A Multivariate Mixture Model to Estimate the Accuracy of Glycosaminoglycan Identifications Made by Tandem Mass Spectrometry (MS/MS) and Database Search

Abstract

We present a statistical model to estimate the accuracy of derivatized heparin and heparan sulfate (HS) glycosaminoglycan (GAG) assignments to tandem mass (MS/MS) spectra made by the first published database search application, GAG-ID. Employing a multivariate expectationmaximization algorithm, this statistical model distinguishes correct from ambiguous and incorrect database search results when computing the probability that heparin/HS GAG assignments to spectra are correct based upon database search scores. Using GAG-ID search results for spectra generated from a defined mixture of 21 synthesized tetrasaccharide sequences as well as seven spectra of longer defined oligosaccharides, we demonstrate that the computed probabilities are accurate and have high power to discriminate between correctly, ambiguously, and incorrectly assigned heparin/HS GAGs. Heparin and heparan sulfate (HS), members of the glycosaminoglycan (GAG) family, are linear polysaccharides composed of repeating disaccharide building blocks of variously sulfated hexuronic acid (134) D-glucosamine units that structurally differ solely by the length of the oligosaccharide and degree of modification, with heparin being more heavily sulfated and having less N-acetylation than HS. Interacting with proteins, heparin/HS play essential roles in a wide variety of biological processes, including anticoagulation (1), cell proliferation (2, 3), and carcinogenesis (4, 5). The specificity of these interactions is driven by the pattern of modification of heparin/HS oligosaccharide sequences. To understand the molecular role of heparin/HS, it is necessary to correlate function with the fine structure of the carbohydrate. However, the non-template-driven biosynthesis of heparin/HS results in extremely diverse structures. Analyzing heparin/HS is challenging for three reasons: the presence of multiple isomeric sequences in a complex mixture of oligosaccharides, the difficulty of separating the isomers, and the facile loss of sulfates in MS/MS (6).We previously introduced a method for structurally sequencing heparin/HS oligosaccharides that involves chemical derivatizations to replace labile sulfates with stable acetyl groups (7). This derivatization scheme allows for the use of reverse-phase liquid chromatography (LC) for high-resolution separation of isomeric heparin/HS oligosaccharides and MS/MS for sequencing them. However, the data from this derivatization method cannot be easily incorporated into current glycomic software, such as GlycoWorkbench (8), due to the multistep derivatizations and lack of a scoring algorithm that accurately evaluates the matches. We recently reported the development of a software tool for the high-throughput analysis of LC-MS/MS data from these derivatized heparin/HS oligosaccharides, entitled GAG-ID (9), which is the first database-driven software package for this purpose. GAG-ID produces a GAG sequence assignment for each input spectrum; however, some assignments are true matches and some are false. False matches arise from low-qua...

Developments in Mass Spectrometry for Glycosaminoglycan Analysis: A Review

“…Determining specific sites of modification within GAG molecules is essential to understanding GAG–protein binding and biological function. The simplest way to assign structural characteristics is through brute force methods that generate all possible fragments and match theoretical fragments with experimental data ( 138 , 139 , 140 ). Even with the processing power of modern computers and CPUs, this method can take an exceedingly long time for larger GAG precursors.…”

“…GAG-ID was developed by Chiu et al. to automate assignment of derivatized Hp/HS that are separated and fragmented with LC-MS/MS ( 138 , 139 ). Duan et al.…”

GAGrank: Software for Glycosaminoglycan Sequence Ranking Using a Bipartite Graph Model

“…GAG-ID automates the interpretation of permethylated HS LC/MS 2 data using a multivariate hypergeometric distribution with detected peaks separated into high-, medium-, and low-intensity bins. The same group later used a multivariate mixture model to determine GAG-ID identification accuracy given database search scores and ambiguity values among identifications [32]. Finally, Duan and colleagues recently published over two publications a J o u r n a l P r e -p r o o f method for interpreting CS GAG MS 2 data that uses a genetic algorithm to assign a likelihood score to each sequence for a given MS 2 spectrum [33,34].…”

Analysis of the Glycosaminoglycan Chains of Proteoglycans

“…This analysis makes it possible to filter large MS/MS database search results with predictable false identification error rates. 82 For HP/HS analysis, GlycCompSoft is designed to enable the comparison of top-down analytical glycomics data on two or more LWMHs. In a test based on three lots of Lovenox, Clexane, and three generic enoxaparin samples, the program proved its low error rate and good time efficiency when processing large data sets.…”