Method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics

Abstract

The present invention relates to the technical field of fracture combined with deep vein thrombosis prediction, and in particular to a method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics, comprising the following steps: P1, screening out a DVT genotype SNP feature set; P2, establishing a fracture combined with DVT prediction model and a model evaluation index; P3, optimizing the fracture combined with DVT prediction model, and determining the optimal feature SNP subset of the fracture combined with DVT prediction model; P4, extracting the optimal DVT genotype SNP feature set from the genotype data of fracture patients whose DVT is unknown according to the optimal feature SNP subset, comparing the DVT genotype SNP feature set with the optimal DVT genotype SNP feature set, and predicting the risk value of whether the fracture patient has DVT. The present invention combines genomic DNA sequence information with a machine learning method, and uses a random forest algorithm in ensemble learning to further extract genomic sequence features, and predict the risk of DVT in fracture patients, thereby providing accurate clinical guidance for the subsequent treatment and medication of the patients.

Description

Technical Field

[0001] The present invention relates to the technical field of prediction of fracture combined with deep vein thrombosis, and specifically to a method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics. Background Art

[0002] Deep Vein Thrombosis (DVT) is a common cardiovascular disease affected by genetic factors. It refers to the coagulation of venous blood in the deep veins of the lower extremities, mainly manifested as the formation of blood clots in the deep veins of the muscle tissue of the legs, pelvis or abdomen. The important harm of DVT to health is that in addition to impairing venous blood flow, causing swelling and pain in the legs, ankles and feet, it can also cause complications such as pulmonary embolism (PE) that can cause death. Clinical data statistics show that the proportion of patients with DVT after trauma, fractures and surgical operations is as high as 50%-60%. Pulmonary embolism is the most serious and common complication of DVT, with a short-term survival rate of less than 60%, which has seriously threatened people's life activities and quality of life.

[0003] At present, clinical DVT is mainly confirmed by Doppler vascular ultrasound and deep venography of the limbs, but it takes a long time and delays precious rescue time. Moreover, DVT is easily confused with trauma, congestion, and edema after fracture, which can easily cause delayed treatment. In addition, related studies have found multiple risk factors for thrombosis and established a risk assessment system based on this. These research results provide a certain reference for the prevention and treatment of clinical DVT, but there are also deficiencies and limitations. At present, the commonly used thrombosis scoring standard in clinical practice is the Caprini thrombosis risk assessment table, which contains about 40 different risk factors for thrombosis, but this assessment system relies too much on the clinical experience of doctors and cannot objectively and effectively predict the risk of lower limb DVT after fracture, and cannot accurately guide the treatment of patients, thereby increasing the risk of postoperative bleeding.

[0004] Studies have found that the occurrence of DVT is regulated by multiple genes, and its heritability is about 60%, indicating that genetic factors play an important role in the development of DVT. So far, the genome-wide association study (GWAS) for thrombosis-related phenotypes has found more than 30 disease-susceptible single nucleotide polymorphism (SNP) sites, but the specific relationship between these susceptibility sites and DVT is very complex, and it is impossible to simply use a combination of several SNPs to accurately assess and predict the risk of DVT. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics, comprising the following steps:

[0008] P1, collect known DVT susceptibility SNPs, screen out characteristic SNP genotype sets, and establish a sample feature matrix based on the characteristic SNP genotype sets;

[0009] P2, based on the sample feature matrix, the random forest algorithm was used to establish a fracture combined with DVT prediction model and model evaluation indicators;

[0010] P3, optimize the fracture combined with DVT prediction model according to the model evaluation index, and determine the optimal characteristic SNP subset of the fracture combined with DVT prediction model;

[0011] P4, based on the optimal characteristic SNP subset, extract the genotype of the optimal SNP feature set from the genotype data of fracture patients whose DVT is unknown. Based on the genotype of the optimal SNP set, use the established fracture combined with DVT model to predict the risk value of DVT in fracture patients whose DVT is unknown.

