N-GlyCat: A Skyline tool featuring glycan spectral catalogs for automated analysis and curation of structure data

Introduction

Protein glycosylation mediates a broad range of protein and cellular functions. Proteins can be glycosylated at multiple sites and at each site it is possible to have multiple different glycan structures. Glycan structures cannot be predicted, but must be empirically measured. Recent developments in mass spectrometry (MS) and related software tools facilitate the analysis of glycans. However, widespread implementation of these approaches has been limited by the lack of approachable, automated tools for peak assignment of glycan structures. Here, we compiled glycomic data from GlycoPost to generate a non-redundant N-Glycan Catalog ( N-GlyCat) containing >400 structures. This spectral library can be exploited within the framework of GlyCat, a new external tool for Skyline, for automated glycan structure analyses.

Methods

930 raw MS data files deposited by multiple laboratories were retrieved from GlycoPost. Files were limited to those generated by porous graphitic carbon (PGC)-LC-MS/MS, a technique that characterizes and quantifies glycan structures by PGC normalized retention time (glucose unit, GU) and negative mode MS/MS. The top 100 product ions from each MS2 were used to create glycan structure spectral libraries with accompanying GU values for every glycan structure reported in each dataset. The quality of composition identifications was evaluated by the full isotopic distribution (including the M-1 isotopic ion) while structure identifications were made by a consensus agreement across datasets for the GU values and MS/MS spectra. An external graphical user interface for Skyline, GlyCat, was written in C#.

Preliminary Data

Glycan structure data interpretation requires manual peak picking and spectral interpretation, and automated methods for identifying incorrect annotations within published data are not available. This presents considerable challenges to the interpretation, reporting, sharing, and use of glycomics data. To address this, we developed GlyCat, a Skyline tool that enables automated glycan data curation. Briefly, spectral quality scores are based on isotopic distribution of a glycan chemical formula and MS/MS to annotated spectral libraries. This enabled identification of an acceptable MS1 isotopic dot product (idotp) score (>0.85) for confident composition matching. The contribution of the M-1 precursor ion to the idotp score discriminates glycan compositions that can be subject to the off-by-one error from incorrect monoisotopic precursor determination (e.g. Neu5Ac vs 2x fucose). MS2 spectra provide structural characterization, although inter-instrument and intensity variation limits the effectiveness of the dotp score (>0.70). Thus, GU is the primary structural determinant. GlyCat enables inspection of published data with objective measures. To demonstrate, we utilized the N-GlyCat library to inspect published PGC-LC-MS/MS data and found that, based on false glycan composition assignments, a 15% false discovery rate exists in published data. The most frequent errors were off-by-one errors. GlyCat also enables retention time calibration to reduce differences due to varying instrument configurations, column conditions and sample matrices within and among datasets. Although a dextran ladder internal standard can be used for retention time calibration, traditional approaches would require this standard to be present in every sample. We demonstrate that by having confident structural identifications (low isomeric complexity and dissimilar MS/MS) in a dataset, PGC-based glycomic datasets can be calibrated without a dextran ladder, allowing their incorporation into N-GlyCat. GlyCat is a freely available external tool for Skyline, N-GlyCat and associated assays are available in Panorama, and future iterations are expected to include O-glycan libraries.

Novel Aspect

GlyCat is the first automated structure assignment tool for MS1, MS2 and GU-based assessment of glycan structure data.