Anxiety disorders are one of the most common mental health conditions with a high rate of everyday life disability. Connectivity is steadily gaining relevance to increase our knowledge of psychiatric diseases. Graph signal processing (GSP) is a new framework to integrate structural connectivity and brain function. We propose here a graph-based analysis using GSP metrics and classification procedure, to identify anxiety biomarkers. Results suggest that the joint consideration of structure-function features improves their discriminatory accuracy, and our understanding of the pathophysiology of anxiety.