Optimization-Based Peptide Mass Fingerprinting for Protein Mixture Identification

In current proteome research, the most widely used method for protein mixture identification is probably peptide sequencing. Peptide sequencing is based on tandem mass spectrometry (MS/MS) data. The disadvantage is that MS/MS data only sequences a limited number of peptides and leaves many more peptides uncovered. Peptide mass fingerprinting (PMF) has been widely used to identify single purified proteins from single-stage MS data. Unfortunately, this technique is less accurate than the peptide sequencing method and cannot handle protein mixtures, which hampers the widespread use of PMF. In this article, we tackle the problem of protein mixture identification from an optimization point of view. We show that some simple heuristics can find good solutions to the optimization problem. As a result, we obtain much better identification results than previous methods. Through a comprehensive simulation study, we identify a set of limiting factors that hinder the performance of PMF-based protein mixture identification. We argue that it is feasible to remove these limitations and PMF can be a powerful tool in the analysis of protein mixtures, especially in the identification of low-abundance proteins, which are less likely to be sequenced by MS/MS scanning.