Objective: To find a panel of proteins in antemortem cerebrospinal fluid (CSF) that could be used to differentiate between samples from Alzheimer's disease (AD) patients and samples from healthy and neurological control subjects.
Methods: For a test cohort, antemortem CSF proteins from 34 AD and 34 non-AD patients were separated using two-dimensional gel electrophoresis. The resulting protein patterns were analyzed using the random forest multivariate statistical method. Protein spots of interest were identified using tandem time-of-flight mass spectrometry. A validation cohort consisting of CSF from 18 AD patients and 10 non-AD subjects was analyzed in a similar way.
Results: Using the test cohort, a panel of 23 spots was identified that could be used to differentiate AD and non-AD gels with a sensitivity of 94%, a specificity of 94%, and a predicted classification error rate of only 5.9%. These proteins are related to the transport of beta-amyloid, the inflammatory response, proteolytic inhibition, and neuronal membrane proteins. The 23 spots separately classified the validation cohort with 90% sensitivity (probable AD subjects), 83% specificity, and a predicted classification error rate of 14% in a blinded analysis. The total observed sensitivity is 93%, the total observed specificity is 90%, and the predicted classification error rate is 8.3%.
Interpretation: A panel of possible CSF biomarkers for AD has been identified using proteomic methods.