smcr: a new approach to recovering temporal metabolic ...process unknown • retardation of...

Nestlé Research Center

SMCR: A new approach to recoveringtemporal metabolic signal modulation inNMR spectroscopic datasets

Selena Richards

Application to a life-long caloricrestriction study in dogs


Caloric Restriction• Natural intervention

• Increase longevity• Reduce onset and prevalence of late

life diseases• McCay et al. 1930’s

• Yeast, worms, fruit flies, rodents anddogs

• Nestle Research Centre• CR in Labrador Retrievers

• 1.8 years longer median lifespan• Osteoarthritis and neoplastic diseases

• Biochemical and physiologicalprocess unknown• Retardation of cellular, DNA and

macromolecular damage

Caloric RestrictedCaloric RestrictedControl Fed

Aged 6 years Aged 6 years


Dynamic biofluid

• Inter-subject variability is large• Severely complicates retrieval of statistically valid

biomarkers• Variation in recency of dietary intervention and

differences between individuals metabolic phenotype• Identify an individuals healthy state

• Advantages• Readily available & relatively non-invasively• Rich source of nutrients transit from one organ to

another• Reflect current state of health and disease

• Nutritional metabonomics (1H NMR blood serum)• Distinctive biomarkers associated with dietary intervention

• Dynamic concentration changes in lipids, lipoproteins and ketone bodies assystem maintains homeostasis

• Mask subtle systematic changes associated with diet


Objectives

YoungYoung

OldOld

Aging

11H NMR metabolic profileH NMR metabolic profile

Expression Expression of diabetesof diabetes


ObjectivesM

etab

olic

con

stitu

ents

δ ppm

1

2

3

Lipidic

SugarProtein


sugarsLife-long trajectoryLife-long trajectory

Objectives

Old

Time (years)0 13

lipidic

proteins

Young Middle aged


Self Modeling Curve Resolution


• Underlying Principle of SMCR• Bilinear model

• Initial Estimates• Quantitative Iterative Target Transformation Factor Analysis

(QITTFA)• Starting estimates approximate the final solution• Advantages

• Refinement of Initial estimates prior to ALS• Absence of unstructured variance

• Five Steps of the QITTFA routine

SMCR Theory

sT1 sT2= +c1 c2

E = ||D - C ST||

D + E

Spectrotype 1 Spectrotype 2

Richards, S. E.; Walmsley, A. D. Journal of Chemometrics 2007, 22, 63-80.


Initial Estimates (QITTFA)1. Needle Spectrum

Uni

t int

ensi

ty

5.0 ppm

δ ppm

in1 = (1, 0 0,….0)

2. Singular Value Decomposition

vT1= t1D vT2

t2

4. Needle Output Spectra Constrained

in’1 = out’1

inte

nsity

out1 = in1 V* V*T

3. Needle Output Spectra

inte

nsity

NegativityNegativity


Initial Estimates (QITTFA)

δ ppmout No.

5. Selection of Output Spectra(Initial Estimates)

Needle Output Matrix

out1 = in1 V* V*T

out2 = in2 V* V*T

outm = inm V* V*T

Which is the purest Which is the purest outout (initial estimate)? (initial estimate)?

out450

δ ppm

I

SIMPLISMA

Initial estimate spectrum No. 1Initial estimate spectrum No. 1

Purity spectrumPurity spectrum

p

out450

Out no.

!+=

j

j

jx

sp

α


Validation of SMCR

Rotational AmbiguityRotational AmbiguityD D == C C SSTTDD = ( = (CCTT) () (TT-1-1SSTT))

Intensity AmbiguityIntensity Ambiguity nnDD= = ΣΣ (1/ (1/kkiiccii) () (kkiissiiTT)) i=1 i=1

E = ||D - C ST||

ST = C+ D

C = D (ST)+

Constrain? Constrain?

