The discovery of biomarkers in easily accessible body fluids such as serum is one of the most challenging topics in proteomics requiring highly efficient separation and detection methodologies. Here, we present the application of a microfluidics-based LC-MS system (chip-LC-MS) to the label-free profiling of immunodepleted, trypsin-digested serum in comparison to conventional capillary LC-MS (cap-LC-MS). Both systems proved to have a repeatability of approximately 20% RSD for peak area, all sample preparation steps included, while repeatability of the LC-MS part by itself was less than 10% RSD for the chip-LC-MS system. Importantly, the chip-LC-MS system had a two times higher resolution in the LC dimension and resulted in a lower average charge state of the tryptic peptide ions generated in the ESI interface when compared to cap-LC-MS while requiring approximately 30 times less (~5 pmol) sample. In order to characterize both systems for their capability to find discriminating peptides in trypsin-digested serum samples, five out of ten individually prepared, identical sera were spiked with horse heart cytochrome c. A comprehensive data processing methodology was applied including 2-D smoothing, resolution reduction, peak picking, time alignment, and matching of the individual peak lists to create an aligned peak matrix amenable for statistical analysis. Statistical analysis by supervised classification and variable selection showed that both LC-MS systems could discriminate the two sample groups. However, the chip-LC-MS system allowed to assign 55% of the overall signal to selected peaks against 32% for the cap-LC-MS system.