We establish brain connectivity features that represented the disease signatures and identify Parkinson’s disease (PD) subtypes by data-driven approaches. Canonical correlation analysis (CCA) was performed to define the clinical related connectivity features, which were then used in hierarchical cluster analysis to identify the distinct biotypes of PD. Multimodal MRI including gray matter functional connectivity and white matter microstructure were further used to explore the neuropsychological significance of these biotypes. CCA revealed two significant clinical-related patterns in PD. Hierarchical cluster analysis identified three neurophysiological biotypes: mild, progressive depression-dominant and progressive motor-dominant. These three biotypes characterized by different neural substrate.