the treatment of gallstones in the past few decades, while generating excessive medical economic burdens, its postoperative complications also pose a threat to patients' health. Therefore, early prediction of gallstones and timely clinical screening and preventive intervention / treatment of high-risk patients are the key to preventing the occurrence and development of gallstones. The identification of risk factors related to gallstone formation plays an important role in risk prediction. Current studies have shown that there are 96 risk factors for gallstone formation, and studies have pointed out that intestinal flora disorders may play an important role in the auxiliary diagnosis of gallstones, but due to the inconvenience of stool sampling, its clinical use value is limited. Therefore, it is crucial to establish a convenient, efficient and economical gallstone risk prediction model and take appropriate intervention measures in a timely manner for the prevention or early diagnosis of gallstones.

[0003] In the paper "Construction and Multicenter Validation Study of Gallstone Risk Prediction Model" published by the inventor on May 20, 2024, a method for predicting the risk of gallstones is disclosed. The method believes that women, old age, higher body mass index, fasting blood sugar, uric acid, total bilirubin, γ-glutamyl transpeptidase and fatty liver are positively correlated with the risk of gallstones. Gallbladder polyps, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol and aspartate aminotransferase are negatively correlated with the risk of gallstones. Although this method can also be used to predict the risk of gallstones, the above characteristics also have the problem of inconvenient sampling. Summary of the invention

[0004] In view of this, the present invention provides a method and system for predicting the risk of gallstones based on oral flora, which uses oral flora as a prediction indicator to facilitate sampling. In order to solve the above technical problems, the technical solution of the present invention is to adopt a method for predicting the risk of gallstones based on oral flora, comprising: Collecting samples, wherein the samples include oral bacteria species and their abundance in a gallstone patient group and a healthy control group; Compare the gallstone patient group samples with the healthy control group samples to determine the association between oral flora and gallstone risk; Select several oral bacteria species as input feature species for the prediction model; The feature values ​​corresponding to the input feature types of the subject are input into the random forest model to predict the risk of gallstones in the subject.

[0005] As an improvement, the method of selecting several oral bacteria species in the oral flora as input feature species of the prediction model includes: The prediction model is a random forest model, which can use the collected samples to predict the risk of gallstones and output the importance score of oral bacteria species; The prediction results of the prediction model are verified by using the cross-validation method and the area under the curve validation method, and the error rate and area under the curve value of the prediction model are obtained respectively; Oral bacteria species were ranked from high to low according to their importance scores, and several oral bacteria species ranked higher than a threshold were selected as input feature species for the prediction model based on the error rate and the area under the curve. As a further improvement, the random forest model is: ; Among them, I(.) is the indicator function, and its value is 1 when the condition in the brackets is met, otherwise it is 0; hm(x) is the decision tree function with sequence number m in the random forest rule; k is the predicted value; H(x) is the random forest model function; and n is the number of decision trees. As another further improvement, the steps of the cross-validation method include: Divide the collected samples into several sub-datasets of equal size; Use a sub-dataset as the test set and the remaining sub-datasets as the training set to cross-validate the random forest model and calculate the single validation error rate; repeat this step until all sub-datasets are used as test sets; The error rate of the random forest model is calculated using the single-shot validation error rate.

[0006] As an improvement, using the formula: ; Calculate the single verification error rate; where ErrorRate i is the single validation error rate of the i-th cross validation; m i is the number of samples with prediction errors; For the test set The number of samples; Using the formula: ; Calculate the error rate of the random forest model; where ErrorRate cv ErrorRate iis the single validation error rate of the i-th cross validation; j is the number of sub-datasets.

[0007] As an improvement, the area under the curve verification method uses the formula: ; Calculate the area under the curve; AUC is the area under the curve; TPR is the true positive rate, and FPR is the false positive rate.

[0008] As an improvement, saliva samples were collected from the gallstone patient group and the healthy control group during sample collection and subjected to 16S rRNA gene sequencing.

[0009] As an improvement, methods to determine the association between oral flora and gallstone risk include: The Ace index, Chao1 index, Shannon index, and Simpson index based on OTU levels were used to test the differences between the gallstone patient group and the healthy control group. Compare the results of several distance algorithms based on their contribution to sample differences; Differences in oral flora at the genus level were tested between samples from the gallstone patient group and the healthy control group.

[0010] As an improvement, after selecting the types of input features, the accuracy of the prediction model was verified using the collected samples through receiver operating characteristic analysis.

