Machine learning study on predicting depressive symptoms and genetic correlations in Parkinson's disease

Depressive symptoms are prevalent in individuals with Parkinson's disease. Previous research has demonstrated a significant association between the triglyceride glucose (TyG) index and depression. Leveraging multicenter clinical data, the present study evaluates the predictive capacity of the TyG index for depressive symptoms in PD patients, aiming to establish its potential role in identifying individuals at risk for depression. A comparative analysis of multiple machine learning models was conducted to predict depression in PD patients, ultimately selecting the most effective model. Key predictive variables, including diabetes status, sex, cholesterol levels, triglycerides, blood glucose, and sleep disturbances, were incorporated into a support vector machine (SVM)-based nomogram to assess depression risk in PD patients. Additionally, a genome-wide association study (GWAS) utilizing external databases confirmed a causal relationship between the TyG index and depression. Furthermore, this study explores the biological functions and molecular mechanisms underlying shared transcriptomic proteins between PD and depression, providing insights into potential pathophysiological links between the two conditions.