An Improved Approach for Glycan Structure Identification from HCD - - PowerPoint PPT Presentation

Introduction Method Experiments

An Improved Approach for Glycan Structure Identification from HCD MS/MS Spectra

Weiping Sun, Yi Liu, Gilles Lajoie, Bin Ma and Kaizhong Zhang

Department of Computer Science, Western University wsun63@uwo.ca

October 3, 2016

1 / 18

Introduction Method Experiments

Overview

1

Introduction

2

Main Method

3

Experiments and Discussion

2 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Glycosylation

One of the most important PTMs 70% human proteins are glycosylated

3 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Glycosylation

One of the most important PTMs 70% human proteins are glycosylated Types of Glycosylation N-linked glycosylation

– Attached to N(Asn) – Motif; core structure

O-linked glycosylation

– Linked to S/T, or hydroxylysine residues

3 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

MS-Based Glycoproteomic Analysis

4 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Tandem Mass Spectrometry

Three common fragmentation techniques: CID: B- and Y-ions HCD: B- and Y-ions, as well as A- and X-ions ETD/ECD: C- and Z-ions

5 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Glycan Identification from HCD Spectrum

6 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Glycan Identification from HCD Spectrum

7 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Computational Approaches for Spectral Data Interpretation

Database search – Search from a glycan database to find matched glycan candidates. Examples: GlycoSearchMS, GlycoWorkBench, MAGIC, GlycoMaster DB, etc.

8 / 18

Introduction Method Experiments Glycosylation Glycoproteomics Approaches

Computational Approaches for Spectral Data Interpretation

Database search – Search from a glycan database to find matched glycan candidates. Examples: GlycoSearchMS, GlycoWorkBench, MAGIC, GlycoMaster DB, etc. VS De novo sequencing – Computation does not rely on glycan database knowledge, instead the algorithms directly construct glycans from MS/MS. Examples: Glycan: STAT, GlyCH, StrOligo, etc. Glycopeptide: GlycoMaster, etc.

8 / 18

Introduction Method Experiments Mathematical Model Method

Motivations

De novo sequencing need high-quality mass spectra. Database search has the ability to obtain more reliable results. Our previous de novo sequencing method can at least provide useful structures.

9 / 18

Introduction Method Experiments Mathematical Model Method

Glycan Database Search Problem

Glycan: a labelled rooted unordered tree with bounded degree. Glycan database search problem: Input:

• An MS/MS spectrum M
• A glycan database D
• A predefined mass error tolerance δ

Output:

A glycan structure T in D that satisfies,

• |T + P + H2O + 1 − Mp| ≤ δ;
• Matching score between M and T is maximized.

10 / 18

Introduction Method Experiments Mathematical Model Method

Main Idea

Use de novo sequencing result to filter glycans selected from database.

11 / 18

Introduction Method Experiments Mathematical Model Method

Step 1: Peptide mass calculation

12 / 18

Introduction Method Experiments Mathematical Model Method

Step 2: Glycan candidate selection and raw score calculation

• 1. Calculate glycan mass
• 2. Screen glycan database for possible glycan candidates
• 3. Calculate their raw score

Sraw = α

• f (mBi, hBi)+β
• f (mY j, hY j)+θ
• f (mI k, hI k)

13 / 18

Introduction Method Experiments Mathematical Model Method

Step 3: Filtration

A list of de novo sequencing glycans: Ln = {R1, R2, . . . , Rm} A list of database glycans: Ld = {Q1, Q2, . . . , Qn} Scomp(Qi, Rj) = Salign(Qi, Rj) × e

1 rank(Rj )

S(Qi) = K

k=1 Scomp(Qi, Rk) × 1 K × Sraw(Qi)

14 / 18

Introduction Method Experiments Dataset Results

Dataset

Protein samples:

• Alpha-1-acid glycoprotein (Bovine)
• Ovomucoid (Chicken)
• Ig gamma-3 chain C region (Human)

Thermo Scientific Orbitrap Elite hybrid mass spectrometer HCD fragmentation technique GlycoMaster DB was used for comparison 46 HCD spectra of glycopeptides were contained

15 / 18

Introduction Method Experiments Dataset Results

Experimental Results

Software program: GlycoNovoDB

Table: Performance of De Novo Sequencing Algorithm and

GlycoNovoDB Compared with GlycoMaster DB

Rank1 De Novo Sequencing Algorithm GlycoNovoDB Number Percentage(%) Number Percentage(%) 1 40 86.96 45 97.83 2 1 2.17 1 2.17 3 1 2.17 0.00 4 ∼ 10 1 2.17 0.00 > 10 2 4.35 0.00 can’t find 1 2.17 0.00

1”Rank” refers to the ranking status of the reference structure (the top

glycan structure reported by GlycoMaster DB) in our results for a spectrum.

16 / 18

Introduction Method Experiments Dataset Results

Experimental Results

GlycoNovoDB can report more confident results than GlycoMaster DB.

• There are 6 spectra that GlycoMaster DB reported more than
• ne top-ranked glycans with the same score.
• An example:

17 / 18

Introduction Method Experiments Acknowledgement

Thank you!

Questions?

University of Western Ontario

• Prof. Kaizhong Zhang
• Prof. Gilles A. Lajoie
• Dr. Weiping Sun
• Dr. Yi Liu

University of Waterloo

• Prof. Bin Ma

This work was supported in part by the NSERC Discovery Grant and a Discovery Accelerator Supplements Grant.

18 / 18