[0011] The present invention also provides a gallstone disease risk prediction system based on oral flora, which is characterized by comprising: A sample collection module, used to collect samples, wherein the samples include oral bacteria species and their abundance in a gallstone patient group and a healthy control group; A correlation determination module is used to compare the samples of the gallstone patient group with the healthy control group to determine the correlation between the oral flora and the risk of gallstones; A feature type selection module is used to select several oral bacteria species as input feature types of the prediction model; The prediction module is used to input the feature values ​​corresponding to the input feature types of the test subject into the random forest model to predict the risk of gallstones in the test subject.

[0012] The present invention is beneficial in that: The present invention determined the correlation between oral flora and the risk of gallstones by comparing samples from the gallstone patient group with samples from the healthy control group. 21 biomarkers (oral bacterial species) in the oral flora that predict the risk of gallstones were found through random forest models, cross-validation, and AUC validation, and the biomarkers were used as input features of the prediction model, thereby greatly improving the prediction accuracy of the model and avoiding the interference of irrelevant features on the prediction results.

[0013] In the present invention, only a saliva sample of the subject needs to be collected to obtain the types and abundance of oral bacteria in the sample through 16S rRNA gene sequencing. Compared with the existing prediction models, the difficulty of sample collection is greatly reduced, which is more conducive to popularization. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of Example 1 of the present invention.

[0015] Figure 2 (a) is a comparison chart of Ace index results.

[0016] Figure 2 (b) is a comparison chart of Chao1 index results.

[0017] Figure 2 (c) is the comparison chart of Shannon index results Figure 2 (d) is a comparison chart of Simpson index results.

[0018] Figure 3 This is a PCoA analysis diagram based on the OTU level.

[0019] Figure 4 This is a species diagram showing differences between groups at the genus level.

[0020] Figure 5 Schematic diagram of cross-validation results of the random forest prediction model with important features.

[0021] Figure 6 Schematic diagram of the AUC verification results of the random forest prediction model with important features.

[0022] Figure 7 Schematic diagram of the top 21 bacterial genera in importance in random forest analysis (after removing genera with "unclassified" in their names).

[0023] Figure 8 Schematic diagram of ROC analysis (training set) of oral bacteria species obtained based on random forest analysis.

[0024] Fig. 9Schematic diagram of ROC analysis (test set) of biomarkers obtained based on random forest analysis. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with specific implementation methods.

[0026] Example 1 like Figure 1 As shown, the present invention provides a method for predicting the risk of gallstones based on oral flora, comprising: S1 collects samples, which include oral bacteria species and their abundance in a gallstone patient group and a healthy control group.

[0027] In this example, 34 patients with gallstones and 78 healthy controls were included. When collecting samples, saliva samples were collected from the gallstone patient group and the healthy control group, and the species and abundance of each oral bacteria genus in the saliva samples were obtained by 16S rRNA gene sequencing.

[0028] The so-called bacterial genus abundance refers to the content or quantity of a certain bacterial genus in a specific environment. The bacterial genus abundance reflects the distribution of different bacterial genera in a specific environment or sample, and is an important parameter in microbial ecology research.

[0029] S2 compared the samples of the gallstone patient group with those of the healthy control group to determine the association between oral flora and the risk of gallstones.

[0030] In the research process, it is necessary to first determine whether there is a statistical significance between oral flora and the risk of gallstones, otherwise the subsequent prediction work lacks corresponding theoretical support. This step compares the α diversity and β diversity of oral flora between the two groups, and conducts species difference analysis to explore whether changes in oral flora are associated with gallstones.

[0031] Alpha diversity is mainly used to describe the diversity of species within a single sample or community, measuring the number and uniformity of different species within a sample. Beta diversity is used to describe the differences in species composition between different samples or communities, measuring the changes in species composition between samples.

[0032] In this embodiment, the method for determining the correlation between oral flora and the risk of gallstones includes: S21 The Ace index, Chao1 index, Shannon index and Simpson index based on OTU level were used to test the inter-group differences between the gallstone patient group and the healthy control group.

[0033] OTU (Operational Taxonomic Unit) refers to "operational taxonomic unit", which is a taxonomic unit used to represent microbial communities with high similarity in microbiome research. OTU is usually based on sequence data. By calculating the similarity between sequences, sequences with similarity above a certain threshold (such as 97%) are clustered together to form an OTU. Each OTU represents a small group of similar sequences in the microbial community.

[0034] Ace Index: The full name is “Abundance-based Coverage Estimator”, which is used to estimate the number of species in a sample, especially for low-abundance species.

