34
Sensory Evaluation of Aroma Models for Flavor Characterization Keith Cadwallader University of Illinois at Urbana-Champaign TM Kenneth A. Spencer Award Symposium Kansas City Section of ACS October 27, 2008

Sensory Evaluation of Aroma Models for Flavor Characterization

  • Upload
    tannar

  • View
    80

  • Download
    0

Embed Size (px)

DESCRIPTION

TM. Sensory Evaluation of Aroma Models for Flavor Characterization. Keith Cadwallader University of Illinois at Urbana-Champaign. Kenneth A. Spencer Award Symposium Kansas City Section of ACS October 27, 2008. Overview:. Rationale: why conduct sensory studies?. General approach. - PowerPoint PPT Presentation

Citation preview

Page 1: Sensory Evaluation of Aroma  Models for Flavor Characterization

Sensory Evaluation of Aroma Models for Flavor Characterization

Keith CadwalladerUniversity of Illinois at Urbana-Champaign

TM

Kenneth A. Spencer Award SymposiumKansas City Section of ACSOctober 27, 2008

Page 2: Sensory Evaluation of Aroma  Models for Flavor Characterization

Overview:

Rationale: why conduct sensory studies?

General approach

Some important considerations

Some common types of model studies

Sensory methods (tools) used in sensory studies

Example of a dose-response study

Example of an omission study

Final thoughts

Example of an addition study

Page 3: Sensory Evaluation of Aroma  Models for Flavor Characterization

Why conduct sensory studies?

Cannot accurately predict the effect (sensory perception) caused by altering the chemical composition of odor mixtures based on only flavor dilution values or odor-activity values (OAVs).

Omission of a compound with a high OAV may not necessarily alter the sensory perception of the overall ‘flavor’ concept.

Page 4: Sensory Evaluation of Aroma  Models for Flavor Characterization

GCO screening of odorants

General approach for performing model studies

AEDA, DHDA, GCO-H, post-peak intensity scaling

identification by GC-MS, RIs and odor properties

concentrations and OAVs calculation of OAVs from threshold data

GC-MS with IS and SIDA methodology

aroma modelconstruction preliminary testing/adjustments

selection of appropriate matrix

sensory testingof aroma model omission studies (n-1) with difference testing and

descriptive analysis

dose-response studies (descriptive analysis)

Page 5: Sensory Evaluation of Aroma  Models for Flavor Characterization

Some things to consider:

Are all key odorants accounted for?

Are quantitative data accurate?

Is an appropriate matrix available or can it be (re-)created?

What is the objective of study?

• Impact (cause-and-effect relationship) of a single odorant

• (Re-)creation of an aroma system (model)

• Relative impact (or influence) of all aroma components on the aroma system

What is an appropriate experimental approach?

• Experimental design options

• Sensory methods of analysis

Page 6: Sensory Evaluation of Aroma  Models for Flavor Characterization

Some limitations in methods used to indicate key odorants

Odor-activity values (OAVs) – based on quantitative data (OAV = concentration/odor detection threshold).

Aroma-impact based of GCO data: - (e.g. post-peak) scaling of odorant intensity - flavor dilution factors or CHARM-values (from dilution analysis).

• Only useful for compounds of known identity

• Must have accurate concentration and odor threshold data

• number of odorants detected and the their perceived intensities depend on arbitrarily selected parameters: sample size, isolation method, degree of concentration of aroma extract, etc.

Page 7: Sensory Evaluation of Aroma  Models for Flavor Characterization

Let’s assume we have all relevant or key odorants identified and

accurately quantified, and an appropriate matrix is available.

What’s next?

Page 8: Sensory Evaluation of Aroma  Models for Flavor Characterization

Need to consider:

Objective and experimental design

Sensory method(s) for evaluation

Page 9: Sensory Evaluation of Aroma  Models for Flavor Characterization

Omission (n – 1) studies

- Sensory comparison of the aroma of the complete mixture against the same mixture in which an odorant (or group of odorants) have been omitted.- Suitable for the determination of potential impact of individual (or groups of) odorants on aroma system.

Dose response studies

- Sensory evaluation of a suitable product matrix that has been spiked with an odorant (or group of odorants) to determine if the addition causes an increase in the intensity of a specific flavor attribute.- Suitable technique to evaluate ‘cause and effect’ relationship between odorant and sensory attribute.

