A major challenge of proteomic studies is the accurate quantitation of proteins. LC-MS/MS-based methods are especially suited for profiling proteins in large sample sets. In this setup, the measurement of relative protein abundance relies on the correct quantitation of tryptic peptides. However, peptide intensities often do not unequivocally reflect the abundance of the native proteins in the sample. In this study, we show that peptides that accurately reflect relative protein abundances in large-scale sample sets can be selected based on the correlation to each other. This strategy was tested in a well-controlled experiment using a set of spiked serum samples as well as 55 clinical serum samples from schizophrenia patients and healthy volunteers. The peptide correlation analysis we present here provides an intuitive and simple procedure to obtain a high quality quantitative information from proteomics data.