In so doing, it describes analytical methods that can be applied to identify variability arising from technical sources across the workflow, from sample procurement through LC-MS. It also compares two approaches to quantify proteins from peptide data. Further, it presents baseline statistical data for 81 relatively abundant CSF proteins within a neurologically normal group of older individuals, calculates the technical variability and gradient effects observed in the measurement of these proteins within multiple replicates of a pooled CSF sample, and illustrates how selective subsets of these proteins might be used to classify samples that differ by biological phenotype. Thus, it provides a framework for future experiments that will evaluate CSF samples from individuals with neurological diseases, in search of relevant biomarkers. Recently, other reports have also described the use of this technique for CSF proteome characterization and biomarker discovery. Impressively, some have identified and quantified hundreds or thousands of proteins in a single CSF sample, by referencing the unique LC retention times and m/z values of the extracted ion chromatograms of peptides to an annotated ‘library’ of retention times and m/z values compiled from previous MS/MS analyses of similar CSF preparations. For this current experiment, no such annotated AMT library was available. Instead, MS1 and MS2 R428 scanning were performed simultaneously to enable the annotation of peptides in real time. This approach availed the identification of a comparatively smaller number of proteins, but was wholly adequate for the purpose of this experiment, which was not intended to discover novel rare CSF proteins or to maximize the number of proteins identified. As it happens, recent instrumentation advances during the short interval since this experiment was conducted now allow for the annotation of many more peptides in real time, effectively increasing the sensitivity of simultaneous scanning; these changes have reduced the popularity of the more laborious AMT approach. Regardless, even without such advances, either approach is likely to identify promising candidate biomarkers. Indeed, even in this limited experiment, in which tandem mass spectrometry was triggered solely on the basis of relatively high peptide abundance, a comparatively modest list of 81 proteins generated sufficient diagnostic potential to allow perfect segregation of ‘individual’ and ‘pooled’ sample replicates with a much smaller subset of 24 selected proteins. It is also encouraging that many of the 81 proteins have already been reported as potential biomarkers for AD by multiple independent groups. Indeed, because none of these previously reported candidate biomarkers have been vetted sufficiently to be applied in clinical trials, they will have to be studied further: individually and in combination; in larger cohorts and in different diseases. Thus, particularly with recent advances, this technique is well suited for application in future AD research studies to facilitate the U0126 validation of promising biomarkers. A final point of discussion addresses the purpose and the implications of the hierarchical clustering analyses performed in this study. Such analyses are employed here to illustrate the potential of this technique to measure ensembles of proteins that can classify samples according to desired characteristics. In most biomarker discovery studies, such clustering analyses would be preceded by a selection process in which candidate biomarkers are vetted on the basis of statistical association with a diagnosis of interest. In the current study, because the samples analyzed do not strictly represent two different disease states, the proteins were evaluated, instead, for their ability to segregate CSF from different sources.