Common types of sensory studies . . .

Comparison of aroma model to real product (validation)- Use of sensory difference test and/or descriptive analysis

Page 10: Sensory Evaluation of Aroma  Models for Flavor Characterization

Sensory methods used in model studies . . .

Conventional Difference Tests

Do not require intensive training of panelists.

Statistical analysis is straightforward (well established).

Task is easy to understand and perform.

Sensitive to small differences provided enough observations (tests) are made.

Not intended to measure direction or degree of difference.

Use difference-from-control test if degree of difference is required.

Page 11: Sensory Evaluation of Aroma  Models for Flavor Characterization

Sensory methods used in model studies

provides qualitative and quantitative comparisons of the model against the product or the omission mixture.

provides descriptive terms for attributes and allows quantification of their perceived intensities.

In order to detect small differences between products, the performance level of the panel must be sufficient in terms of reproducibility (precision), discrimination power, and agreement among panelists (improved with training, use of external references and by increased number of panelists).

Descriptive Analysis

Complement difference tests

Terminology (lexicon) should be developed based not only on attributes of product being studied, but also based on attributes of all n-1 combinations (attributes cannot be predicted).

Page 12: Sensory Evaluation of Aroma  Models for Flavor Characterization

Example of a Dose-Response Study(with sensory descriptive analysis)

Page 13: Sensory Evaluation of Aroma  Models for Flavor Characterization

Farmhouse Cheddar Cheese . . .

Results of gas chromatography-olfactometry (GCO) and Aroma Extract Dilution Analysis (AEDA) indicated 2-isopropyl-3-methoxypyrazine (3-7 ppb) and p-cresol (200 ppb) to be “most likely” responsible for cowy/barny and earthy/bell pepper flavors, respectively.

Additional sensory testing was conducted to measure impact of compounds on perceived intensities of corresponding flavor descriptors (blind study).

Compounds spiked into a bland cheese matrix across concentration found in Farmhouse cheeses.

Evaluation by descriptive sensory panel in a blind study.

Suriyaphan, O.; Drake, M.A.; Chen, X.Q.; Cadwallader, K.R. Characteristic aroma componentsof British Farmhouse Cheddar cheese. J. Agric. Food Chem. 2001, 49, 1382-1387.

Page 14: Sensory Evaluation of Aroma  Models for Flavor Characterization

e.g. Flavor Profile of British Farmhouse Cheddar Cheese

0

2

4

6Bitter

Brothy

Cooked

Cowy/Barny

Diacetyl

Earthy/Bell Pepper (Aroma)

Earthy/Bell Pepper (Flavor)

Free Fatty Acid

FruityLactone

Nutty

Prickle

Salty

Sour

Sulfur

Sweet

Umami

Whey

description - aromatics associatedwith barns and stock trailers reference - p-cresol, Band-aid, phenol

Page 15: Sensory Evaluation of Aroma  Models for Flavor Characterization

Linking aroma analysis results to flavor lexicon terms

Relationship between p-cresol concentration and “cowy/barny flavor” intensity

00.5

1.0

1.52.0

2.53.0

3.5

0 65 100 165 300

p-Cresol (ppb)

Av

era

ge

Fla

vo

r In

ten

sit

y

2-isopropyl-3-methoxypyrazine (ppb)

Av

era

ge

In

ten

sit

y

0

1

2

3

4

5

6

7

0 3.5 7

earthy/bell pepper flavorearthy/bell pepper aroma

Relationship between 2-isopropyl-3-methoxypyrazine concentration and “earthy aroma/ flavor” intensity

p-Cresolthreshold = 55 ppb (in water)

2-isopropyl-3-methoxypyrazinethreshold = 0.002 ppb (in water)

Farmhouse Cheddar Cheese . . .

Page 16: Sensory Evaluation of Aroma  Models for Flavor Characterization

Example of an Addition Study(with difference/similarity scaling)

Page 17: Sensory Evaluation of Aroma  Models for Flavor Characterization

Beefy/Brothy Cheddar Cheese . . .

The unambiguous linking of sensory descriptors with causative chemical components permits researchers to precisely relate sensory flavour quality with the chemistry and technology of Cheddar cheese production.

The objective of this study was to identify volatile aroma compounds responsible for the beefy/brothy flavor note in Cheddar cheese.