Alternating Least Squares (ALS)Alternating Least Squares (ALS)

Constrained ( Constrained ( --- ) ) Unconstrained (Unconstrained (- -))

δ ppm

I

δ ppm

I

Poor matchPoor matchActive constraints

Lack of fit (LOF)Lack of fit (LOF)

!

LOF(%) =100

dij" ˆ d

ij( )2

ij

#

dij2

ij

#

Good matchGood matchFew active constraints

!

r2(%) =100"

ˆ d ij

2

ij

#

dij

2

ij

#

% Variance explained% Variance explained


Experimental


Experimental Design & Analysis

• Animal Handling• 48 dogs

• paired gender and weaning weight• Randomly assigned (CF/CR)• Initiated 8 weeks• CR (75% of CF)

• NMR spectroscopy• 400 µl of saline (10% D2O)in 200 µl blood plasma• Bruker DRX 600 NMRspectrometer

• 600.13 MHz for 1H• NOESY

• Binned Data (0.005 ppm)

Caloric RestrictedCaloric RestrictedControl Fed


Experimental Design & Analysis

• Multivariate Metabolic Trajectory• PCA Trajectory plots• Reveal patterns and trends in the data• Scaled to UV

• SMCR Analysis• QITTFA Initial Estimates

• Non-negativity constraints, maximum iteration 500• ALS

• Non-negativity constraints in C and S and normalization of S


Results and Discussion


Multivariate Metabolic Trajectory (PCA)

Metabolites associated with the first two PC indicators of diet NOT aging

Control fed (CF)

Caloric Restricted (CR)


Initial Estimates

Lipoprotein fatty acyl(CH2)n

CH=CH

C=CH-CH2-CH=C

-CH2-CO

-CH2-C=

Lipoprotein fatty acyl(CH3)

Cholesterol in HDL

Phosphatidylcholine-N(CH3)3

Phosphatidylcholine-OCH2

Phosphatidylcholine-NCH2

Dominantly Lipidic

GlucoseGlucose


Initial Estimates

Sugar-protein

Lactate

Lactate Citrate

Glucose

Alanine

N-Acetyl glycoprotein

Albumin Lysylgroups

Mainly albuminCH3


Final SMCR SolutionDominantly Lipidic

Sugar-protein

Lipoprotein fatty acyl(CH2)n

CH=CH

C=CH-CH2-CH=C

-CH2-CO

-CH2-C=

Lipoprotein fatty acyl(CH3)

Cholesterol in HDL

Phosphatidylcholine-N(CH3)3

Phosphatidylcholine-OCH2

Phosphatidylcholine-NCH2

Glucose Glucose

Lactate

LactateCitrate

Glucose

Alanine

N-Acetyl glycoprotein

Albumin Lysyl groups Mainly albumin CH3


Final SMCR ResultsContour plot of the R2 Correlation Coefficients

Cholesterol in HDL

Lipoproteins

Phosphatidylcholine

Glucose


Semi-quantitative trajectory

Period 1 (puppies) Period 2 (middle aged) Period 3 (elderly)

Dominantly lipidic (CF)

Dominantly lipidic (CR)

Sugar-protein (CF)

Sugar-protein (CR)


Conclusion


Conclusion

• Identified predominant sources of variation• No a priori information

• Pinpointed age groups where aging and dietbecame significant• Age groups which were phenotypically different

• Addition of new chemometric tool• Metabonomics toolbox

• Diverse application with other biomedical problems• Subtle time dependent changes

Richards, S. E.; Wang, Y.; Lawler, D.; Kochhar, S.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Analytical Chemistry 2008, 80, 4876-4885


Acknowledgments

• Imperial College• Yulan Wang• Elaine Holmes• John Lindon• Jeremy Nicholson• Anthony Maher• Olaf Beckonert

• Nestle Research Centre• Dennis Lawler• Sunil Kochhar• Ziad Ramadan

• Funding• Nestle Research Centre

smcr: a new approach to recovering temporal metabolic ...process unknown • retardation of...

Documents