analytical chemistry: from data (pre)-treatment
TRANSCRIPT
EA4041, Groupe de Chimie Analytique de Paris-Sud ‘GCAPS’
Analytical chemistry: from data (pre)-treatment optimization to data mining, data fusion and
big data
April 1st, 2014
Cell membrane lipidomics
Lipids in skin barrier
Lipids: from natural
substances to heritage objects
Lipid analogues for
diagnostic and
therapeutic aims
Four principle themes
Data (pre)-processing Multivariate analysis
Chemometric techniques
Different analytical tools
Data acquisition
Separation techniques,
mainly chromatographic
Coupled mass spectrometry
techniques (LC, GC, GCxGC/MS)
Vibrational spectroscopy
(infrared, near infrared and
Raman)
Schedule
4
10h00-10h30: «Analytical chemistry: cell membrane lipidomics and data analysis», Sana Tfaili 10h30-11h00: «Analytical chemistry and chemometrics: a tool for skin physiological and physiopathological charecterization», Ali Tfayli 11h00-11h10: «Analytical chemistry: from data (pre)-treatment optimization to data mining, data fusion and big data», S&A Tfai(y)li
Lipidomics
• Recent comparing to other omics
• Borrows heavily from metabolomics
• Specific analysis of lipids
6
Analytical chemistry: cell membrane lipidomics and data analysis
Infectious diseases
•Leishmaniasis
•Impact of treatment on the lipid composition of membranes Leishmania
donovani
Collaboration UMR BioCIS / Ph. Loiseau
Analytical chemistry: cell membrane lipidomics and data analysis
Hereditary diseases and RBCs
•Lipid composition and sickle cell disease
Olivier Blanc-Brude Paris Centre de Recherche Cardiovasculaire (PARCC)
•Lipid composition and Gaucher disease (LETIAM)
Institut National de Transfusion Sanguine (Pr. Le Van Kim et Dr. M Franco)
Service de Neuro-Pédiatrie de l’Hôpital Trousseau (Pr. T. Billette de Villemeur et Dr
C. Mignot)
Analytical chemistry: cell membrane lipidomics and data analysis
Lipidomics macrophages & Atherosclerosis
•Impact of membrane incorporation of w3 PUFA on cholesterol efflux from macrophages
•Intracellular trafficking of cholesterol from
macrophages:study of the inhibition of Rab7 and the role of oxysterols
Collaboration EA4529 (cross-cutting theme in the
future unit Lip (Sys) ²)
Lipids and polarity
Solubility parameter d d’Hildebrand
5 10 15 20 25 0
O
O
O
O
HO
O
O O
O
O
H
H
H
O
H
H
O
H
H
H
OH
H
H
HN
OH
O
OHH
HN
O
O
OHH
O
OO
O
O
O
O
HO
O
R1
R2
OOH
O
OHOH
OOH
OH
OHOH
O
O
O
HO
R1
O ONH
2
P
O
O
O
R2
O
O
HO
O ON
+P
O
O
O
R2
R1
O
O
HOH
R O ON
+P
O
O
11
Normal Phase / Reversed Phase Liquid chromatography LC
5
10
1
5
20
2
5
0
d
Solv
ent
stre
ngt
h
Solven
t strength
Lipids are separated according to their polar moeties.
Lipids are separated according to their chain length & number and position of double bonds
LIP
ID C
LASS
ES A
NA
LYSI
S
LIP
ID M
OLE
CU
LAR
SP
ECIE
S A
NA
LYSI
S
12
13
Mass spectrometry MS
e-
ABC ABC
e-
e-
+ Ionisation IE
E interne AB+ + C
fragmentation (BC+)* + A OU
B+ + C Mass spectrometry by electronic impact
(GC-MS)
LC-MS acquisitions: Different ionization modes
+ - OOO
OOO
OOO
SA
SA
SA
Laurent Imbert, PhD thesis, GCAPS,EA4041, 2012, Univ. Paris Sud
Lipid molecular species
Lipidomics: coming to grips with lipid diversity Andrej Shevchenko & Kai Simons Nature Reviews Molecular Cell Biology 11, 593-598 (August 2010)
Lipid Classes LC MS
Sub-classes LC MS or in HRMS LC/MS
Isobars LC MS LC/MS
Isomers LC MS LC/MS
15
LC-MS data matrix
Data = 3D matrix need to concatenate and "unfold" files
In the data matrix: Objects (lines) = sample
Variables (columns) = couple (Tr, m/z)
LC-MS data matrix processing
Univariate statistical analysis using XCMS online: Paired Student t-test between the two
groups of signals
Evident significant difference between the intensities of the ions.
