intro to data preprocessing€¦ · data preprocessing workflow 20 gc-ms lc-ms detect masses from...

Xiuxia Du, Ph.D. Department of Bioinformatics and Genomics

University of North Carolina at Charlotte

Introduction to Preprocessing of Untargeted Metabolomics Data

Outline• Raw untargeted LC/MS and GC/MS metabolomics data

- Profile and centroid data

- Mass vs. retention time map

- TIC

- EIC

• Principles of LC/MS and GC/MS data preprocessing

• Feature identification

- Identification of known compounds

- Identification of unknown compounds

2

3

Raw Untargeted LC/MS and GC/MS Metabolomics data

4

list of scans in raw files• MS scans in blue• MS/MS scans in red• # sequential number• @ retention time• MS level• type of spectrum

• p = profile• c = centroid• t = thresholded

• polarity of ionization• + = positive• - = negative• ? = unknown

List of mass spectra

One mass spectrum

5

One mass spectrum

6

One mass spectrum

7

Zoom in one mass spectrum

8

profile mode

Mass spectra in centroid mode

9

Mass spectra in centroid mode

10

Spectrum in centroid mode• Data files are much smaller than files in profile mode.

• We will use the centroid data for practicing data pre-processing using XCMS and MZmine 2.

11

LC-MS raw data in 3D

12

Raw data in 3D

13

3D to 2D• Direct processing of the 3D data is NOT trivial

• Instead, we examine 2D

- Mass vs. retention time

- Total ion current vs. retention time: TIC

- Ion current vs. retention time for a particular mass: EIC (Extracted Ion Chromatogram)

14

Mass vs. retention time map

15

TIC

16

EIC

17

EIC

18

19

Principles of LC/MS and GC/MS Data Preprocessing

Data preprocessing workflow

20

GC-MS

LC-MSdetect masses from

mass spectra

detect masses frommass spectra

constructEICs

constructEICs

detectchromatographic peaks

detectchromatographic peaks

deconvolution

annotation

databasesearch

databasesearch

1 2 3 4 5 6alignment /

correspondence

alignment / correspondence

Construct EICs

21

Select one EIC

22

One EIC

23

Detect EIC peaks

24

• Use wavelet transform

• Implemented in XCMS as the centWave method

Detect EIC peaks

25

mexican hat wavelet

−20 −10 0 10 20

−0.4

0.0

0.2

0.4

0.6

0.8

t

ψψs,ττ((

t))

s = 1s = 2s = 8

68

1012

1:dim(wCoefs)[1]

Scal

e

+

++

2850 2900 2950 3000 3050 3100 3150

02

46

810

Seconds

Inte

nsity

* 10

3

ChromatogramGaussian Fit

Tautenhahn, R.; Bottcher, C.; Neumann, S., Highly sensitive feature detection for high resolution LC/MS. BMC bioinformatics 2008, 9, 504.

Detected EIC peaks

26

27

LC/MS-specific Data Preprocessing

Find isotopes

28

Find isotopes

29

Find isotopes

30

31

Alignment

zoom in……

32

Alignment

33

Alignment

Peaks table after alignment

34

35

GC/MS-specific Data Preprocessing

GC-EI-MS

36

• Example: EI fragmentation of methanol

EI fragmentation

37

[CH3OH]•+ �! CH3O+ +H•

[CH3OH]•+ �! CH2O+ +H2

[CH3OH]•+ �! CH+3 + •OH

Deconvolution

38

41

ADAP-GC 2.0

ADAP-GC 2.0: Deconvolution of Coeluting Metabolites from GC/TOF-MS Data for Metabolomics Studies. Analytical chemistry 2012, 84 (15), 6619-29.

42

Feature identification

Feature identification• Apply statistics and machine learning to detect

discriminating peaks

• Identify discriminating peaks

43

Identification of known compounds• Screening search for compound ID based on LC-MS

data

- Searching monoisotopic mass and isotopic distribution against compound databases

• Library match for compound identification from both LC-MS/MS and GC-MS spectra

44

HMDB

45

HMDB

46

MS/MS or GC-MS spectra matching• Library match for compound identification from both

LC-MS/MS and GC-MS spectra

47

Identification of unknown compounds• MS-FINDER

• CSI:FingerID

• CFM-ID

• MetFrag

• MIDAS

• MAGMA

48

MetFrag

49

intro to data preprocessing€¦ · data preprocessing workflow 20 gc-ms lc-ms detect masses from...

Documents