Parkinson’s disease (PD) is a prevalent and nuanced neurodegenerative disease. While the use of neuroimaging data is attractive for PD research, insights from a single imaging modality do not provide a full picture of the disease. Unfortunately, direct comparison across modalities is not feasible. We address this by defining graph representations for each modality as input to a graph convolutional network. We present a pipeline that automates the technical preprocessing procedures and performs the analysis. We hypothesize that incorporating anatomical, structural, functional and volumetric data will yield clinically relevant biomarkers useful for tracking the progression of PD.