The conventional approach in modern proteomics to identify proteins from limited information provided by molecular and fragment masses of their enzymatic degradation products carries an inherent risk of both false positive and false negative identifications. For reliable identification of even known proteins, complete de novo sequencing of their peptides is desired. The main problems of conventional sequencing based on tandem mass spectrometry are incomplete backbone fragmentation and the frequent overlap of fragment masses. In this work, the first proteomics-grade de novo approach is presented, where the above problems are alleviated by the use of complementary fragmentation techniques CAD and ECD. Implementation of a high-current, large-area dispenser cathode as a source of low-energy electrons provided efficient ECD of doubly charged peptides, the most abundant species (65-80%), in a typical trypsin-based proteomics experiment. A new linear de novo algorithm is developed combining efficiency and speed, processing on a conventional 3 GHz PC, 1000 MS/MS data sets in 60 s. More than 6% of all MS/MS data for doubly charged peptides yielded complete sequences, and another 13% gave nearly complete sequences with a maximum gap of two amino acid residues. These figures are comparable with the typical success rates (5-15%) of database identification. For peptides reliably found in the database (Mowse score > or = 34), the agreement with de novo-derived full sequences was >95%. Full sequences were derived in 67% of the cases when full sequence information was present in MS/MS spectra. Thus the new de novo sequencing approach reached the same level of efficiency and reliability as conventional database-identification strategies.