We use Mascot (Matrix Science) for protein identification. Protein inference, grouping and FDR (false discovery rate) estimation is performed in Proteome Discoverer and Scaffold.
Our default database is UniProt. As routine policy, for poorly annotated proteomes we search both the sections of Uniprot i.e. SwissProt (curated) and Trembl (non-curated). After both the searches, non redundant protein list with each IDs supported by at least one unique peptide is submitted to the user. Both databases are updated every month.
The reporting of a FDR has become mandatory. The FDR is essentially a statistical and computational measure to establish the certainty of the protein identification in a given dataset. There are several methods, we typically use a decoy database to measure the false discovery rate.