Contaminant profiling: A set of next generation analytical tools to deal with contaminant complexity

Contaminant profiling: A set of next generation analytical tools to deal with contaminant complexity Associate Professor Jan H Christensen Department of Basic Sciences and Environment Slide 1

Risk Assessment How can we improve current risk assessment procedures? (1) Measure the entire contaminant complexity ( the chemical fingerprint ) (2) Investigate combination effects / Cocktail effects Comprehensive investigation of single compounds Chemical analysis and effect studies Slide 2

A paradigm shift in (environmental) analytical chemistry? Top-down approach Traditional mechanistic or bottom-up approach Extensive sample preparation (to isolate) Chemical target analysis Fully quantitative analysis Extrapolation to complex systems Limited sample preparation Chemical profiling analysis (semi)quantitative analysis or patterns (fingerprints) Identification of relevant compounds Slide 3

Analytical data are multivariate Slide 4

a trend that is increasing Slide 5

Human pattern recognition Kevin Thomas Barack Obama Slide 6

Human pattern recognition Slide 7

Some mathematics + = It is Elvis and not Berlusconi eyes + Brain + = It is oil pollution Sensor + Computer with multivariate software Slide 8 8

PCA the most common multivariate method Height and shoe size of 19 persons. Each sample (person) can be described by two variables (height and shoe size) Height Shoe size 1.98 44.5 1.57 36.0 2.03 45.0 2.11 46.5 2.02 42.5 2.07 43.5 1.98 46.0 1.72 42.0 1.86 44.0 1.64 36.5 2.02 43.5 1.57 36.0 1.80 40.0 1.57 35.0 1.73 35.5 1.98 44.5 1.89 44.5 1.71 40.5 2.10 47.5 Shoe Size Skostr 48 46 44 42 40 38 36 34 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 Højde Slide 9 Height

PCA projection (Scores) The first score (t 1 ) is found by projection of the original points onto the 1. loading (PC1 loading). Now each sample is represented by one number (t 1 ). That s a model! 2 PC1 loading t 1 0.9895-1.9458 1.2485 1.7989 0.7963 1.1601 1.2533-0.3647 0.4633-1.6113 0.9784-1.9535-0.4184-2.1039-1.4234 1.0027 0.6545-0.6732 1.9366 Shoe size Skostr (scaled) (skaleret) 1.5 1 0.5 0-0.5-1 -1.5-2 -2-1.5-1 -0.5 0 0.5 1 1.5 2 Højde (skaleret) Height (scaled) Slide 10

Principal Component Analysis (PCA) X = TP' One Excel sheet = Two new Excel sheets Scores: Info about samples Loadings: Info about measurements Slide 11 Lars Nørgaard, lan@life.ku.dk / 11

Sensors: Analytical data 1. Spectroscopy Ultraviolet-visual Near-infrared, FT-Infrared, Raman Fluorescence (excitation or emission spectra) Fluorescence (time-resolved, excitation-emission landscapes) Nuclear Magnetic Resonance 2. Mass spectrometry Ionization (Desorption, spray and gas phase methods) Mass analyzers (nominal vs. accurate mass) 3. Chromatography Liquid chromatography (LC, HPLC, UPLC) Gas chromatography (GC) Electrophoresis (migration) Univariate detectors (ECD, TCD, sulphur, FID etc) 4. Hyphenated techniques Liquid chromatography spectroscopy (fluorescence and UV/VIS) Gas or liquid chromatography with mass spectrometry detection (LC-MS, GC- MS) Two-dimensional chromatography (GC GC, LC LC) Slide 12 Lars Nørgaard, lan@life.ku.dk / 12

Example 1: Oil hydrocarbon Fingerprinting Christensen JH, Tomasi G and Hansen AB. Chemical Fingerprinting of Petroleum Biomarkers using Time Warping and PCA, Environmental Science and Technology, 2005, 39 (1), 255-260. Slide 13

Conventional methods rely on peak integration Slide 14

The CHEMSIC method (How to avoid peak integration) Part A): Sample Preparation and chemical analysis Building a sample database Part B): Data processing Data import and cropping Baseline removal Retention time alignment Variable selection Normalization Part D) Chemometrics Slide 15

PART A: Chemical analysis (SIC) PART B: Data pre-processing Step I: Baseline removal The baseline is removed by calculating the first derivative of the chromatograms Slide 16

PART B: Data pre-processing Step II: Retention Time Alignment The effects of correlation optimized warping (COW) on a chromatographic section of m/z 217 SIC in 10 oil samples Slide 17

PART B: Data pre-processing Step III: Normalization (I) Internal standards (II) x N nj x J nj j 1 x 2 nj (III) Normalization using selected data points - Less affected by closure - Less sensitive to noise Slide 18

PART C: Data structure Chemical profiles of 100 crude oils or refined produts Size of data set (excel sheet): 100 samples x 3600 RTs = 360,000 data points Relative abundance Slide 19

PART C: Map of samples (the score plots) Slide 20

PART C: Chemical interpretation (PC2 loadings) Distinguish oils originating from source rocks with different clay content Why are the oils different? RS RS RS RS High PC2 High DAS/RS High clay content Abundance Low PC2 Low DAS/RS Low clay content and anoxic conditions DAS DAS DAS DAS Retentions time (min) Slide 21

PART C: Chemical interpretation (PC4 loadings) Distinguish oils originating from source rocks with different thermal maturity Why are the oils also different? High PC4 High / ratio of C 27 -C 29 regular steranes Low thermal maturity Seems to be independent of source organic matter input Slide 22

Modifications to the CHEMSIC method adding more SICs (information) RS 29ab RS RS RS Tm Ts 30ab DAS DAS DAS 28b DAS Slide 23

Example 2: Does the CHEMSIC method also work under warmer climates? January 2000: 4 million litre of crude oil escaped from a burst underwater pipeline at the Petrobras refinery into and Guanabara Bay Slide 24

PART C: Map of samples (the score plots) Slide 25

PART C: Chemical interpretation (the loading plots) PC2 PC3 Slide 26

PAH pollution sources, patterns and gradients Slide 27

Modifications to the CHEMSIC method and more SICs (information) Naphthalenes Dibenzothiophenes Fluorenes Dibenzofurans Phenanthrenes/Anthracenes C1-BNT Flt/Pyrenes B(a)A/Chrysenes 5/6-Ring PAH 0.8 0.6 0.4 [A.U.] 0.2 0-0.2-0.4 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0.6 D ata poin ts 0.4 0.2 [A.U.] 0-0.2-0.4-0.6 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 D ata poin ts Slide 28

and to GCxGC data Archetypal analysis Slide 29

Concluding remarks Contaminant profiling is one way to deal with contaminant complexity and it is complementary to conventional quantitative analysis of selected contaminants New tools are under development and we are applying them to numerous different areas - Source apportionment - Tracking bioremediation efficiency - Environmental metabolomics - Risk assessment of contaminant mixtures (correlation to toxicity) - Slide 30

Acknowledgements Post Doc Giorgio Tomasi Lab technician Jette Petersen Assist Prof Nikoline Nielsen Post Doc: Raquel F Varela PhD. Stud: Nichlas Weezel PhD. Stud: Linus Malmquist PhD. Stud: Sarah Mohamed PhD Stud: Esther S Boll PhD Stud Søren Furbo PhD Stud Rune Græsbøl Slide 31 PhD Stud Majbrit Hansen PhD Stud Kuba Modrzynski PhD Stud Peter Christensen