Statistically significant differences between sample groups. Given sufficient sample and analytical time, extensive annotated libraries that match XICs to their unique peptide sequences can be accumulated for a given sample type. This accurate mass and time tag approach can be applied retrospectively or prospectively, reducing or eliminating the need for tandem mass spectrometry in subsequent studies of that biofluid within a given laboratory. Alternatively, MS2 can be performed in series with LC-MS during primary quantitative data acquisition. In this work, we apply quantitative label-free LC-MS/MS to the analysis of replicate CSF samples: to identify sources of technical variability that can be mitigated; to assess the inter-individual and residual technical variance with which this technique measures numerous proteins in cognitively normal subject samples; to compare alternative strategies for protein quantification from quantitative peptide data; and to test the ability of its output to segregate biological samples according to desired biomarker characteristics. In this way, we demonstrate the suitability of quantitative label-free LC-MS/MS as a tool for CSF biomarker discovery. During the performance of each of these LC-MS analyses, the mass spectrometer was operating in data dependent mode and automatically isolated the most abundant ions for MS2. After processing, the MS2 data were then searched against the UNIPROT database for annotation as described in Fig. S3. The annotated features were processed using a visual script that was executed with Rosetta ElucidatorTM software. The annotated peak intensities were normalized as described in Fig. S3. Because some peptides were detected in more than one charge state, the individual charge states for each peptide were combined, yielding 1360 annotated isotope groups. Additionally, when multiple isotope groups within an LC-MS analysis were found to be associated with a common peptide sequence, their peak intensities were summed, yielding 926 annotated, aligned peptide ion chromatograms across the 25 samples. The data from each step of the sequential processing diagrammed in Fig. 2A were exported into a spread sheet, grouped as gene names and imported into DAnTE-R software for further analysis. To evaluate variability in the proteomics workflow at the level of annotated peptides, non-clustering heat maps of Axitinib Pearson correlation coefficients were used. PCC values were generated using the log2 transformed peptide intensity data for all pair-wise sample comparisons. Fig. 2B shows the symmetrical matrix of PCC values from all pairwise correlations among the pooled and individual sample aliquots, calculated from all annotated charge groups; for the purpose of comparison, Fig. S11 shows a similar matrix, calculated from all aligned charge groups. The diagonal white squares represent self-comparisons that yield a perfect correlation of 1.0 on a scale of 0.65 to 1.0; as shown by the color bar, imperfect correlations are AbMole BioScience kinase inhibitors represented by increased shading that ranges from yellow to orange to red, with the lowest values appearing as black squares. These Pearson correlation heat maps corroborate the ion current results; pairwise comparisons of the pooled sample with the lowest ion current yielded uniformly low PCC’s, represented by intense black bars in Fig. 2B and Fig. S11, with samples P13b and P5 showing the poorest correlation. The corresponding P13 vs. P5 scatter plot appears as a wide, homogeneous cloud of log2 transformed aligned charge group intensities. In contrast, the scatter plot of two pooled samples with a high PCC, represented in Fig. 2D, appears predominantly as a tight linear cluster. We concluded from these data that TIC values below, 25% of the mean value of the data set are not normalized by the algorithm described in Materials and Methods.