Peptide identifications and false discovery rates using different mass spectrometry platforms

Abstract

Characterization of endogenous neuropeptides produced from post-translational proteolytic processing of precursor proteins is a demanding task. A variety of complex prohormone processing steps generate molecular diversity from neuropeptide prohormones, making in silico neuropeptide discovery difficult. In addition, the wide range of endogenous peptide concentrations as well as significant peptide complexity further challenge the structural characterization of neuropeptides. Liquid chromatography-mass spectrometry (MS), performed in conjunction with bioinformatics, allows for high-throughput characterization of peptides. Mass analyzers and molecular dissociation techniques render specific characteristics to the acquired data and thus, influence the analysis of the MS data using bioinformatic algorithms for follow-up peptide identification. Here we evaluated the efficacy of several distinct peptidomic workflows using two mass spectrometers, the Thermo Orbitrap Fusion Tribrid and Bruker Impact HD UHR-QqTOF, for confident peptide discovery and characterization. We compared the results in several categories, including the numbers of identified peptides, full-length mature neuropeptides among all identifications, and precursor proteins mapped by the identified peptides. We also characterized the peptide false discovery rate (FDR) based on the occurrence of amidation, a known post-translational modification (PTM) that has been shown to require the presence of a C-terminal glycine. Thus, amidation events without a preceding glycine were considered false-positive amidation assignments. We compared the FDR calculated by the search engine used here to the minimum FDR estimated via false amidation assignments. The search engine severely underestimated the rate of false PTM assignments among the identified peptides, regardless of the specific MS platform used.

Keywords: Amidation; De novo sequencing; Neuropeptides; Peptidomics; Proteomics.

Fig. 1

Schematic of post-translational prohormone processing. (A) The prohormone is acted upon by several endopeptidases, followed by other enzymes, to form various PTMs, leading to the full-length neuropeptide hormones. (B) Artifacts resulting from additional enzymatic and non-enzymatic processing/degradation leading to a series of shortened peptides.

Fig. 2

Average number of unique peptide and neuropeptide sequences identified over n = 3 technical replicates for four different instrumental methods, where neuropeptides are defined as peptides derived from prohormones. Error bars represent the standard deviation. Complete lists of the peptides and proteins are provided in the supplementary material (Figs. S3–S5);*p <0.05, **p <0.01.

Fig. 3

Average number of neuropeptide precursor proteins identified over n = 3 technical replicates for four different instrumental methods. Error bars represent the standard deviation;**p <0.01.

Fig. 4

Average number of peptides identified from n = 3 technical replicates for four different instrumental methods using two different databases: exclusively Aplysia californica (target, A.C.) and mixed species with precursor proteins from (A.C) + Homo sapiens (H.S.). *p <0.05, **p <0.01. For the peptides from the A.C. + H.S. database, all pairwise comparisons, except HCD-IT and QTOF-CID, are significantly different; p <0.05.

Fig. 5

Comparison of the search engine-reported peptide FDR percentages to experimentally determined percentages for the occurrence of false positives using all of the four tested platforms. The x-axis corresponds to the constant PSM FDR threshold cutoff used to filter the results from the PEAKS search. The y-axis corresponds to the peptide level FDR percentage estimated by the search engine and experimentally determined value via evaluation of incorrect amidations. (A) HCD-OT, (B) CID-IT, (C) HCD-IT and the (D) CID-QTOF.