Lung Cancer Risk Early Warning Method Based on Single Nucleotide Polymorphism and Artificial Intelligence

Abstract

This invention provides a lung cancer risk warning method based on single nucleotide polymorphisms and artificial intelligence, characterized by the following steps: Step S1, detecting corresponding site features of patients based on multiple preset SNP sites; Step S2, obtaining the patient's epidemiological information, which includes at least smoking characteristics, gender characteristics, age characteristics, and family history characteristics; Step S3, based on the site features and epidemiological information, providing a lung cancer risk warning for the patient through a predetermined lung cancer risk warning model, wherein the lung cancer risk warning model is a susceptibility model established by combining multiple SNP sites and epidemiological information through logistic regression and extreme gradient boosting algorithms.

Description

Technical Field

[0001] This invention belongs to the field of medical technology, specifically relating to a lung cancer risk early warning method based on single nucleotide polymorphism and artificial intelligence. Background Technology

[0002] Early detection of lung cancer is crucial for reducing its mortality rate. A common approach is to construct lung cancer prediction models to classify the population by risk, and then conduct universal screening of high-risk groups to predict lung cancer. However, existing models based on germline variation have performed poorly, necessitating the development of new models.

[0003] Currently, some studies have incorporated SNPs into population-based modeling. For example, some models use only SNP sites for prediction, while others combine smoking with logistic regression. However, the predictive performance of these models remains insufficient, with an AUC < 0.7. The area under the receiver operating curve (AUC) is used to assess the presence of SNPs and the contribution of machine learning methods to risk prediction models. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence, using extreme gradient enhancement to establish early lung cancer diagnosis. The technical solution adopted in this invention is as follows:

[0005] This invention provides a lung cancer risk warning method based on single nucleotide polymorphisms and artificial intelligence, characterized by the following steps: Step S1, detecting corresponding site features of patients based on multiple preset SNP sites; Step S2, obtaining the patient's epidemiological information, which includes at least smoking characteristics, gender characteristics, age characteristics, and family history characteristics; Step S3, based on the site features and epidemiological information, providing a lung cancer risk warning for the patient through a predetermined lung cancer risk warning model, wherein the lung cancer risk warning model is a susceptibility model established by combining multiple SNP sites and epidemiological information through logistic regression and extreme gradient boosting algorithms.

[0006] The lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence provided by the present invention may also have the following technical features, wherein the smoking feature includes at least the smoking intensity.

[0007] The lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence provided by this invention may also have the following technical features, wherein the preset multiple SNP sites are SNP sites with a detection rate >95%, Hardy-Weinberg equilibrium P >0.01 in the control group, and minor allele frequency >0.01.

[0008] The lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence provided by this invention may also have the following technical features, wherein there are 61 SNP sites.

[0009] Invention Function and Effect

[0010] The lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence according to the present invention utilizes 61 SNP loci, along with epidemiological information such as smoking characteristics, gender characteristics, age characteristics, and family history characteristics, and generates a lung cancer risk warning model based on logistic regression and extreme gradient enhancement. Compared with traditional lung cancer risk warning models, the model of the present invention selects more loci, has a wider coverage, and considers more epidemiological factors. Furthermore, the model is optimized based on extreme gradient enhancement, thus the predictive performance of this model is better than that of traditional models, and the prediction accuracy is significantly improved. Attached Figure Description

[0011] Figure 1 This is a flowchart of the lung cancer risk warning method in an embodiment of the present invention;

[0012] Figure 2 This is one of the schematic diagrams comparing the effects of various models designed in the embodiments of the present invention;

[0013] Figure 3 This is the second schematic diagram comparing the effects of various models designed in the embodiments of the present invention. Detailed Implementation

[0014] The specific embodiments of the present invention will be described below with reference to the accompanying drawings and examples.

[0015] <Example>

[0016] In this embodiment, all tests were conducted based on the epidemiological characteristics of 974 lung cancer patients and 1005 controls. Among them, male patients (71.1%) outnumbered female patients. In all lung cancer patients, ADC accounted for 48.8% and SCC for 36.5%.

[0017] Figure 1 This is a flowchart of the lung cancer risk warning method in an embodiment of the present invention.

[0018] like Figure 1 As shown, the lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence specifically includes the following steps:

[0019] Step S1: Detect the patient based on multiple preset SNP sites to obtain the corresponding site features.

[0020] In this embodiment, 61 SNP loci were selected from lung cancer-related loci in previous genome-wide association studies (GWAS) and published articles. Successful SNP genotyping criteria were: detection rate >95%, Hardy-Weinberg equilibrium (HWE) P >0.01 in the control group, and minor allele frequency (MAF) >0.01.

[0021] Step S2: Obtain the patient's epidemiological information, which includes at least smoking characteristics, gender characteristics, age characteristics, and family history characteristics.

