Transcript Development of a UV-Vis Spectral Model for the American

Phenolics and Tannin Assays
for Practical Use in
Winemaking
Giovanni Colantuoni
John Thorngate
Research and Development
Outline
Introduction
 Grape and Wine Phenolics
 Measuring Phenolics
Adams-Harbertson Assays
 Gage R&R Analysis
 Creating a Standardized SOP
The UV-Vis Predictive Model
 Chemometrics — Model Calibration and Deployment
 Comparison to Skogerson-Downey-Boulton
 Using the Model
Summary
Research and Development
Chemists interested in polyphenols, in common
with the majority of scientists, tackle today’s
problems with yesterday’s tools, i.e., current
problems are attacked with methods which are
inadequate and to that extent are already out of
date.
The discovery and quick application of new methods
or developments and extensions of existing
methods is therefore of first importance.
B.R.Brown, In Methods of Polyphenol Chemistry, 1964
Research and Development
Introduction
Why focus on phenolics?
Important for:
Color
Taste
Mouthfeel
Wine aging
Research and Development
Introduction
Why measure phenolics?
Identify higher quality lots more
easily
Use phenolic data for:
Press decisions
Heavy press additions
Blend balancing
Evaluation of processing
Research and Development
Grape and Wine Phenolics
Phenolic compounds of interest to the
winemaker:
Phenolic acids
Flavonoids
Anthocyanins
Tannins
Polymeric Pigment
J.A. Kennedy, Grape and wine phenolics: Observations and recent findings,
Ciencia e Investigación Agraria 35:77-90, 2008
Research and Development
Phenolic Acids
Kennedy, 2008
Research and Development
Flavonoids
Quercetin
A.L. Waterhouse, Wine Phenolics,
Annals of the New York Academy of Sciences 957:21-36, 2002
Research and Development
Anthocyanins
Kennedy, 2008
Research and Development
Tannins
Schofield et al., Analysis of Condensed Tannins: A Review
Animal Feed Science and Technology 91:21-40, 2001
Research and Development
Polymeric Pigments
Kennedy, 2008
Research and Development
Phenolic Levels in Wine
Waterhouse, 2002
Research and Development
Measuring Phenolics
Total Phenolics
 A280
 Folin-Ciocalteu
Tannins
 Acid Butanolysis
 Aldehyde
Pigments
Nota bene: unless you are chromatographically separating discrete
compounds all measures of phenolics are methodologically defined
Research and Development
Total Phenolics
Absorbance at 280 nm
Pro’s: Simple; just requires UV-transparent
cuvette and a UV-capable
spectrophotometer (express as A280 in AU)
Con’s: Subject to interferences from other
aromatic ring containing compounds (e.g.,
nucleotides, aromatic amino acids)
Nota bene. . .these are relatively small effects
Research and Development
Total Phenolics
Folin-Ciocalteu
Pro’s: Measures all mono- and
dihydroxylated phenolics; automatable
Con’s: Subject to interferences from
fructose and SO2; spent reagent has to be
disposed of as hazardous waste
Research and Development
Tannins
Acid Butanolysis
Pro’s: Specific for tannins; anthocyanidin
color measured with spectrophotometer
(relative abundance)
Con’s: Low reaction yields; highly
dependent upon reaction conditions and
the tannin structure
Research and Development
Tannins
Aldehydes (Vanillin, DMCA*)
Pro’s: Measures flavan-3-ols and polymers
(m-dihydroxy’s); color measured with
spectrophotometer
Con’s: Rate and extent of color
development solvent dependent; vanillin
adduct absorbs at 500 nm (problematic
for red wines)
*dimethylaminocinnamaldehyde
Research and Development
Pigments
Any number of spectrophotometric
assays for pigments are available
These procedures have been
extensively researched by Chris
Somers in Australia (e.g., The Wine
Spectrum, Winetitles: Marleston, SA, 1998)
 e.g., A520, A420 and all their permutations
Research and Development