Step 1: file by file, detection of ions (> threshold) scan by scan
Step 2: ion chromatogram generation, file by file
Step 3: file by file, peak detection table (ion; rt, intensity) (data file)
Step 4: Alignment: Setting a tolerance window (m/z, rt) based in general on the first chromatogram. (data set)
18
LC-MS data matrix processing
Alignment tools using MzMine
Principal component analysis Unsupervised method Data mining PLS Discriminant analysis Supervised method
http://fiehnlab.ucdavis.edu/staff/kind/Statistics/Concepts/OPLS-PLSDA
http://www.nlpca.org/pca_principal_component_analysis.html
19
LC-MS data matrix processing
Orthogonal PLS-DA (OPLS-DA)
Maximizes the discrimination between two classes in its first component
S-Plot expresses the relationship between the original variables (rt, m/z)
and scores on the selected axis.
Published in: Susanne Wiklund; Erik Johansson; Lina Sjöström; Ewa J. Mellerowicz; Ulf Edlund; John P. Shockcor; Johan Gottfries; Thomas Moritz; Johan Trygg; Anal. Chem. 2008, 80, 115-122.
20
LC-MS data matrix processing
Chemometric tools for LC-MS lipidomics profiles analysis.
Samples
LC-MS profiles
Highlighting specific MS signals of the 2 groups
Stat
isti
cal
anal
ysis
Biomarkers
Analytical chemistry and chemometrics: a tool for skin physiological and
physiopathological characterization
Ali Tfayli
Largest organ in human body Barrier function
Sensation
Heat regulation
Secretion end excretion
Dermis Epidermis
- Stratum Basale - Struatum Spinosum - Stratum Granulosum
- Stratum Corneum
- Superficial - Deep
General structure of the skin
Introduction
28
Skin Barrier function
Introduction
Barrier against
External insult
Water loss
STRATUM
CORNEUM
Keratin
Intercellular lipids
SC structure
29
Skin Barrier function
Introduction
CHOLESTEROL
CERAMIDES
Composition of SC lipids
OTHERS
FATTY ACIDS
Structure of Ceramides
Different polar heads
Different chain lengths
Presence of double bonds
30
Skin Barrier function
Introduction
Conformational order
Lateral packing of SC lipids
trans
gauche
Orthorhombic Disordered Hexagonal
HIGHLY ORGANIZED
GOOD BARRIER PROPERTIES
31
Cosmetology
Toxicology
Dermatology
Characterization of skin barrier
Skin aging
-Physiological status
-Physiopathological status
Skin hydration and dry skin diseases
Mechanical stress
32
Analyses of skin barrier
Composition and profiling
Organization, lateral packing
Vibrational spectroscopies: Infrared and Raman
Separative techniques - mass
33
Vibrational spectroscopies
µ = BA
BA
mm
mm
.
k
CC 2
1_
ν = nombre d’onde, ν = fréquence, µ = masse réduite
INTERACTION
RAYONNEMENT - MATIERE
ABSORPTION EMISSION
DIFFUSION
SPECTROSCOPIES OPTIQUES
6N3
1nn0n00
n
0
0000
6N3
1nn0
n
p
0ptotttQ
dQ
dE
2
1tEtQ
dQ
d
coscos)cos(cos
)(34
Vibrational signal collection
Individual spectral collection In depth spectral collection
0 10 20 30 40 50 60 70 80 90 0
0.01
profondeur (µm)
inte
nsité
(a
.u.)
x: profondeur y: intensité du pic à 1191 cm-1
36
Data pre-processing
-Dark current
-CCD response correction
-Optical components contribution
-Smoothing
-Baseline correction
-Normalization
38
Physiological / Physiopathological status
Raman descriptors of SC barrier
Vibrational spectroscopies VS barrier function
Structural information
Amide I band
dCH3 Rocking
Intensity (a.u.) 20 15 10 5 0
2700
2800
2900
3000
3100
Wav
enu
mb
er (
cm-1
)
800
1000
1200
1400
1600
1800
(CH2) (CH3)
(C-C)
dCH2 scissoring
Conformational order
Compacity of packing
Chain end conformers
Organisation
Polar heads interactions
Chain conformation
39
Ex vivo et in vivo
TFAYLI A. et al. EJD 2012
Volontaires
13 F & 7 M (22 à 64 ans)
Méthode
Surface nettoyée
2ème
acquisition in vivo
1ère
acquisition in vivo
Extraction des lipides
Ex vivo
Pas de séparation
Spectres in vivo avant l’extraction
–
Spectres in vivo après l’extraction
=
Signal in vivo des lipides
In vivo
Physiological / Physiopathological status
Skin aging
Vibrational spectroscopies VS barrier function
40
in vivo
TFAYLI A. et al. EJD 2012
Observations directes
2800 2850 2900 2950 3000
0
0.02
0.04
0.06
0.08
0.1
0.12
Nombre d’ondes (cm-1)
Inte
nsi
té (u
.a.)
