Quantifying the Impact of the Peptide Identification Framework on the Results of Fast Photochemical Oxidation of Protein Analysis

Abstract

Fast Photochemical Oxidation of Proteins (FPOP) is a promising technique for studying protein structure and dynamics. The quality of insight provided by FPOP depends on the reliability of the determination of the modification site. This study investigates the performance of two search engines, Mascot and PEAKS, for the data processing of FPOP analyses. Comparison of Mascot and PEAKS of the hemoglobin--haptoglobin Bruker timsTOF data set (PXD021621) revealed greater consistency in the Mascot identification of modified peptides, with around 26% of the IDs being mutual for all three replicates, compared to approximately 22% for PEAKS. The intersection between Mascot and PEAKS results revealed a limited number (31%) of shared modified peptides. Principal Component Analysis (PCA) using the peptide-spectrum match (PSM) score, site probability, and peptide intensity was applied to evaluate the results, and the analyses revealed distinct clusters of modified peptides. Mascot showed the ability to assess confident site determination, even with lower PSM scores. However, high PSM scores from PEAKS did not guarantee a reliable determination of the modification site. Fragmentation coverage of the modification position played a crucial role in Mascot assignments, while the AScore localizations from PEAKS often become ambiguous because the software employs MS/MS merging.

Keywords: FPOP; search engine; structural proteomics.

Figure 1

Intersections (orange) of IDs by Mascot (red) and PEAKS (green) in the (A) hemoglobin sample, (B) haptoglobin sample, and (C) HbHp complex sample.

Figure 2

Principal component analysis of modified peptides identified in the HbHp complex sample by (A) Mascot and (B) PEAKS. The plot of color-coded modifications shows their distribution within dimensions defined by the PCA. The vectors show the correlation of the variables with the PCA dimensions. The logAScore variable represents the probability of site determination. In both cases, the IDs are distributed into 3 clusters. For Mascot, most of the IDs are in the top cluster, while for PEAKS, the majority of the IDs are in the middle cluster.

Figure 3

Fragmentation spectra of selected IDs from the PCA model of the Mascot search results for the HbHp complex. The ⧫ sign labels the precursor ion. (A) Trp7 oxidation (+15.995) in the peptide SAVTALWGK in its singly charged form at m/z 948.51 and RT 9.2 min from the top cluster, y3 −H₂O undetected by Mascot was annotated manually; (B) Leu2 carbonyl (+13.979) in the peptide VLSPADKTNVK in its doubly charged form at m/z 593.33 and RT 12.5 min from the middle cluster; (C) Tyr10 oxidation (+15.995) in the peptide DYAEVGRVGYVSGWGR in its triply charged form at m/z 596.29 and RT 9.8 min from the bottom cluster. The representative IDs show that the PCA is able to separate reliable, questionable, and unreliable modified peptides.

Figure 4

Fragmentation spectra of selected IDs from the PCA model of the PEAKS search results for the HbHp complex: (A) Val3 carbonyl (+13.979) in the peptide TNVKAAWGK in its singly charged form at m/z 988.52 at RT 7.27 min from the top cluster; (B) Val3 oxidation (+15.995) in the peptide with the sequence LRVDPVNFK in its doubly charged form at m/z 552.31 at RT 8.19 min from the middle cluster; (C) His13 oxidation (+15.995) in the peptide LLGNVLVCVLAHHFGK in its doubly charged form at m/z 897.00 at RT 21.01 min from the bottom cluster. The selected IDs show that the fragmentation coverage of the site of modification and signal intensity do not have a clear relation to the AScore and position within the PCA clusters.