Zi-Tong Guo,1 Xiao-Lin Yu,2 Hui Cheng,2 Tuersunjiang Naman2 

1Department of Cardiology, First Affiliated Hospital of Xinjiang Medical University, Urumqi, Xinjiang, People’s Republic of China; 2Department of Cardiology, People’s Hospital of Xinjiang Uygur Autonomous Region, Urumqi, Xinjiang, People’s Republic of China

Correspondence: Tuersunjiang Naman, Email tursunjan1016@163.com

Objective: This study investigates the relationship between serum γKlotho levels and coronary heart disease (CHD) risk and develops a machine learning model for CHD prediction.
Methods: A total of 1435 subjects were enrolled for analysis and randomized as training (n =  969, 70%) or validation (n =  466, 30%) group. The training group was used for univariate regression. Thereafter, least absolute shrinkage and selection operator (LASSO) regression was conducted for selecting independent risk factors for CHD. Using independent risk factors for CHD, nine machine learning models were developed, the best model was selected by evaluating them, and the model was validated by decision curve analysis (DCA).
Results: The factors independently associated with CHD risk were age, the serum level of γKlotho, LDL-C, sex, diabetes, hypertension, and smoking status. We used these risk factors to construct nine popular machine-learning models. Among all models, the RF model was better appropriate; thus, we visualized and validated this model, which showed promising clinical application.
Conclusion: Serum γKlotho levels are novel biomarker which positively related to CHD risk. Additionally, the RF model can better predict the risk of CHD, and RF model is better appropriate to predicting the CHD risk in clinics.

Keywords: coronary heart disease, γKlotho, random forest, RF, prediction model