[0012] Preferably, in P1, the specific steps are:

[0013] P11, a collection of known susceptibility SNPs for DVT was collected from the public database CAUSALdb and the latest large-scale DVT GWAS studies;

[0014] P12, based on the genotype data published by the 1000 Genomes Project, calculate the allele frequency of each susceptible SNP in the known susceptible SNP set in P11 in Asians and Europeans, and use statistical methods to screen out all SNPs suitable for Asians from the known susceptible SNP set of DVT and form a characteristic SNP set;

[0015] P13, according to the characteristic SNP set, all characteristic SNP genotype data are extracted from the genotype data of all fracture patients whose DVT is known, and then the characteristic SNP genotype data of all samples are merged into a characteristic SNP genotype matrix, that is, the sample characteristic matrix; the genotype data of the known fracture patients are obtained using gene chips.

[0016] Preferably, Fisher's exact test is used when screening the characteristic SNP set.

[0017] Preferably, the process of establishing the prediction model for fracture combined with DVT is as follows:

[0018] P21, data classification: fracture patients were divided into positive and negative sets according to whether DVT occurred after fracture. Fracture patients with DVT were regarded as positive set, and fracture patients without DVT were regarded as negative set. The positive and negative sets were randomly divided into independent training and test sets by stratified sampling method, and the training set feature matrix and test set feature matrix were obtained according to the sample feature matrix.

[0019] P22, a prediction model for fracture combined with DVT was established using the random forest algorithm for the training set feature matrix; the prediction model for fracture combined with DVT was evaluated using the test set special matrix.

[0020] Preferably, in P21, the training set accounts for 90% of the DVT genotype SNP feature set, and the test set accounts for 10% of the DVT genotype SNP feature set.

[0021] Preferably, the specific steps of P22 are: first, generate a decision tree based on the training set, randomly and non-repetitively select d features at each node of the decision tree, and use the d features to divide the training set respectively to obtain the division features; repeat the above steps k times, k is the number of decision trees in the fracture combined with DVT prediction model; finally, use the test set to evaluate the fracture combined with DVT prediction model, and use the voting method to obtain the model evaluation index.

[0022] Preferably, a voting method is used when obtaining model evaluation indicators.

[0023] Preferably, model evaluation indicators include sensitivity, specificity, precision, accuracy, F1 score and AUC value, etc.

[0024] Preferably, in P3, a feature selection method is used during optimization.

[0025] Preferably, in P3, the steps for obtaining the optimal characteristic SNP subset are: using the Gini index to evaluate the feature importance of the fracture combined with DVT prediction model, and constructing multiple feature subsets according to the feature importance as the number of features increases from 1 to the maximum, and determining the optimal characteristic SNP subset according to the model evaluation index; when optimizing the fracture combined with DVT prediction model, the fracture combined with DVT prediction model derives a risk threshold based on the training data, and when the predicted risk value is greater than or equal to the risk threshold, the fracture patient has a high risk of DVT.

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

[0027] The present invention provides a method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics. The method combines genomic DNA sequence information with a machine learning method, extracts genomic sequence features using a random forest algorithm, predicts the risk of deep vein thrombosis after fracture, obtains a risk value, and uses the risk value as a reference indicator for the possibility of DVT occurrence in fracture patients, thereby providing accurate clinical guidance for the patient's subsequent treatment and medication.

[0028] The present invention is a method for predicting fracture combined with deep vein thrombosis based on the genetic characteristics of fracture patients, and is widely applicable to the prediction of fracture combined with deep vein thrombosis in the Chinese Han population. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The present invention is a flow chart of a method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics. DETAILED DESCRIPTION

[0030] The present invention is further described in detail below in conjunction with specific embodiments, which are intended to explain the present invention rather than to limit it.

[0031] The present invention discloses a method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics, referring to Figure 1 , including the following steps:

[0032] P1, based on the DVT susceptibility characteristic SNP set, the DVT genotype SNP feature set is screened from the genotype data of known fracture patients. The specific steps are:

[0033] P11, a collection of known susceptibility SNPs for DVT was collected from the public database CAUSALdb and the latest large-scale DVT GWAS studies;

[0034] P12, based on the genotype data published by the 1000 Genomes Project, the allele frequency of each susceptible SNP in the known susceptible SNP set in P11 in Asians and Europeans was calculated, and the Fisher's exact test method in the statistical method was used to screen out all SNPs suitable for Asians from the known susceptible SNP set of DVT and form a characteristic SNP set;

[0035] P13, according to the characteristic SNP set, all characteristic SNP genotype data are extracted from the genotype data of all fracture patients whose DVT is known, and then the characteristic SNP genotype data of all samples are merged into a characteristic SNP genotype matrix, that is, the sample characteristic matrix; the genotype data of the known fracture patients are obtained using gene chips.