[0035] Chao1 index: used to estimate species richness in a sample, with a particular focus on the discovery of low-frequency species.

[0036] Shannon index: A measure of the diversity of microbial communities that takes into account species richness and evenness.

[0037] Simpson Index: It is also a way to measure diversity, with a particular focus on the impact of dominant species, where lower values ​​indicate higher diversity.

[0038] The results are as follows Figure 2 As shown in (a)~2(d), the results of Ace index and Chao1 index showed that the richness of oral flora in the gallstone patient group was significantly lower than that in the healthy control group (p < 0.05), and Shannon index and Simpson index showed the diversity between the two groups.

[0039] S22 compares the results obtained by several distance algorithms according to their contribution to sample differences.

[0040] This step uses PCoA (Principal Coordinates Analysis) based on the OTU (Operational Taxonomic Unit) level, which is a multivariate statistical method used to explore and visualize the similarities and differences in microbial community structure and diversity. PCoA reduces the dimension of the distance matrix between samples and projects high-dimensional data into a low-dimensional space, so that the similarities or differences between samples can be intuitively displayed in two-dimensional or three-dimensional space.

[0041] The results are as follows Figure 3 As shown in Figure 2, there was a significant separation between the gallstone group and the healthy control group in terms of overall community structure (p < 0.05) (p ANOSIM <0.01).

[0042] The p-value is an important parameter used in statistics to judge the results of hypothesis testing. It indicates the probability of observing a more extreme result than the current sample result when the hypothesis is true. In this embodiment, the p-value is used to measure whether there is a significant difference in the overall community structure between the gallstone group and the healthy control group.

[0043] p ANOSIM It is an important statistic in ANOSIM (Analysis of Similarities) analysis, which is used to determine the significance of differences between groups. ANOSIM <0.01, it can be seen that there is a very significant separation in the microbial community structure between the gallstone group and the healthy control group, which is of great significance for understanding the microbiome mechanism of gallstones.

[0044] S23 performed a difference test on the oral flora at the genus level between the gallstone patient group samples and the healthy control group.

[0045] First, we excluded taxa that lacked taxonomic information starting from the phylum level in the database. Figure 4 As shown, there were significant differences in 10 bacterial genera between the two groups (p < 0.05). Compared with the healthy control group, the relative abundance of 7 bacterial genera such as Rothia and F0058 in the gallstone group was higher, while the relative abundance of Gemella, Leuconostoc and Blastococcus was lower in the gallstone group.

[0046] According to the results of the above three groups of comparisons, it can be seen that there is a considerable correlation between oral flora and gallstones. Therefore, oral flora can be used as an indicator to predict the risk of gallstones.

[0047] S3 selects several oral bacteria species as input feature species of the prediction model.

[0048] It is foreseeable that there are many types of bacteria in the oral cavity, but their importance to the prediction model varies. The error rate and AUC value of the model constructed by inputting different types and numbers of bacteria are different. Therefore, the purpose of this step is to screen out the oral bacteria species that can best predict gallstones as the input feature species of the prediction target. In this embodiment, the prediction model is a random forest model, which can use the collected samples to predict the risk of gallstones and output the importance score of the oral bacteria species. The mathematical expression of the random forest model is: ; Among them, I(.) is the indicator function, and its value is 1 when the condition in the brackets is met, otherwise it is 0; hm(x) is the decision tree function with sequence number m in the random forest rule; k is the predicted value; H(x) is the random forest model function; and n is the number of decision trees.

[0049] The random forest model is an ensemble learning method that improves the accuracy and stability of predictions by building multiple decision trees and combining their prediction results. The random forest algorithm combines the simplicity of decision trees with the powerful ability of ensemble learning and is widely used in tasks such as classification and regression. Its construction process is: 1) Random sampling: Random forest randomly selects multiple subsets from the training data set through sampling with replacement (Bootstrap), and these subsets are used to build different decision trees.

[0050] 2) Random selection of features: When building each decision tree, the random forest algorithm also randomly selects a portion of features (rather than all features) for node division, which further increases the diversity and generalization ability of the model.

[0051] 3) Building multiple trees: Random forests consist of multiple decision trees, each of which is independently built on a different subset of data and features.

[0052] 4) Prediction results: For classification tasks, random forest determines the final classification result by majority voting; for regression tasks, random forest obtains the final regression result by averaging the prediction values ​​of each tree.