Compounds spiked into a bland cheese matrix across concentration found in beefy/broth cheese.

Evaluation by similarity-to-control and descriptive sensory analysis.

Cadwallader, K.R., Drake, M.A., Carunchia-Whetstine, M.E. and Singh, T.J. 2006. Characterisation of Cheddar cheese flavour by sensory directed instrumental analysis and model studies. In Flavour Science: Recent Trends. Bredie, W.P. and Peterson, M.A. (Eds.), Developments in Food Science 43, Elsevier, New York, pp. 157-160.

Potential beefy/brothy compounds identified by GCO.

Page 18: Sensory Evaluation of Aroma  Models for Flavor Characterization

Differentiating odorants detected in beefy/brothy Cheddar cheeses

Compound Odor

Description

FD Factor

RIa AEDAb DHDAc

FFAP DB-5 Nd B1d B2d N B1 B2

2-Methyl-3-furanthiol(MFT)

1312 873 Vitamin, meaty

nd 3 9 nd nd nd

3-(Methythio)Propanal

1455 907 Potato 9 2187 2187 5 25 25

2-Methyl-(3-methyldithio) furan

1682 1170 Vitamin, meaty

nd n.d. n.d. 5 25 25

Maltol 1998 1175 Burnt sugar nd 9 9 nd nd nd

FuraneolTM (HDMF) 2035 1058 Burnt sugar 9 729 2187 5 25 25

Homofuraneol 2090 1160 Burnt sugar 3 9 3 nd 1 1

Bis-(2-methyl-3-furyl) disulfide

2132 1542 Vitamin, meaty

Nd nd nd 1 1 5

aRetention index. bFlavor dilution factor determined by aroma extract dilution analysis. cFlavor dilution factor determined by dynamic headspace dilution analysis. dN indicates control (not brothy) cheese, B1 and B2 refer to beefy/brothy cheeses.

Page 19: Sensory Evaluation of Aroma  Models for Flavor Characterization

Sensory Analysis of Beefy/Brothy Model Cheeses

CombinationAroma Description

(Intensity)a

Overall Similarity

Scoreb

No spike Not beefy/brothyc 0

MFT (2 ng/g) + Furaneol (10 g/g) Beefy/brothy (2) 8

MFT (4 ng/g) + Furaneol (20 ug/g) Beefy/brothy (3)Burnt sugar/fruity (3)

7

MFT (2 ng/g) + Furaneol (10 g/g) + Methional (80 g/g) Beefy/brothy (2) 9

MFT (4 ng/g) + Furaneol (20 g/g) + Methional (80 g/g) Beefy/brothy (3) 7

aAverage aroma intensity on a 15-point universal scale (nine trained panelists). bSimilarity to typical beefy/brothy Cheddar cheese (10-point scale). cCheese received the following aroma/ flavour scores: diacetyl=2, whey=4, cooked=3.5, milkfat=3.5, salty=4, sour=3, sweet=1.5.

Page 20: Sensory Evaluation of Aroma  Models for Flavor Characterization

Example of an Omission Study(with R-index method)

Page 21: Sensory Evaluation of Aroma  Models for Flavor Characterization

Four critical steps in omission studies

Choice of target material

Sensory validation of mixture (?)

Construction of synthetic mixture (model)

Choice of experimental approach and sensory method(s) for evaluating model

Omission studies . . .

Page 22: Sensory Evaluation of Aroma  Models for Flavor Characterization

Omission studies . . .

Example: Evaluation of key odorants of chipotle peppers

Cadwallader, K.R.; Lorjaroenphon, Y.; Kim, H.; Lee, S-Y. Evaluation of key odorants in chipotle pepper byquantitative analysis, calculation of odor-activity values and omission studies. In Recent Highlights in Flavor Chemistry & Biology. Proceedings of the 8th Wartburg Symposium. Hofmann, T., Meyerhof, W. and Schieberle, P.(eds), Deutsche Forschungsanstalt für Lebensmittelchemie, Garching, Germany.