[0036] P2, based on the sample feature matrix, the random forest algorithm was used to establish a fracture combined with DVT prediction model and model evaluation indicators.

[0037] P21, data classification: fracture patients were divided into positive and negative sets according to whether DVT occurred after fracture. Fracture patients with DVT were regarded as positive set, and fracture patients without DVT were regarded as negative set. The positive set and negative set were randomly divided into independent training set and test set by stratified sampling method, and the training set feature matrix and test set feature matrix were obtained according to the sample feature matrix. Among them, the training set accounted for 90% of the DVT genotype SNP feature set, and the test set accounted for 10% of the DVT genotype SNP feature set.

[0038] P22, the prediction model of fracture combined with DVT was established using the random forest algorithm for the training set feature matrix; the prediction model of fracture combined with DVT was evaluated using the test set special matrix. The specific steps are as follows:

[0039] First, a decision tree was generated based on the training set. At each node of the decision tree, d features were randomly and non-repetitively selected, and the training set was divided using the d features to obtain the division features. The above steps were repeated k times, where k was the number of decision trees in the fracture combined with DVT prediction model. Finally, the fracture combined with DVT prediction model was evaluated using the test set, and the model evaluation index was obtained by the voting method.

[0040] Model evaluation indicators include sensitivity, specificity, precision, accuracy, F1 score and AUC value. The calculation formulas for each model evaluation indicator are as follows:

[0041]

[0042] Sensitivity=TP / P=TP / (TP+FN)

[0043] Specificity=TN / N=TN / (TN+FP)

[0044] Precision=TP / P'=TP / (TP+FP)

[0045] Accuracy=(TP+TN) / (P+N)

[0046] F1=2×TP / (2×TP+FP+FN)

[0047] P3, based on the model evaluation index, the feature selection method is used to optimize the fracture combined with DVT prediction model and determine the optimal characteristic SNP subset of the fracture combined with DVT prediction model. The specific steps are:

[0048] The Gini index was used to evaluate the feature importance of random forests, and multiple feature subsets were constructed according to the feature importance as the number of features increased from 1 to the maximum. The optimal feature SNP subset was determined based on the model evaluation index.

[0049] P4, based on the optimal characteristic SNP subset, extract the genotype of the optimal SNP feature set from the genotype data of fracture patients whose DVT is unknown. Based on the genotype of the optimal SNP set, use the established fracture combined with DVT model to predict the risk value of DVT in fracture patients whose DVT is unknown.

[0050] When optimizing the fracture combined with DVT prediction model, the fracture combined with DVT prediction model derives a risk threshold based on training data. When the predicted risk value is greater than or equal to the risk threshold, the fracture patient has a higher risk of DVT.

[0051] Example

[0052] A sample of 1000 Chinese Han fractures (500 with DVT and 500 with non-DVT) was taken as an example.

[0053] P1, collected and obtained known susceptibility SNPs for DVT from the public database CAUSALdb and the latest large-scale DVT GWAS study, and then further screened out the SNP feature set whose allele frequencies were suitable for Asians based on the genotype data published by the 1000 Genomes Project, a total of 175 SNPs, as the feature SNP set for building the model.

[0054] The Illumina ASA-24v1 gene chip was used to obtain the genotype data of 1000 fracture patients, and the genotype data of each characteristic SNP in the characteristic SNP set were extracted.

[0055] P2, using the stratified sampling method, 1000 fracture patients were randomly divided into training set and test set, where the number of training set samples accounted for 90% (900) of the total number of samples, and the number of test set samples accounted for 10% (100) of the total number of samples. The model was trained using the random forest algorithm, and the model was evaluated using the test set feature matrix.

[0056] P3, the Gini index was used to evaluate the feature importance of random forest and feature selection was performed according to the feature importance ranking. Finally, the model established using the 45 optimal feature SNP subsets performed best, and all indicators during the evaluation were not less than 0.75.

