Blood serum was used to identify protein biomarkers for diagnosis of Parkinson's disease (PD) using analytically validated quantitative 2D-gel electrophoresis, and single variable and multivariate statistics. Using banked samples from a first medical center, we identified 57 specific protein spot biomarkers with disease-specific abnormal levels in serum of patients with PD, Alzheimer's disease, amyotrophic lateral sclerosis and similar neurodegenerative conditions (337 samples), when compared to age-matched normal controls (132 samples). To further assess their clinical usefulness in Parkinson's disease, we obtained prospective newly drawn blood serum samples from a second (56 PD, 30 controls) and third (6 PD, 48 controls) medical center. The protein concentrations of the 57 biomarkers were assessed by 2D-gel electrophoresis. Stepwise linear discriminant analysis selected a combination of 21 of the 57 as optimal to distinguish PD patients from controls. When applied to the samples from the second site, the 21 proteins had sensitivity of 93.3% (52 of 56 PD correctly classified), specificity of 92.9% (28 of 30 controls correctly classified); 15 of 15 patients with mild, 28 of 30 with moderate to severe symptoms, and all of the 6 PD patients from the third site were correctly classified. Eleven of the 21 proteins showed statistically significant abnormal concentrations in patients with mild symptoms, and 14 in patients with moderate-severe symptoms. The protein identities reflect the heterogeneity of Parkinson's disease, and thus may provide the capability of monitoring the blood for a diverse range of PD pathophysiological mechanisms: cellular degeneration, oxidative stress, inflammation, and transport.