Development of a UV-Vis Spectral Model for the American Wine Industry

1
Phenolics and Tannin Assays for Practical Use in
Winemaking Giovanni Colantuoni John Thorngate
2
Outline

• Introduction
• Grape and Wine Phenolics
• Measuring Phenolics
• Adams-Harbertson Assays
• Gage RR Analysis
• Creating a Standardized SOP
• The UV-Vis Predictive Model
• Chemometrics Model Calibration and Deployment
• Comparison to Skogerson-Downey-Boulton
• Using the Model
• Summary

3

• Chemists interested in polyphenols, in common
  with the majority of scientists, tackle todays
  problems with yesterdays 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  inadequate and to that extent are already out of
  date.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
4
Introduction

• Why focus on phenolics?
• Important for
• Color
• Taste
• Mouthfeel
• Wine aging

5
Introduction

• Why measure phenolics?
• Identify higher quality lots more easily
• Use phenolic data for
• Press decisions
• Heavy press additions
• Blend balancing
• Evaluation of processing

6
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 3577-90, 2008
7
Phenolic Acids
Kennedy, 2008
8
Flavonoids
Quercetin
A.L. Waterhouse, Wine Phenolics, Annals of the
New York Academy of Sciences 95721-36, 2002
9
Anthocyanins
Kennedy, 2008
10
Tannins
Schofield et al., Analysis of Condensed Tannins
A Review Animal Feed Science and Technology
9121-40, 2001
11
Polymeric Pigments
Kennedy, 2008
12
Phenolic Levels in Wine
Waterhouse, 2002
13
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
14
Total Phenolics

• Absorbance at 280 nm
• Pros Simple just requires UV-transparent
  cuvette and a UV-capable spectrophotometer
  (express as A280 in AU)
• Cons Subject to interferences from other
  aromatic ring containing compounds (e.g.,
  nucleotides, aromatic amino acids)
• Nota bene. . .these are relatively small effects

15
Total Phenolics

• Folin-Ciocalteu
• Pros Measures all mono- and dihydroxylated
  phenolics automatable
• Cons Subject to interferences from fructose
  and SO2 spent reagent has to be disposed of as
  hazardous waste

16
Tannins

• Acid Butanolysis
• Pros Specific for tannins anthocyanidin color
  measured with spectrophotometer (relative
  abundance)
• Cons Low reaction yields highly dependent
  upon reaction conditions and the tannin structure

17
Tannins

• Aldehydes (Vanillin, DMCA)
• Pros Measures flavan-3-ols and polymers
  (m-dihydroxys) color measured with
  spectrophotometer
• Cons Rate and extent of color development
  solvent dependent vanillin adduct absorbs at 500
  nm (problematic for red wines)

dimethylaminocinnamaldehyde
18
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

19
Adams-Harbertson Assays

• Functional assays providing quantitative
  information on various phenolic classes
• Total iron-reactive phenols
• Analogous to Folin-Ciocalteu
• Caveat doesnt 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

20
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 doesnt generate hazardous
  waste
• The protein-precipitable tannin is highly
  correlated to perceptual astringency

21
Tannin vs. Astringency
Kennedy et al., Analysis of Tannins in Red Wine
Using Multiple Methods Correlation with
Perceived Astringency, AJEV 57481-485, 2006
22
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

23
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
  90s)
• 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,

24
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 3.02
• SPP 2.01
• LPP 4.86
• Tannins 2.79
• IRP 3.78
• But observed spikes of 7.6, 11.7, 27.5
• ANOVA analysis needed Used MINITAB

25
Gage R R

• Operator Contribution 3.3 , of Categories 7

Automotive Industry Action Group (AIAG)
Measurement Systems Analysis (June 1998)
26
Gage R R

• Operator Contribution 34.4 , of Categories 1
  

Automotive Industry Action Group (AIAG)
Measurement Systems Analysis (June 1998)
27
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

28
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 AH Assay SOP
• Currently being cast in ISO format
• Review and finalization to follow
• Gage RR 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

29
UV-Vis Spectroscopy

• Early in Primary Fermentation

30
UV-Vis Spectroscopy

• Later in Primary Fermentation

31
Calibration / Modeling
Calibration / Modeling
Linear Curve-fitting
AH Assay Results Predicted
UV-Vis Spectrum
MODEL



anthocyanins



absorbance _at_ 520 nm
32
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

33
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
34
UV-Vis Based A-H Assay

• Validation

laboratory analytical instrumentation (lab-based
HPLC, GC/MS, )
MEASURED VALUES
MRSEV or MRSEP
standardized measurements
FIELD VALIDATION SAMPLES
model building deployment (multivariate PCR,
PLS, ANN, )
process analytical instrumentation (at-line or
in-line UV/Vis, IR, )
SAMPLE RESULTS
SPECTRA
PC / Notebook
TEST SAMPLES
35
UV-Vis Based A-H Assay

• Deployment

model building deployment (multivariate PCR,
PLS, ANN, )
process analytical instrumentation (at-line or
in-line UV/Vis, IR, )
SAMPLE RESULTS
SPECTRA
PC / Notebook
TEST SAMPLES
36
The Predictive Model (Ver. 4)
37
Model Comparisons
Data ranges of current data and Skogerson data Data ranges of current data and Skogerson data Data ranges of current data and Skogerson data Data ranges of current data and Skogerson data Data ranges of current data and Skogerson data
Current Current Skogerson et al. 2007 Skogerson et al. 2007
Min Max Min Max
Anthocyaninsa 0 1419 0 1096
IRPb 72.6 4979 19.8 2272
Tanninsb 0 2667 -8.1 798
Prediction statistics for the Skogerson et al. (2007) model using our data Prediction statistics for the Skogerson et al. (2007) model using our data Prediction statistics for the Skogerson et al. (2007) model using our data Prediction statistics for the Skogerson et al. (2007) model using our data Prediction statistics for the Skogerson et al. (2007) model using our data
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
NOTE Skogerson data was for Australian
wines Current data was for domestic wines.
amg/L malvidin-3-glucoside equivalents bmg/L
catechin equivalents
38
That being said. . .
Validation statistics for the prediction of phenolic components (n248) Validation statistics for the prediction of phenolic components (n248) Validation statistics for the prediction of phenolic components (n248) Validation statistics for the prediction of phenolic components (n248) Validation statistics for the prediction of phenolic components (n248)
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 malvidin-3-glucoside equivalents bmg/L
catechin equivalents
39
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

40
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

41
Acknowledgments