[0057] P4, extract the genotypes of 45 optimal characteristic SNPs from the genotype data of fracture patients whose DVT is unknown, and use the established fracture combined with DVT model to predict the risk of DVT in fracture patients whose DVT is unknown. Those fracture patients whose risk exceeds the model threshold are at high risk of DVT.

[0058] Experimental results: For Chinese Han samples, the fracture combined with venous thrombosis prediction model established by this method had various indicators in the test data that were not less than 0.75.

[0059] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. However, the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and 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 fracture combined with deep vein thrombosis based on genetic characteristics, characterized in that: The following steps are involved: P1, collect known DVT susceptibility SNPs, screen out characteristic SNP genotype sets, and establish a sample feature matrix based on the characteristic SNP genotype sets; P2, based on the sample feature matrix, the random forest algorithm was used to establish a fracture combined with DVT prediction model and model evaluation indicators; specifically: P21, data classification: fracture patients were divided into positive and negative sets according to whether DVT occurred after fracture. Fracture patients with DVT were regarded as positive set, and fracture patients without DVT were regarded as negative set. The positive and negative sets were randomly divided into independent training and test sets by stratified sampling method, and the training set feature matrix and test set feature matrix were obtained according to the sample feature matrix. P22, the prediction model of fracture combined with DVT was established using the random forest algorithm for the training set feature matrix; the prediction model of fracture combined with DVT was evaluated using the test set special matrix; The specific steps are as follows: first, generate a decision tree based on the training set, randomly select d features without duplication at each node of the decision tree, and use the d features to divide the training set respectively to obtain the division features; Repeat the above steps k times, where k is the number of decision trees in the fracture combined with DVT prediction model; finally, use the test set to evaluate the fracture combined with DVT prediction model, and use the voting method to obtain the model evaluation index; P3, optimize the fracture combined with DVT prediction model according to the model evaluation index, and determine the optimal characteristic SNP subset of the fracture combined with DVT prediction model; Specifically, the Gini index was used to evaluate the feature importance of the fracture combined with DVT prediction model, and multiple feature subsets were constructed according to the feature importance and the number of features increased from 1 to the maximum, and the optimal feature SNP subset was determined according to the model evaluation index; When optimizing the prediction model for fracture combined with DVT, the prediction model for fracture combined with DVT derives a risk threshold based on the training data. When the predicted risk value is greater than or equal to the risk threshold, the fracture patient has a high risk of DVT. P4, based on the optimal characteristic SNP subset, extract the genotype of the optimal SNP feature set from the genotype data of fracture patients whose DVT is unknown. Based on the genotype of the optimal SNP set, use the established fracture combined with DVT model to predict the risk value of DVT in fracture patients whose DVT is unknown.

2. The method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics according to claim 1, characterized in that: In P1, the specific steps are: P11, a collection of known susceptibility SNPs for DVT was collected from the public database CAUSALdb and the latest large-scale DVT GWAS studies; P12, based on the genotype data published by the 1000 Genomes Project, calculate the allele frequency of each susceptible SNP in the known susceptible SNP set in P11 in Asians and Europeans, and use statistical methods to screen out all SNPs suitable for Asians from the known susceptible SNP set of DVT and form a characteristic SNP set; P13, extract all characteristic SNP genotype data from the genotype data of all fracture patients with known DVT according to the characteristic SNP set, and then merge all characteristic SNP genotype data into a characteristic SNP genotype matrix, i.e., the sample characteristic matrix; The genotype data of patients with known fractures were obtained using gene chips.

3. The method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics according to claim 2, characterized in that: Fisher's exact test was used to screen the characteristic SNP set.

4. The method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics according to claim 1, characterized in that: In P21, the training set accounted for 90% of the DVT genotype SNP feature set, and the test set accounted for 10% of the DVT genotype SNP feature set.

5. The method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics according to claim 1, characterized in that: A voting method is used to obtain model evaluation indicators.

6. The method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics according to claim 1, characterized in that: Model evaluation indicators include sensitivity, specificity, precision, accuracy, F1 score and AUC value.

7. The method for predicting the risk of fracture combined with deep vein thrombosis based on genetic characteristics according to claim 1, characterized in that: In P3, feature selection method is used for optimization.

Images