Recent findings have challenged the prevailing histology- or imaging-based definition of the “vulnerable plaque”. To investigate molecular characteristics associated with “clinical instability” of atherosclerosis, we performed a proteomics comparison of the vascular extracellular matrix and associated molecules in human carotid endarterectomy specimens from symptomatic versus asymptomatic patients. The proteomics data were integrated with gene expression profiling and an analysis of protein secretion by lipid-loaded human vascular smooth muscle cells. The molecular signature of plaques from symptomatic patients identified by proteomics and at least one of the other two approaches comprised matrix metalloproteinase-9, chitinase-3-like protein 1, S100A8/S100A9, cathepsin B, fibronectin and galectin-3-binding protein. These biomarker candidates were measured in 685 subjects of the Bruneck Study and found to significantly predict the progression to advanced atherosclerosis (as assessed by repeated carotid ultrasound) and the incidence of cardiovascular disease over a 10-year follow-up period, highlighting the strength of tissue-based proteomics for biomarker discovery.