A key problem in computational proteomics is distinguishing between correct and false peptide identifications. We argue that evaluating the error rates of peptide identifications is not unlike computing generating functions in combinatorics. We show that the generating functions and their derivatives (spectral energy and spectral probability) represent new features of tandem mass spectra that, similarly to Δ-scores, significantly improve peptide identifications. Furthermore, the spectral probability provides a rigorous solution to the problem of computing statistical significance of spectral identifications. The spectral energy/probability approach improves the sensitivity-specificity trade-off of existing MS/MS search tools, addresses the notoriously difficult problem of "one-hitwonders" in mass spectrometry, and often eliminates the need for decoy database searches. We therefore argue that the generating function approach has the potential to increase the number of peptide identifications in MS/MS searches.

While bacterial genome annotations have significantly improved in recent years, techniques for bacterial proteome annotation (including post-translational chemical modifications, signal peptides, proteolytic events, etc.) are still in their infancy. At the same time, the number of sequenced bacterial genomes is rising sharply, far outpacing our ability to validate the predicted genes, let alone annotate bacterial proteomes. In this study, we use tandem mass spectrometry (MS/MS) to annotate the proteome of Shewanella oneidensis MR-1, an important microbe for bioremediation. In particular, we provide the first comprehensive map of post-translational modifications in a bacterial genome, including a large number of chemical modifications, signal peptide cleavages, and cleavages of N-terminal methionine residues. We also detect multiple genes that were missed or assigned incorrect start positions by gene prediction programs, and suggest corrections to improve the gene annotation. This study demonstrates that complementing every genome sequencing project by an MS/MS project would significantly improve both genome and proteome annotations for a reasonable cost.

Trypsin is the most commonly used enzyme in mass spectrometry for protein digestion with high substrate specificity. Many peptide identification algorithms incorporate these specificity rules as filtering criteria. A generally accepted "Keil rule" is that trypsin cleaves next to arginine or lysine, but not before proline. Since this rule was derived two decades ago based on a small number of experimentally confirmed cleavages, we decided to re-examine it using 14.5 million tandem spectra (2 orders of magnitude increase in the number of observed tryptic cleavages). Our analysis revealed a surprisingly large number of cleavages before proline. We examine several hypotheses to explain these cleavages and argue that trypsin specificity rules used in peptide identification algorithms should be modified to "legitimatize" cleavages before proline. Our approach can be applied to analyze any protease, and we further argue that specificity rules for other enzymes should also be re-evaluated based on statistical evidence derived from large MS/MS data sets.

The target-decoy approach (TDA) has done the field of proteomics a great service by filling in the need to estimate the false discovery rates (FDR) of peptide identifications. While TDA is often viewed as a universal solution to the problem of FDR evaluation, we argue that the time has come to critically re-examine TDA and to acknowledge not only its merits but also its demerits. We demonstrate that some popular MS/MS search tools are not TDA-compliant and that it is easy to develop a non-TDA compliant tool that outperforms all TDA-compliant tools. Since the distinction between TDA-compliant and non-TDA compliant tools remains elusive, we are concerned about a possible proliferation of non-TDA-compliant tools in the future (developed with the best intentions). We are also concerned that estimation of the FDR by TDA awkwardly depends on a virtual coin toss and argue that it is important to take the coin toss factor out of our estimation of the FDR. Since computing FDR via TDA suffers from various restrictions, we argue that TDA is not needed when accurate p-values of individual Peptide-Spectrum Matches are available.