Application of PLS and GPA in understanding the relationships between descriptive and headspace flavor profiles

Transcription

1 Application of PLS and GPA in understanding the relationships between descriptive and headspace flavor profiles S. J. CHUNG, H. Heymann, and I. U. Gruen. Department of Food Science University of Missouri

2 Flavor Research Odor active compounds Basic Tastants Physical properties Appearance Perceptual process Flavors

3 Application of multivariate techniques in flavor research I Sensory Analysis Descriptive Analysis- descriptors Time-Intensity Analysis- time dependent data points Chemical Analysis Headspace Analysis- odor-active volatile compounds

4 Application of multivariate techniques in flavor research II Application of multivariate techniques for exploratory purpose Multi-disciplinary approach Need of understanding the relationship between data sets

5 Partial Least Square Analysis Asymmetric method Predicting one data set from the other Solution for each set.in ADDITION Best solution of X to explain Y Two data sets

6 Generalized Procrustes Analysis Symmetric method Extracts common structure from multiple data sets Centering, rotating, adjusting scale Multiple data sets

7 The role of milk fat in ice cream Fatty, full, and rich sensation Solvent for many aroma compounds Reversible interaction Fat controls the intensity & the release rate of the aroma compounds

8 Objectives Discussion of Applying various multivariate statistical analysis techniques to individual data sets Using GPA and PLS to link the sensory perception to the flavor profile obtained by dynamic headspace analysis

9 Materials and Methods

10 Samples 4 fat levels and 3 flavor compounds = 12 samples 0% MF 3% MF 6% MF 12% MF Vanillin d-decalactone Hexanal

11 Descriptive Analysis 12 (-1) Panelists (6 females, 6 males) Training sessions 30 min X 4 times Definition of attributes Paper ballot Sample serving order: 12*12 William Latin Square Design (WLSD) Triplicate

12 Modified Dynamic Headspace (MDH) 50g ice cream 25 ml synthetic saliva 37 C Magnetic stir bar Nitrogen gas flow (100ml/min) Volatiles collected in Tenax trap GC-MS

13 Statistical Analysis I Descriptive Analysis General Linear Model Canonical Variate Analysis Modified Dynamic Headspace analysis General Linear Model Principal Component Analysis

14 Statistical Analysis II Sensory and Chemical data correlation Types of Data Descriptive (DA) vs. Chemical data (CHEM) Descriptive (DA) vs. Log transformed chemical data (LCHEM) Types of Analyses Generalized Procrustes Analysis Partial Least Square Analysis

15 Results and Discussion

16 Descriptive Analysis Texture Descriptors Iciness, Crumbliness, Denseness, Stickiness, Melt rate, Thickness Flavor Descriptors Overall flavor, Sweet, Vanilla, Coconut, Stale flavor, Cream flavor, Eggy flavor, Cooked flavor After Taste Descriptors Mouth coating, Powdery

17 Sensory characteristics of ice creams an overview by CVA coconut, overall fl. CV2 stale, powdery FLAVOR 12vh 12dlh 12hh cc 0dlh 1 ol. 6dlh 3dlh eg 0.5 ds cmb ck sw ic mr 0 tk sk vn6vh 6hh mc pd 3hh 0vh -0.5 crmsl 3vh thick, dense cream, cooked FAT CV icy, crumble, melt rate 0hh

18 Volatile characteristics of ice creams an overview by PCA PC2 Fat 12dlh 12vh 6dlh 6vh fdlh 3vh hh 12hh 3hh hh -1 0vh 0dlh hexanal vs. others PC1

19 Effect of fat on flavor release Aldehydes ( ) hexanal, heptanal, Z-2-heptenal, octanal, nonanal, benzaldehyde, etc. Alcohols ( ) 1-octen3-ol, dec-2-enol, 1-ethyl-2-hexanol Ester ( ) propyl octanoate Methyl ketones ( ) 2-hexanone, 2-heptanone, 2-octanone, 2- nonanone, etc.

20 Correlating Sensory and Chemical Data Sets

21 PLS- CHEM & DA The second PLS axis 0dlh 0vh hh dlh 3dlh -1 3vh 6vh 6hh hh -2 The first PLS axis 12vh 12dlh 12hh The second PLS axis meltr,icy, stale, crumble aldehydes stick 1 thick mcoat dense 11 cream eggy ov flav coconut cook vanilla 12 sweet powdery stale crumble icy meltr The first PLS axis creamy, thick, dense methyl ketones

22 PLS- LCHEM & DA The second PLS axis 0dlh 0hh 0vh 3vh 3hh dlh 6vh 6hh 6dlh 12dlh 12vh 12hh The first PLS axis The second PLS axis icy 20 meltr stick 16 thick mcoat dense 0.5 eggy cream powdery coconut cook 6 sweet 18vanilla ov flav stale crumble The first PLS axis

23 GPA- CHEM & DA The second dimension m0hh 0hhCH dlhDA 0vhCH 3dlhDA m0dlh0dlhch m0vh 0vhDA 3vhDA 0.5 m3dlh m3vh3dlhch 3hhDA 3vhCH 6dlhDA 6dlhCH m6vh m6dlh 6vhCH 12dlhCH 0hhDA 0 6vhDA m3hh 12vhCH 12hhCH 6hhDA m12dlh 3hhCH -0.5 m12vh m6hh m12hh12dlhda 6hhCH 12vhDA The first dimension 12hhDA The second dimension 0.7 icy crumble 7 meltr coconut sweet stale vanilla ov eggy flav cook stick 2 4 cream thick mcoat 1dense powdery -0.7 The first dimension

24 GPA- LCHEM & DA The second dimension dlhCH 12vhCH 1 m12dlh m12vh 12hhCH 0dlhCH 3dlhCH m12hh 0.5 3vhCH 0hhCH m0dlh 12vhDA 12dlhDA m3dlh 3hhCH 12hhDA m3vh 0vhCH m3hhm0vh m0hh 6dlhDA 0 6hhDA 3dlhDA3hhDA 0dlhDA 0hhDA vhDA 0 1 3vhDA 2 0vhDA m6dlh m6hh -1 m6vh dlhCH 6hhCH-2 6vhCH -2.5 The first dimension The second dimension creamy, thick, dense methyl ketones mcoat powdery dense eggy thick stick cook cream ov vanilla flav coconut sweet stale10 22 crumble icy 5 197meltr The first dimension meltr,icy stale crumble aldehydes 3 20

25 Summary and Conclusion I Effect of fat.. on sensory properties Texture- smooth, dense Flavor-enhancement and suppression on volatile profile Aldehydes, esters and alcohols decreased as fat level increase Methyl ketones increased as fat level increase

26 Summary and Conclusion II PLS Predictive CHEM LCHEM GPA Applicable to multiple data sets CHEM LCHEM

27 Thank you! Questions?