Dave Horn3, Daniel Lopez Ferrer3, Bernard Delanghe2, Daniel Zolg1, Martin Frejno1
International Hupo 2020
Most database search algorithms compare experimental fragmentation spectra of peptides with lists of theoretical fragment masses corresponding to peptides from an in silico digest of a protein database to calculate similarity measures, while largely disregarding the intensity dimension. Automatically matching peptides to spectra in this way will yield false identifications of low-quality spectra or misrepresent their confidence. The standard method to control for erroneous matching of such spectra is the target-decoy approach that estimates the False Discovery Rate (FDR) in bottom-up proteomics experiments. Machine learning methods such as Percolator are commonly used to separate incorrect from correct matches, but their performance heavily depends on the calculated scores. Here, we show how intensity-based scores successfully circumvent common issues and challenges in peptide identification.
1MSAID GmbH, Garching, Germany; 2Thermo Fisher Scientific (Bremen) GmbH, Bremen, Germany; 3Thermo Fisher, San Jose, CA
United States