[0022] In this embodiment, smoking characteristics include smoking intensity and smoking duration. According to the inventors' experiments, smoking intensity is the most important factor in the model, followed by smoking duration, age, and sex. Among the 17 SNPs included in the lung cancer model, BAG6rs1077393, TERT rs2735845, and CAMKK1 rs7214723 are more important than family history.

[0023] Step S3: Based on site characteristics and epidemiological information, a predetermined lung cancer risk warning model is used to provide an early warning of the patient's lung cancer risk.

[0024] In this embodiment, based on 61 SNP loci, epidemiological information such as smoking, gender, age, and family history was combined to construct a susceptibility model as a lung cancer risk warning model by using logistic regression (LR) and extreme gradient boosting algorithm (XGBoost) to stratify lung cancer, lung adenocarcinoma (ADC), and lung squamous cell carcinoma (SCC).

[0025] Figure 2 This is one of the schematic diagrams comparing the effects of various models designed in the embodiments of the present invention.

[0026] The two main subtypes of lung cancer, ADC and SCC, have different mutational characteristics. Therefore, this embodiment tested the risks of lung cancer, ADC, and SCC separately. Figure 2As shown, for lung cancer risk prediction, the LR AUC of the epidemiological model alone (excluding SNPs) was 0.703, while the AUC of XGBoost was 0.744. Compared with the epidemiological LR model alone, further adding SNPs and applying XGBoost increased the AUC of the XGBoost model to 0.759 (P<0.001). In the XGBoost model for lung cancer prediction, BAG6rs1077393 was the most important predictor among all SNPs, followed by TERT rs2735845 and CAMKK1rs7214723. When XGBoost was applied and SNPs were added to the model, further stratification of lung adenocarcinoma (ADC) showed a significant improvement from 0.639 to 0.699 (P=0.009), while the best model for predicting lung squamous cell carcinoma (SCC) was the complete LR-based model (AUC=0.833).

[0027] In both overall lung cancer and different subtypes of lung cancer, the AUC of the lung cancer risk warning model in this embodiment reached above 0.7, with the highest reaching 0.833.

[0028] Figure 3 This is the second schematic diagram comparing the effects of various models designed in the embodiments of the present invention.

[0029] This embodiment also stratifies based on gender, age, smoking status, and family history. For example... Figure 3 As shown, lung cancer risk prediction performed well in men (AUC = 0.791), those aged 60 and above (AUC = 0.761), smokers (AUC = 0.785), those without a family history (AUC = 0.734), and those with a family history (AUC = 0.790). Specifically, for some subgroups (men, those aged 60 and above, smokers, and those with a family history), the SNP-based full model constructed by XGBoost was the best approach, while for other subgroups (women, those under 60 years old, non-smokers, and those without a family history), logistic regression (LR) was the best approach. Before and after adding SNPs, all stratified models showed a trend of improved model performance in the LR model, with some models showing significant improvements (women and non-smokers).

[0030] All tests were two-tailed, and the statistical significance was set at p < 0.05. All analyses were performed using Rv3.6.2.

[0031] Functions and effects of the embodiments

[0032] The lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence provided in this embodiment utilizes 61 SNP loci, along with epidemiological information such as smoking characteristics, gender characteristics, age characteristics, and family history characteristics, and generates a lung cancer risk warning model based on logistic regression and extreme gradient enhancement. Compared to traditional lung cancer risk warning models, the model in this embodiment selects more loci, has a wider coverage, and considers more epidemiological factors. Furthermore, the model is optimized based on extreme gradient enhancement, resulting in better predictive performance and significantly improved prediction accuracy compared to traditional models.

[0033] The above embodiments are only used to illustrate specific implementations of the present invention, and the present invention is not limited to the scope of the description of the above embodiments.

Claims

1. A lung cancer risk early warning method based on single nucleotide polymorphism and artificial intelligence, characterized in that, The method comprises the following steps: Step S1, detecting a patient according to a plurality of SNP sites to obtain corresponding site characteristics; Step S2, obtaining epidemiological information of the patient, the epidemiological information at least including smoking characteristics, gender characteristics, age characteristics and family history characteristics; Step S3, based on the site characteristics and the epidemiological information, warning the lung cancer risk of the patient through a predetermined lung cancer risk warning model, Wherein the lung cancer risk warning model is a susceptibility model established by combining the plurality of SNP sites and the epidemiological information through a logic regression and an extreme gradient boosting algorithm for stratification of lung cancer, lung adenocarcinoma and lung squamous cell carcinoma, The SNP site is 61, and the SNP site is an SNP site with a detection rate > 95%, a control group Hardy-Weinberg balance P > 0.01 and a minor allele frequency > 0.

01.

2. The lung cancer risk warning method based on single nucleotide polymorphism and artificial intelligence according to claim 1, wherein: wherein The smoking characteristics at least include smoking intensity and smoking time, For the lung cancer risk warning model, the smoking intensity is the most important factor, followed by the smoking time, and among the SNP sites, BAG6 rs1077393, TERT rs2735845 and CAMKK1 rs7214723 are more important than the family history characteristics.

Images