Peau jeune
Peau mature
kruskal Wallis
2845-3020 cm-1
Physiological / Physiopathological status
Skin aging
Vibrational spectroscopies VS barrier function
41
in vivo
-1 0 1 2 3 4
-3
-2
-1
0
1
2
PC1
PC2
32M
32M
60M
60M
59F
59F
22M
33F
33F
42F
42F
30M
30M 27F
27F
26F
26F
58F
58F
27F
28M
28M
26F
26F
24F
24F 39F
42M
42M
31F 31F
64M
64M
Analyse en composantes principales sur la gamme:
2845-3020 cm-1
Trois groupes en fonction de l’âge: •22-30 ans •30-42 ans •50-64 ans
TFAYLI A. et al. EJD 2012
Physiological / Physiopathological status
Skin aging
Vibrational spectroscopies VS barrier function
42
Water structure
RH : 2.5% » 75%
Ex vivo
VYUMVUHORE R. et al. Analyst, 2013
Nombre d'Ondes (cm-1)
Inte
nsi
té (u
.a.)
500 1 000 1 500 2 000 2 500 3 000 3 500
0.0
0.2
0.4
0.6
0.8
1.0
RH=4%
RH=75%
RH=98%
Eau
Physiological / Physiopathological status
Skin hydration / dry skin diseases
Vibrational spectroscopies VS barrier function
43
VYUMVUHORE R. et al. Analyst, 2013
3465.8
3343.3
3280.0
3212.0
1200
1000
800
600
400
200
0 3200 3300 3400 3500 3600
Inte
nsi
té (u
.a.)
Nombre d’Ondes (cm-1)
Eau fortement liée
Eau partiellement liée
Eau non-liée
Ex vivo
Water structure
Physiological / Physiopathological status
Skin hydration / dry skin diseases
Vibrational spectroscopies VS barrier function
44
0 10 20 30 40 50 60 70 80
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Rap
po
rt: T
ran
s/ga
uch
e
Humidité Relative (%)
ν C-C (1060+1130/1080 cm-1)
Ex vivo
Conformation des chaînes lipidiques
VYUMVUHORE R. et al. Analyst, 2013
Structure secondaire de la kératine:
bande Amide I
Physiological / Physiopathological status
Skin hydration / dry skin diseases
Vibrational spectroscopies VS barrier function
45
Tension mécanique du SC Bande totale νOH
Ex vivo
VYUMVUHORE R. et al. J. Biophotonics, 2014
Courbure du substrat
due au séchage
sx
sy
Lamelle en verre
SC hydraté
séch
age
Physiological / Physiopathological status
Hydration and mechanical stress
Vibrational spectroscopies VS barrier function
46
0 1 2 3 4 5 6 7 81.1
1.15
1.2
1.25
Rapport
: S
1060 /
S1080
Temps (h)
Ex vivo
01
23
45
67
81.1
1.1
5
1.2
1.2
5
Rapport : S1060 / S1080
Tem
ps (h
)
0 1 2 3 4 5 6 7 8 37
37.5
38
38.5
39
39.5
40
40.5
41
41.5
42
Aire Sous Courbe 1652 (%)
Temps (h)
Hélices α
0 1 2 3 4 5 6 7 8 25
25.5
26
26.5
27
27.5
28
28.5
Aire Sous Courbe 1671 (%)
Temps (h)
Feuillets β
Conformation des lipides Structures secondaires des protéines
VYUMVUHORE R. et al. J. Biophotonics, 2014
Physiological / Physiopathological status
Hydration and mechanical stress
Vibrational spectroscopies VS barrier function
47
Pics discriminants
Analyse statistique multivariée
2 3 4 5 6 7 8 9 10 11 12 13
2 3 4 5 6 7 8 9 10 11 12 13
Yva
r (T
ract
ion
(%
))
Ypred (Traction (%))
Régression PLS , Modèle + Traction 4%, 7%, 10%
y=1*x+4.556e-007 R2=0.9903
Interprétation structurale
État moléculaire lié au stress mécanique
VYUMVUHORE R. et al. J. Raman spectroscopy, 2013
Ex vivo
SC mechanical strains
Vibrational spectroscopies VS barrier function
48
Deux domaines
zone élastique
zone plastique
Compacité des lipides
Structure secondaire des protéines
VYUMVUHORE R. et al. J. Raman spectroscopy, 2013
SC mechanical strains
Vibrational spectroscopies VS barrier function
