Parkinson’s Disease (PD) is a progressive neurodegenerative disorder characterized by motor dysfunction and, in many cases, cognitive impairment. While substantial progress has been made in understanding PD’s clinical manifestations, distinguishing PD patients from healthy controls and identifying cognitive impairment subtypes remains challenging. In this study, we present a Graph Neural Network (GNN) framework that leverages both structural anatomical data (SA) and functional connectivity (FC) to enhance the classification of PD patients and the stratification of cognitive impairment subtypes. Our approach includes an ablation study to evaluate the individual and combined contributions of SA and FC in distinguishing between PD patients and healthy controls (HC) and between PD patients with normal cognition (PD-NC) and those with mild cognitive impairment (PD-MCI). Experimental results show that the GNN-based approach reveals the distinct roles of anatomical and functional connectivity in disease phenotypes.