Inspection of the corresponding scatter plot showed many data points above and below the primary linear cluster. Similarly, the heat map in Fig. 2B showed uniformly poor correlations for P3, and the scatter plot comparing pooled sample aliquots P3 vs. P4 displayed a group of points with higher values in the P4 sample than in P3. To investigate the source of this variability, we examined the sequences of peptides that were poorly correlated in pairwise comparisons and considered the proteins from which they were derived. Examination of the identified proteins revealed that some samples contained “process contaminant” proteins. The peptides from these proteins were removed from subsequent analyses of all samples and a new correlation heat map was generated. Although this exclusion of keratins and MAF proteins improved the correlations between selected pairwise comparisons, it did not improve all poor correlations. Further examination of the scatter plot of annotated peptide intensities for one of these unchanged pairwise comparisons showed a distinct cloud of points with lower values in 
sample 1a. All the sequences of these discordant peptides contained at least one methionine residue. Removal of these methionyl peptides from subsequent PCC calculations resulted in an increased correlation coefficient for 1a vs. 1b, and a modest increase in the correlation coefficient for 5a and 5b. However, the removal of methionyl peptides did not significantly alter the poorer correlation between P3 and the other pooled samples. A scatter plot of the peptide intensities of P3 vs. P4 showed a discrete group of peptide intensities that were lower in aliquot P3 than in P4. Upon inspection, these discordant peptides were found to have early Reversine elution times during liquid chromatography. The markedly decreased ion current for P3 compared to P4 during this early elution window is shown in Fig. S4, panels A and B. As an example of the marked reduction in signal for P3 during this elution time window, all sample intensities for an early-eluting NCAM-1 peptide are shown in Fig. S4, panels C and D. Scatter plots of the peptide intensities of P3 vs. P4 before and after removal of the early eluting peptides show improved correlation after exclusion. In kind, a new non-clustering heat map of PCC values, created after excluding these early-eluting peptides from all PCC calculations, showed improved correlations and a narrower range of PCC values between P3 and all other pooled samples. Although the reason for the low intensities of earlyeluting peptides found in sample P3 remains unclear and does not appear to affect other samples substantially, these early-eluting peptides were excluded from subsequent analyses of all samples. To estimate the residual technical variability with which this technique quantifies each of these proteins, coefficients of variation were calculated from replicates of the pooled CSF sample; to determine the inter-individual variability of each of these proteins within a group of cognitively normal individuals, subject variance was calculated from the paired aliquots from individual samples. These assessments of variability are important for biomarker studies because, hypothetical proteins with relatively higher CVs and/or relatively high subject variance among controls will be more likely than those with lower CVs and lower subject variances to show overlapping ranges between diseased and control cohorts, given equivalent fold-changes associated with disease. Greater overlaps mean lower sensitivity and specificity and, therefore, relatively less potential for a candidate to serve as an effective disease biomarker. This study evaluates the technical merits and potential of a nonbiased proteomics technique, label-free MK-1775 quantitative LC-MS/MS.