Fig. 3: Evaluation of modeling approach on orthogonal controls and other datasets.
a The effect of network smoothing on target- and decoy-peptide pairs for individual spectra showing differences in score and differences in glycan smoothing term ug for the mouse tissue dataset, b Validation of network smoothing with entrapment FDR at a regular GPSM 1% FDR showing it does not appreciably change compared to the base search though see supplementary section 9 for more details, c The GPSM counts by search strategy in the human serum dataset PXD005931 at 1% FDR with individual replicates shown as swarm points colored across strategies, with median line and error bars reflecting inter-quartile range of GPSM count within each search strategy from the 24 serum samples, d The GPSM counts for the human serum dataset PXD009654 using either the base scoring method, the fragmentation model trained on PXD005931, or the fragmentation model trained on PXD009654 to measure transferability within instrument/species at different collision energies, with median line and error bars reflecting inter-quartile range over the three replicates per condition, e The GPSM counts at 1% FDR for the mouse tissue datasets to compare the transferability of the fragmentation model trained on PXD005931 across instruments, species, and collision energies. Source data are provided as a Source Data file.
