Abstract
In MS-based proteomics, one of the main tasks of bioinformatics is still the matching of peptide sequences to experimental spectra. This is done by search engines, of which there are currently many. The main reason for choosing one of these is how well it fits into a desired workflow or simply the researcher's preference. While some algorithmic parts of each search engine are similar, there are also differences that ultimately lead to variations in results.Due to the common problem of false positive identifications in high throughput methods, the initial results need to be controlled for a false discovery rate. For this post-processing step, several approaches have been developed in recent years, most of which are applicable to any search engine result.
Here we present the results of a comparison of widely used proteomics search engines in combination with the most common post-processing approaches, namely the standard target-decoy approach, Percolator and rescoring of PSMs using machine learning. We ran this workflow on publicly available datasets to compare the overlap between the different approaches and to highlight distinctive spectra or peptides found only by certain combinations. This evaluation allowed us to assess whether it is more important to choose a specific search engine or post-processing method. In addition, we were able to show that although there are large differences in the initial results, these can be largely harmonised using post-processing methods.