Analysis of the pathogenesis of non-alcoholic fatty liver disease based on Bayesian network model
Objective To identify key influencing factors of non-alcoholic fatty liver disease(NAFLD)and construct a Bayesian network model to explore its pathogenesis.Methods The study was based on the Beijing Health Management Co-hort.The LASSO regression was employed to screen for relevant influencing factors of NAFLD,and 10-fold cross-validation was used to select the optimal penalty parameter λ.Based on the selected variables,structural learning was conducted using the Bayesian network model,and the model fit was evaluated through the Bayesian Information Criterion(BIC)and cross-val-idation for path analysis.Results During the 5-year follow-up period,22.83％of the 30 001 study participants developed NAFLD(n=6 850),with an incidence density of 45.7 per 1 000 person-years.Six key influencing factors were identified:AGE,BMI,HDL-C,FPG,ALT,and DBP,and a Bayesian network model was constructed based on these factors.The Bayesian network analysis revealed 23 pathways for the development of NAFLD,with the pathway having BMI as the root node being the most frequent(12 pathways),followed by age groups(5 pathways).The pathway AGE→DBP→FPG→NAFLD showed a significantly higher incidence probability in individuals under 45 years compared to those over 45.In cases of low BMI,the pathway BMI→HDL→ALT→NAFLD indicated a higher probability of NAFLD compared to other BMI groups.Conclusion The Bayesian network model elucidated the pathways of NAFLD development and clarified the impact of key factors on its onset,providing a theoretical basis for early screening and intervention of NAFLD.
NETL
NSTL
万方数据