49
Analyse multi-paramétrique du SC
1. Identification des relations entre les différents paramètres
du Stratum Corneum
2. Développement d’un outil multi-informationnel
Patients
pH : pH mètre
PIE : Tewl-mètre
Hydratation globale du SC : Cornéomètre
Composition lipidique : Chromatographies
Information moléculaire + profondeur : Raman
11 volontaires F sains
âgés de 57 à 62 ans
Analyse sur bras et
mollet
In vivo
VYUMVUHORE R. et al. JBO, 2014
Physiological status
“QR code” of the skin
50
∑
=
CER AG Chol PIE Hydr pH
Y1
Y2
Yn
Paramètres de la Peau
No
uve
aux
Ech
anti
on
s
BCER BAG BChol BPIE Bhydr BpH
Coefficients de Régression (B)
X1
X2
Xn
Nouveaux Spectres Raman
Xm : spectre moyen Xstd : écart-type des spectes Ym : moyenne de Y Ystd : écart-type de Y
Données centrées-réduites; besoin de:
Modèle de prédiction
VYUMVUHORE R. et al. JBO, 2014
Physiological status
“QR code” of the skin
51
Sources de peau
Peau animale
Peau humaine
Restriction Interdiction
Manque de reproductibilité
disponibilité
Restriction
Peau synthétique
Culture de keratinocytes
Raman et peau synthétique
HDR: 05 février 2014
52
Evaluation des peaux
Validité en tant que substits
Morphologie
marqueurs biomécaniques
Tests d’irritation
Tests de phototoxicité
Composition protéique
Composition lipidique
Perméabilité
Perméabilité PLUS ÉLEVÉE
Les classes lipidiques majoritaires
sont présentes
Comparaison de la composition et de
l’organisation des lipides
Raman et peau synthétique
HDR: 05 février 2014
53
Composition des lipides
Composition globale
Cholestérol
esters de cholestérol
triglycérides
Acides gras
céramides
Peau humaine: bleu
Peau synthétique: noir
Raman et peau synthétique
HDR: 05 février 2014
54
Distribution hétérogène des lipides
Imagerie Raman + NCLS
-Taille de l’image: 600*600 µm2
-Taille du pixel: 4 µm
-Pas: 20 µm
Kératine Cholestérol Cer. et acides gras
Stratum corneum
reconstruit
Raman et peau synthétique
HDR: 05 février 2014
55
Coupes de SC: imagerie Raman + NCLS
Stratum corneum humain
Stratum corneum
synthétique
1. Lumière blanche
2. Kératine
3. Cholestérol
4. Cer. et acides gras
Raman et peau synthétique
HDR: 05 février 2014
56
Analytical chemistry: from data (pre)-treatment optimization to data mining, data fusion and big data
Sana et Ali Tfai(y)li
Raman analyses: correlation with LC/GC/MS.., biometric data
pH, Hydration, TEWL,…
LC/
GC
/ M
S..
Physiopathological state: Diagnosis
… In vivo
Raman
59
LC-MS acquisitions: Different ionization modes (Data fusion?)
+ - OOO
OOO
OOO
SA
SA
SA
Laurent Imbert, PhD thesis, GCAPS,EA4041, 2012, Univ. Paris Sud
61
Data fusion and data mining
Perspectives: Data fusion between: • RPLC and NPLC in chromatography • Different ionization modes in mass spectrometry • Increase the separation dimensionality LCxLC MS… (new treatment approach) • Between Separation techniques, coupled mass spectrometry • Between several techniques: Raman, IR, separative techniques • Multi-block analysis (specific algorithms for data processing and fusion)?
Additional analysis will increase the time for data processing: other approaches
for data processing ?
Virtual data project (LAL): work on a cloud and save the image Buy cores (possible demand through a project in process)