Once the protein is extracted, it is necessary to cut the protein into peptides as most LC–MS/MS technologies are best suited for measuring peptide fragments 5–25 amino acids long. There are a numerous proteases that can digest proteins into peptides. In proteomic studies, the most commonly used protease is trypsin which cleaves the proteins at lysine and arginine residues. When a protein sample is treated with trypsin, a peptide complex composed of peptides flanked by either a lysine or an arginine residue is generated (Table 23.2). The peptide complex generated by trypsin digestion is then separated by LC. The most commonly used LC approaches use solvent gradients to separate and resolve the peptides based on hydrophobic characteristics. In this way, peptides with similar chemical characteristics move together and are sprayed into the MS for mass detection. MS could only measure the mass of a peptide, if the peptide carries an ionic charge. As most native peptides are charge neutral, ionization of the peptides is required before they are loaded to MS. For peptides in solution as described here, this is achieved by electrospray ionization (ESI). After separation by LC, the solution containing the peptides is forced through a very fine needle exposed to high voltage which leads to the removal of the solvent carrying the peptides and ionization of the peptides. The ionized peptides are sprayed into the MS. In tandem MS/MS analysis, the first MS measures the mass to charge (m/z) ratio of the parent peptide. Then, peptides selected based on predetermined criteria are directed to a collusion cell where the peptides are fragmented upon collusion with an inert gas such as helium. This process is called collusion-induced dissociation (CID). The fragments formed by the CID are measured in the second MS. Each peptide present in the human proteome has a unique CID pattern which makes it possible to predict the amino acid sequence of the peptide being analyzed by MS/MS. A number of complex computational algorithms have been developed to predict the amino acid sequences. The algorithms compare the observed fragmentation pattern of each peptide to the theoretical fragmentation pattern of all human tryptic peptides predicted by the reference human genome, and assign a probability score for individual peptide as well as a protein probability score for identification of a given protein often based on multiple peptides derived from that protein [19]. Although this is essentially a computational process, and not true sequencing of a protein, the MS instruments have such a high precision and computational methods have become so sophisticated that the results are extremely accurate and reproducible. The work flow for LC–MS/MS-based proteomic analysis is summarized in Fig. 23.1.

The MS-based proteomic analysis has a number of limitations. A given protein can only be identified if peptide fragments with appropriate size for MS can be generated after enzymatic digestion. For example, a number of human proteins contain large areas lacking trypsin cutting residues lysine or arginine; therefore, no peptides suitable for LC–MS/MS detection can be generated. In shotgun proteomic approaches, it may be difficult to detect low abundance proteins/peptides as signals from these peptides may be buried among massive amount of information obtained from more abundant proteins, and MS simply may not be able to scan them. The other important limitation of MS-based proteomics is the reliance of computational predictive algorithms to a reference human genome obtained from publicly available databases such as Swissprot. The algorithms could only match the observed peptide fragmentation data to the protein sequences available in the public databases. Therefore, germline polymorphisms or somatic mutations that are not represented in public databases cannot be identified despite MS data from these peptides would have been captured.