[0053] The final prediction result of the random forest model for the sample is 1 or 0, where 1 means there is a risk of gallstones and 0 means there is no risk of gallstones. Another reason why the present invention selects the random forest model as the prediction model is that the random forest model provides a variety of methods to calculate and output feature importance scores, which can help users understand which features contribute most to the prediction results of the model.

[0054] Based on the prediction results of the random forest model, the specific steps for screening oral bacteria species include: S31 uses the cross-validation method and the area under the curve validation method to verify the prediction results of the prediction model, and obtains the error rate and the area under the curve value of the prediction model respectively.

[0055] More specifically, the specific steps of the cross-validation method are: 1) Divide the collected samples into several equal-sized sub-datasets; for example, the ten-fold cross validation used in this embodiment divides the samples into ten equal-sized sub-datasets. The data set D contains N=112 samples. The data set D is divided into 10 subsets D1, D2, ..., D10 of roughly equal size. The number of samples in each N subset is , that is, N=11 (round down).

[0056] 2) Use a sub-dataset as the test set and the remaining sub-datasets as the training set to cross-validate the random forest model and calculate the single validation error rate; repeat this step until all sub-datasets are used as test sets. For example, for the i-th cross-validation (i = 1, 2, ..., 10), the training set By except All subsets except i, Right now , test set .

[0057] 3) Use the single validation error rate to calculate the error rate of the random forest model.

[0058] The number of samples that the model mispredicted in the i-th cross validation is ,in is an indicator function. When the condition in the brackets is met, its value is 1 (suffering from gallstones), otherwise its value is 0 (not suffering from gallstones). Then the error rate of the i-th cross validation is , Is the test set The average error rate after ten-fold cross validation is , j = 10. Among them, ErrorRate i is the single validation error rate of the i-th cross validation; m i is the number of samples with prediction errors; For the test set The number of samples; j is the number of sub-datasets. In this embodiment, the results after cross-validation are as follows: Figure 5 shown.

[0059] Area Under Curve Validation Method Area Under Curve (AUC) validation method is a commonly used model performance evaluation method, mainly used to evaluate the classification ability of classification models, especially in binary classification problems. AUC is the area under the ROC curve (Receiver Operating Characteristic Curve). The ROC curve is obtained by plotting the relationship between the True Positive Rate (TPR, also known as sensitivity) and the False Positive Rate (FPR, also known as 1-specificity). Specifically, its formula is: ; Calculate the area under the curve; AUC is the area under the curve; TPR is the true positive rate, and FPR is the false positive rate.

[0060] In this embodiment, the results after AUC verification are as follows: Figure 6 shown.

[0061] S32 sorts the oral bacteria species from large to small according to the importance score, and selects several oral bacteria species ranked higher than a threshold as input feature species of the prediction model based on the error rate and the area under the curve value.

[0062] Combining the results of cross-validation and ACU validation, it can be seen that using the top 28 oral bacterial genera for prediction not only has a low error rate, but also a high AUC value. In this example, the genera with "Unclassfied" in their names (meaning that no corresponding classification information can be found in the database) were removed, and finally 21 genera were selected as the input feature types of the model. The results are as follows Figure 7 shown.

[0063] S4 used the collected samples to verify the accuracy of the prediction model through ROC (receiver operating characteristic) analysis.

[0064] After determining the input features, in order to verify the performance of the model, you can also use ROC analysis to verify the accuracy of the model. The specific steps are as follows: 70% of the gallstone patient group samples and healthy control group samples (24 samples in the gallstone patient group and 55 samples in the healthy control group) were randomly selected as the training set, and the remaining 30% were used as the test set. After ROC analysis, the AUC value of the training set was 0.73 (95% CI: 0.60-0.86). Figure 8 The AUC value of the test set was 0.76 (95% CI: 0.59-0.93, as shown in Figure 2). Fig. 9 The validation results show that this model has certain diagnostic significance.

[0065] S5 inputs the feature value corresponding to the input feature type of the subject into the random forest model to predict the risk of gallstones of the subject.

[0066] After the prediction model is constructed, a saliva sample of the subject is collected and 16S rRNA gene sequencing is performed to obtain the species and abundance of each oral bacteria genus in the saliva sample. The abundance of the 21 oral bacteria genus species determined in step S3 is selected and input into the prediction model to obtain the prediction result.

