Proposal of a decoy-free FDR approach suitable for metaproteomics.
Résumé
Accurate and fast evaluation of the False Discovery Rate (FDR) of spectra-to-peptide sequences inference is a difficult task in Metaproteomics, because the extra-large databases used are often largely incomplete and include numerous non-sample-specific sequences, particularly when using microbiota gene catalogs or generalist databases.
The traditional approach relying on combined target-decoy databases doubles search time, decreases sensitivity because of a larger search space, and is often biased because of dataset-database inadequate matching. We propose a target-only FDR estimate based on a mixture-model of four beta distributions.
We verified its efficiency on a set of 94 result datasets, including 26 metaproteomics searches, and a specific search with a controlled mismatching metaproteomics database.
Based on these extensive results, we found this method to be adequate for FDR estimation at the Peptide-Spectrum Matches level for proteomics, proteogenomics, and metaproteomics searches.