Globally FDR-Filtered MSDT

Global FDR control was applied in a dataset-wide manner using a target–decoy strategy. Peptide- and protein-level FDR were controlled globally by jointly considering target and decoy entries across all runs, with q-values recomputed under global target–decoy competition.

For FragPipe-derived results, PSMs were rescored using MSBooster, and for each spectrum only the identification with the lowest posterior error probability (PEP) was retained prior to cross-run merging. For Sage-derived results, PEP values were obtained from the search output and similarly filtered to retain the lowest-PEP identification per spectrum and per peptide before merging. Global peptide-level FDR control (1%) was then applied using the calculateFDR function from the PGA R package [1], and the resulting high-confidence peptide set was used to retrieve the corresponding PSMs from individual runs.

Globally FDR-Filtered MSDT files retain only fields directly relevant to spectral representation and downstream modeling, including scan, charge, mz_array, intensity_array, precursor_sequence, proteins, and filename. Results from FragPipe and Sage are stored separately and are not merged, thereby avoiding potential biases arising from differences in statistical assumptions and scoring schemes across search engines, while providing higher-confidence identifications suitable for downstream analysis.