[0067] Example 2 The present invention also provides a gallstone disease risk prediction system based on oral flora, which is characterized by comprising: A sample collection module, used to collect samples, wherein the samples include oral bacteria species and their abundance in a gallstone patient group and a healthy control group; A correlation determination module is used to compare the samples of the gallstone patient group with the healthy control group to determine the correlation between the oral flora and the risk of gallstones; A feature type selection module is used to select several oral bacteria species as input feature types of the prediction model; The prediction module is used to input the feature values ​​corresponding to the input feature types of the test subject into the random forest model to predict the risk of gallstones in the test subject.

[0068] The above are only preferred embodiments of the present invention. It should be noted that the above preferred embodiments should not be regarded as limiting the present invention, and the protection scope of the present invention should be based on the scope defined by the claims. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the spirit and scope of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for predicting the risk of gallstones based on oral flora, characterized in that include: Collecting samples, wherein the samples include oral bacteria species and their abundance in a gallstone patient group and a healthy control group; Compare the gallstone patient group samples with the healthy control group samples to determine the association between oral flora and gallstone risk; Select several oral bacteria species as input feature species for the prediction model; The feature values ​​corresponding to the input feature types of the subject are input into the random forest model to predict the risk of gallstones in the subject.

2. A method for predicting the risk of gallstones based on oral flora according to claim 1, characterized in that Methods for selecting several oral bacteria species in the oral flora as input feature species of the prediction model include: The prediction model is a random forest model, which can use the collected samples to predict the risk of gallstones and output the importance score of oral bacteria species; The prediction results of the prediction model are verified by using the cross-validation method and the area under the curve validation method, and the error rate and area under the curve value of the prediction model are obtained respectively; Oral bacteria species were ranked from high to low according to their importance scores, and several oral bacteria species ranked higher than a threshold were selected as input feature species for the prediction model based on the error rate and the area under the curve.

3. A method for predicting the risk of gallstones based on oral flora according to claim 2, characterized in that The random forest model is: ； Among them, I(.) is the indicator function, and its value is 1 when the condition in the brackets is met, otherwise it is 0; hm(x) is the decision tree function with sequence number m in the random forest rule; k is the predicted value; H(x) is the random forest model function; and n is the number of decision trees.

4. A method for predicting the risk of gallstones based on oral flora according to claim 2, characterized in that The steps of the cross validation method include: Divide the collected samples into several sub-datasets of equal size; Use a sub-dataset as the test set and the remaining sub-datasets as the training set to cross-validate the random forest model and calculate the single validation error rate; repeat this step until all sub-datasets are used as test sets; The error rate of the random forest model is calculated using the single-shot validation error rate.

5. A method for predicting the risk of gallstones based on oral flora according to claim 4, characterized in that Using the formula: ； Calculate the single verification error rate; where ErrorRate i is the single validation error rate of the i-th cross validation; m i is the number of samples with prediction errors; For the test set The number of samples; Using the formula: ； Calculate the error rate of the random forest model; where ErrorRate cv ErrorRate i is the single validation error rate of the i-th cross validation; j is the number of sub-datasets.

6. A method for predicting the risk of gallstones based on oral flora according to claim 2, characterized in that The area under the curve validation method uses the formula: ； Calculate the area under the curve; AUC is the area under the curve; TPR is the true positive rate, and FPR is the false positive rate.

7. The method for predicting the risk of gallstones based on oral flora according to claim 2, characterized in that: When collecting samples, saliva samples were collected from the gallstone patient group and the healthy control group, and the types and abundance of oral bacteria in the samples were obtained by 16S rRNA gene sequencing.

8. The method for predicting the risk of gallstones based on oral flora according to claim 1, characterized in that Methods to determine the association between oral flora and gallstone risk include: The Ace index, Chao1 index, Shannon index, and Simpson index based on OTU levels were used to test the differences between the gallstone patient group and the healthy control group. Compare the results of several distance algorithms based on their contribution to sample differences; Differences in oral flora at the genus level were tested between samples from the gallstone patient group and the healthy control group.

9. The method for predicting the risk of gallstones based on oral flora according to claim 1, characterized in that: After selecting the types of input features, the accuracy of the prediction model was verified using the collected samples through receiver operating characteristic analysis.

10. A gallstone risk prediction system based on oral flora, characterized in that include: A sample collection module, used to collect samples, wherein the samples include oral bacteria species and their abundance in a gallstone patient group and a healthy control group; A correlation determination module is used to compare the samples of the gallstone patient group with the healthy control group to determine the correlation between the oral flora and the risk of gallstones; A feature type selection module is used to select several oral bacteria species as input feature types of the prediction model; The prediction module is used to input the feature values ​​corresponding to the input feature types of the test subject into the random forest model to predict the risk of gallstones in the test subject.

Images