Adams-Harbertson Assays
Functional assays providing quantitative
information on various phenolic classes
 Total iron-reactive phenols
 Analogous to Folin-Ciocalteu
 Caveat: doesn’t measure monohydroxylated phenols
or anthocyanins
 Protein (BSA) precipitable tannins
 Tetrameric tannins and larger
 Polymeric pigments
 Non-SO2 bleachable pigmented fractions
 Non-protein precipitable: small polymeric pigment
 Protein precipitable: large polymeric pigment
 Free Anthocyanins
Research and Development
Adams-Harbertson Assays
Benefits
 Can run the analyses in-house IF you have a
Visible spectrophotometer, a microcentrifuge, a
vortexer and the necessary micropipettes
 The IRP is a measure of total phenolics (minus
anthocyanins) and doesn’t generate hazardous
waste
 The protein-precipitable tannin is highly
correlated to perceptual astringency
Research and Development
Tannin vs. Astringency
Kennedy et al., Analysis of Tannins in Red Wine Using Multiple Methods:
Correlation with Perceived Astringency, AJEV 57:481-485, 2006
Research and Development
Running the A-H Assay
 Sets of up to 24 samples
 4/5 segments, 9 sets of readings, ~ 3 hours
 5 results: anthocyanins, tannins, IRP, SPP, LPP
Research and Development
Gage R & R
OBJECTIVE: Quantify Measurement Error in
Measurement Systems
 Integral Part of SIX SIGMA Methodology
 Quality Systems… Zero Defects… ISO Standards…
 Goal: less than 3.4 defects in a million opportunities
 Early adapters: Motorola & Allied Signal (early 90’s)
 General Electric Co. – most successful implementer
 Two components
 Standard Deviation of Measured Values
 Assessment of Source of Variability
 Contributors to Measurement Variation
 Repeatability – Single Operator, Same Equipment
 Reproducibility – Operators, Protocol, Equipment,…
Research and Development
Gage R & R
Study Conducted in April-June 2008
 Design of Experiments - DOE
 3 wineries, 5 wines, 4 technicians, 4 repetitions
 full-factorial, randomized – 80 test results
 Resulting Standard Deviations
 (free-) Anthocyanins
 SPP
 LPP
 Tannins
 IRP
3.02%
2.01%
4.86%
2.79%
3.78%
 But… observed spikes of 7.6, 11.7,…
27.5%
 ANOVA analysis needed – Used MINITAB
Research and Development
Gage R & R
Operator Contribution 3.3 %, # of Categories* 7
* Automotive Industry Action Group (AIAG)
Measurement Systems Analysis (June 1998)
Research and Development
Gage R & R
Operator Contribution 34.4 %, # of Categories* 1
* Automotive Industry Action Group (AIAG)
Measurement Systems Analysis (June 1998)
Research and Development
Standard Procedure
The Assay Protocol – Essential KEY to Repeatability
& Reproducibility
 Sources of Adams-Harbertson Assay Protocol
 Technical literature and journals
 UC Davis Department of Viticulture & Enology website
 Trade publications
 Individual laboratory adaptations
 In practice… a multitude of ways of running the Assay
 Consequently,
 Large variations in reported results
 And even declarations of intrinsic invalidity
 Moreover,
 A closer look at the assay reveals significant potential for
improving its repeatability and reducing time of execution
Research and Development
Standard Procedure
Road to the Adams-Harbertson Assay SOP
 Initial documented procedure in place at Rubicon Estate
 Set up with the assistance of Dr. Harbertson & Dr. Adams
 Base documents from UC Davis Department of V & E website
 Modifications introduced and validated over time
 Salient results shared with Dr. Adams
 Jointly with Dr. Thorngate determined need for SOP
 Now working with the Gold Standard Group
 Created draft for the “Modified A&H Assay SOP”
 Currently being cast in ISO format
 Review and finalization to follow
 Gage R&R planned for mid-year 2010
 Expected SOP release date – Fall 2010
