Combining high resolution and exact calibration to boost statistical power: A well-calibrated score function for high-resolution MS2 data
To achieve accurate assignment of peptide sequences to observed fragmentation spectra, a shotgun proteomics database search tool must make good use of the very high resolution information produced by state-of-the-art mass spectrometers. However, making use of this information while also ensuring that the search engines scores are well calibrated--i.e., that the score assigned to one spectrum can be meaningfully compared to the score assigned to a different spectrum--has proven to be challenging. Here, we describe a database search score function, the \"residue evidence\" (res-ev) score, that achieves both of these goals simultaneously. We also demonstrate how to combine calibrated res-ev scores with calibrated XCorr scores to produce a \"combined p-value\" score function. We provide a benchmark consisting of four mass spectrometry data sets, which we use to compare the combined p-value to the score functions used by several existing search engines. Our results suggest that the combined p-value achieves state-of-the-art performance, generally outperforming MS Amanda and Morpheus and performing comparably to MS-GF+. The res-ev and combined p-value score functions are freely available as part of the Tide search engine in the Crux mass spectrometry toolkit (http://crux.ms).