Page 23: Sensory Evaluation of Aroma  Models for Flavor Characterization

Predominant Odorants in chipotle peppers by GCO*

A total of 41 odorants were detected by GCO (post-peak intensity scaling, 7 pt scale) of DSE-SAFE aroma extracts from the three dried chipotle pepper samples

16 compounds had high odor intensities 4.0

7 additional odorants had odor intensities 3

2- and 3-methylbutanal, 2-ethyl-3,5-dimethylpyrazine, 2-isobutyl-3-methoxypyrazine, 2-(3)-methylbutanoic acid,

-damascenone, guaiacol, o-cresol, 4-hydroxy-2,5-dimethyl-3(2H)-furanone, octanoic acid, p-cresol, sotolon, syringol,

coumarin, phenyacetic acid and vanillin

* Cadwallader, K.R.; Gnadt, T.A.; Jasso, L. Aroma components of chipotle peppers. In Hispanic Foods: Chemistry and Flavor (Tunick, M.H., González de Mejia, E., eds.); American Chemical Society: Washington, D.C., 2006, 57-66

Omission studies . . .

Page 24: Sensory Evaluation of Aroma  Models for Flavor Characterization

Concentrations and Odor-Activity Values (>100)

102220002253000acetic acid9

1155057362- and 3-methylbutanoic acid16

126 3.74652-methylbutanal2

245 49792,3-butanedione4

28161688linalool12

84732541guaiacol18

105250526202-methylpropanoic acid13

15000.0069ethyl 2-methylbutanoate5

33000.0133dimethyltrisulfide8

76000.005382-isobutyl-3-methoxypyrazine11

110000.00222-damascenone17

150000.005751-octen-3-one7

15350 0.230693-methylbutanal3

156400.046262-ethyl-3,5-dimethylpyrazine10

3760000.001376sotolon27

OAVThreshold

(ng/mL)conc. (ng/g)odorantno.

Page 25: Sensory Evaluation of Aroma  Models for Flavor Characterization

Concentrations and Odor-Activity Values (<100)

<130001497octanoic acid24

>11000014000phenylacetic acid31

46502845o-cresol21

6185010340syringol28

86805300m-cresol26

1025258coumarin29

141000143302-phenylethanol19

152403604butanoic acid14

19475phenylacetaldehyde15

219019144-methylguaiacol20

25317904-hydroxyl-2,5-dimethyl-3(2H)-furanone 23

374.5167hexanal6

3625907vanillin32

39552167p-cresol25

445021814-ethylguaiacol22

83 183methylpropanal1

973291skatole30

OAVThreshold

(ng/mL)conc. (ng/g)odorantno.

Page 26: Sensory Evaluation of Aroma  Models for Flavor Characterization

Composition of Matrix Applied in the Sensory Experiments

722.4 μg/g (dry basis)bnatural capsaicin (Aldrich, St. Louis, MO, USA)

2.1asucrose (Sigma)

2.6acellulose (Sigma, St. Louis, MO, USA)

Ratiobase composition

ga0.3soybean oil

g1.7base

mL100.1 M citrate buffer (pH 4.8)

amountcomposition

a Based on dietary fiber (2.6), sugars (2.1) and total fat (0.9) in 100 g of jalapeno pepper (wet basis) (NutritionData, 2006). b Based on analysis of capsaicin and dihydrocapsaicin in chipotle pepper using method of Thomas et al. (1998).

Page 27: Sensory Evaluation of Aroma  Models for Flavor Characterization

Omission studies – some additional considerations

Eliminating successively (n - 1) all possible components of the mixture

- may not reveal much because of antagonistic effects

Eliminating groups of compounds of the model - e.g. where each group is composed of odorants with similar odor qualities or same chemical class

Chipotle aroma . . .

Page 28: Sensory Evaluation of Aroma  Models for Flavor Characterization

earthy (2-ethyl-3,5-dimethylpyrazine and 2-isobutyl-3-methoxypyrazine)

smoky (guaiacol, 4-methylguaiacol, o-cresol, 4-ethylguaiacol, p-cresol, m-cresol, syringol, coumarin)

sweet aromatics (2,3-butanedione, HDMF, sotolon and vanillin)

floral/fruity (ethyl 2-methylbutanoate, linalool, phenylacetaldehyde, -damascenone, 2-phenylethanol, phenylacetic acid)

malty (methylpropanal, 2- and 3-methylbutanal)

sour/sweaty (acetic, 2-methylproanoic, butanoic, 2/3-methylbutanoic and octanoic acids)

sulfurous (dimethyltrisulfide)

green/plant-like (hexanal, 1-octen-3-one)

Odorant groups* for omission studies

* Terms decided upon by descriptive sensory panel

Page 29: Sensory Evaluation of Aroma  Models for Flavor Characterization

Omission studies . . .