 Preliminary results indicate reduction in error “spikes”,
increased repeatability, and over 1/3 reduction in runtime
Research and Development
UV-Vis Spectroscopy
Early in Primary Fermentation
Research and Development
UV-Vis Spectroscopy
Later in Primary Fermentation
Research and Development
Calibration / Modeling
Linear Curve-fitting
A&H Assay Results – Predicted
UV-Vis Spectrum
anthocyanins
MODEL
*
*
*
*
*
*
absorbance @ 520 nm
Research and Development
UV-Vis Based A-H Assay
Multivariate Modeling - Chemometrics
 Openly-available, widely-used technology
 Commercial software packages can be purchased
 Implemented (and in use) in other process industries
 Applications: lab, virtual sensors, process optimization
 Expected Impact
 Implemented locally in the winery laboratory
 Once in place, no phenolics wet chemistry analyses
 Essentially no sample preparation
 Assay time of one-to-two minutes per sample
 Ideal for real-time vinification decisions
Research and Development
UV-Vis Based A-H Assay
 Development Methodology
laboratory analytical instrumentation
(lab-based; HPLC, GC/MS, …)
MEASURED
VALUES
MRSEC
standardized measurements
CALIBRATION
SAMPLES
(training and testing)
process analytical instrumentation
(at-line or in-line; UV/Vis, IR, …)
model building & deployment
(multivariate; PCR, PLS, ANN,… )
SAMPLE
RESULTS
SPECTRA
PC / Notebook
Research and Development
UV-Vis Based A-H Assay
 Validation
laboratory analytical instrumentation
(lab-based; HPLC, GC/MS, …)
MEASURED
VALUES
standardized measurements
FIELD
VALIDATION
SAMPLES
process analytical instrumentation
(at-line or in-line; UV/Vis, IR, …)
MRSEV
or
MRSEP
model building & deployment
(multivariate; PCR, PLS, ANN,… )
SAMPLE
RESULTS
SPECTRA
PC / Notebook
TEST SAMPLES
Research and Development
UV-Vis Based A-H Assay
 Deployment
process analytical instrumentation
(at-line or in-line; UV/Vis, IR, …)
model building & deployment
(multivariate; PCR, PLS, ANN,… )
SAMPLE
RESULTS
SPECTRA
PC / Notebook
TEST SAMPLES
Research and Development
The Predictive Model (Ver. 4)
Research and Development
Model Comparisons
Data ranges of current data and Skogerson data
Current
Anthocyaninsa
IRPb
Tanninsb
Skogerson et al. 2007
Min
Max
Min
Max
0
1419
0
1096
72.6
4979
19.8
2272
0
2667
-8.1
798
Prediction statistics for the Skogerson et al. (2007) model using our data
amg/L
RMSEP
rpred2
RPD
CVpred
Anthocyaninsa
466
0.20
0.5
105.0
IRPb
909
0.38
0.8
63.3
Tanninsb
406
0.33
1.0
70.3
malvidin-3-glucoside equivalents
bmg/L catechin equivalents
NOTE: Skogerson data was for Australian wines;
Current data was for domestic wines.
Research and Development
That being said. . .
Validation statistics for the prediction of phenolic components (n=248)
RMSEP
rpred2
RPD
CVpred
Anthocyaninsa
149
0.53
1.4
33.0
IRPb
383
0.76
2.1
25.6
Tanninsb
203
0.78
2.1
33.8
There is ample room for improvement!
RMSEP: root mean square error of prediction
rpred2: coefficient of determination of the prediction
RPD: ratio of standard deviation to standard error of prediction
CVpred: coefficient of variation of the prediction
amg/L
bmg/L
malvidin-3-glucoside equivalents
catechin equivalents
Research and Development
Summary




The Adams-Harbertson assays measure
functional classes of phenolic compounds in wine
The Adams-Harbertson assays are repeatable and
reproducible
The Adams-Harbertson assays SOP — a work in
progress
The Predictive Model shows great promise —
additional work is required
Research and Development
Acknowledgments
Dr. James Harbertson (Assoc. Prof.!) and his
laboratory
Dr. Douglas Adams
Gold Standard
Jordan Ferrier
Dr. Roger Boulton, Dr. Mark Downey &
Kirsten Skogerson
Tondi Bolkan, Evan Schiff, Karen
Moneymaker
Research and Development
Acknowledgments
Research and Development