Protein identification at the low nanogram level could in principle be obtained by most nanoscale LC-MS/MS systems. Nevertheless, the complex sample preparation procedures generally required in biological applications, and the consequent high risk of sample losses, very often hamper practical achievement of such low levels. In fact, the minimal amount of protein required for the identification from a gel band or spot, in general, largely exceeds the theoretical limit of identification reachable by nanoscale LC-MS/MS systems. A method for the identification of low levels of purified proteins, allowing limits of identification down to 1 ng when using standard bore, 75 microm id nanoscale LC-MS/MS systems is here reported. The method comprises an offline two-step sample cleanup, subsequent to protein digestion, which is designed to minimize sample losses, allows high flexibility in the choice of digestion conditions and delivers a highly purified peptide mixture even from "real world" digestion conditions, thus allowing the subsequent nanoscale LC-MS/MS analysis to be performed in automated, unattended operation for long series. The method can be applied to the characterization of low levels of affinity purified proteins.