Omission studies – methodology

Subjects were provided with mixtures (signals) marked with 3-digit codes and the complete model (noise) coded as R. A randomized complete block design was used to randomize the samples across subjects.

Subjects were instructed to gently squeeze each sample container, evaluate the odor and rank the samples on how different they were from R, with 1 = least different to 9 = most different. Subjects were allowed to reevaluate samples ad libitum. Subjects were instructed to wait at least 10 seconds between evaluations to minimize adaptation effects.

A response matrix was constructed for the entire panel to calculate the R-indices.

O’Mahony, M. Understanding discrimination tests: A user friendly treatment of response bias, rating and ranking R-index tests and their relationship to signal detection. J. Sensory Stud. 1992, 7, 1-47.

Page 30: Sensory Evaluation of Aroma  Models for Flavor Characterization

Omission studies . . .

Page 31: Sensory Evaluation of Aroma  Models for Flavor Characterization

R-index Values for Omission Test

41.4green/plant-like (6, 7)

44.8sulfurous (8)

48.3sour/sweaty (9, 13, 14, 16, 24)

55.2malty (1, 2, 3)

62.1floral/fruity (5, 12, 15, 17, 19, 31)

62.1sweet aromatics (4, 23, 27, 32)

*69.0smoky (18, 20, 21, 22, 25, 26, 28, 29)

*79.3earthy (10, 11)

R-indexbodorant group omitteda

a Numbers in parentheses indicate odorant numbers omitted. Description of each group was determined by consensus opinion of the trained sensory descriptive panel. b R-index of each model is calculated by using John Brown computations (O’Mahony, 1992) against control (complete model) (n=29; female=21 and male=8). *Significantly different from control at α=0.05 (critical value, expressed in percentage; R-Index = 50% for two-tailed test, α=0.05, n=29 is 17.37).

Page 32: Sensory Evaluation of Aroma  Models for Flavor Characterization

Synergistic and Antagonistic Effects

Some Final Thoughts

Synergistic effects are mainly observed for subthreshold concentrations,i.e. a decrease in detection threshold occurs1.

But models are build from odorants at suprathreshold concentrations - in this region antagonistic effects seem to be most common2.

In general, human subjects are unable to identify individual odorants whenthe mixture contains greater than four odorants in total3. This helps explain why omission of one or more odorants from a complex odor mixture oftenis not distinguished from the intact (complete) mixture.

1 Laska, M.; Hudson, R. A comparison of the detection thresholds of odour mixtures. Chem.Senses 1991, 16, 651-662.

2 Grosch, W. Evaluation of the key odorants of foods by dilution experiments, aroma models and omission. Chem. Senses 2001, 26, 533-545.

3. Liang, D.G. Perceptual odour interactions and objective mixture analyses. Food Qual. Pref. 1994, 5, 75-80.

Page 33: Sensory Evaluation of Aroma  Models for Flavor Characterization

Additional References:

Brown, J. Recognition assessed by rating and ranking. Brit. J. Phychol. 1974, 65, 13-22

Czerny, M.; Mayer, F.; Grosch, W. Sensory study on the character impact odorantsof roasted arabica coffee. J. Agric. Food Chem. 1999, 47, 695-699.

Drake, M.A.; Miracle, R.E.; Caudle, A.D. ; Cadwallader, K.R. Relating sensory and instrumental analyses. In Sensory-Directed Flavor Analysis. Marsili, R. (Ed.), CRC Press/Taylor & Francis Group, LLC, Boca Raton, FL, 2007, pp. 23-54.

Engel, E.; Nicklaus, S.; Salles, C.; Le Quere, J.-L. Relevance of omission tests todetermine flavour-active compounds in food: application to cheese taste. Food Qual.Pref. 2002, 13, 505-513.

Karagul-Yuceer, Y.; Vlahovich, K.N.; Drake, M.A.; Cadwallader, K.R. Characteristicaroma components of rennet casein. J. Agric. Food Chem. 2003, 51, 6797-6801.

Page 34: Sensory Evaluation of Aroma  Models for Flavor Characterization