Anda di halaman 1dari 5

S54 JOURNAL OF FOOD SCIENCEVol. 71, Nr.

1, 2006
Published on Web 1/11/2006
2006 Institute of Food Technologists
Further reproduction without permission is prohibited
S
:

S
e
n
s
o
r
y

&

N
u
t
r
i
t
i
v
e

Q
u
a
l
i
t
i
e
s

o
f

F
o
o
d
JFS S: Sensory and Nutritive Qualities of Food
Aroma Impact Compounds in Three
Citrus Oils: Cross-matching Test and
Correspondence Analysis Approach
M MM MMASAHIR ASAHIR ASAHIR ASAHIR ASAHIRO OO OO C CC CCHIDA HIDA HIDA HIDA HIDA, K , K , K , K , KEIK EIK EIK EIK EIKO OO OO Y YY YYAMASHIT AMASHIT AMASHIT AMASHIT AMASHITA AA AA, , , , , Y YY YYURI URI URI URI URI I II IIZUMIY ZUMIY ZUMIY ZUMIY ZUMIYA AA AA, K , K , K , K , KOKICHI OKICHI OKICHI OKICHI OKICHI W WW WWA AA AAT TT TTANABE ANABE ANABE ANABE ANABE, , , , , AND AND AND AND AND H HH HHIR IR IR IR IRO OO OOT TT TTOSHI OSHI OSHI OSHI OSHI T TT TTAMURA AMURA AMURA AMURA AMURA
ABSTRA ABSTRA ABSTRA ABSTRA ABSTRACT CT CT CT CT: : : : : T TT TTw ww wwenty enty enty enty enty-thr -thr -thr -thr -three odor chemicals fr ee odor chemicals fr ee odor chemicals fr ee odor chemicals fr ee odor chemicals from 3 om 3 om 3 om 3 om 3 C CC CCitrus itrus itrus itrus itrus essential oils (lemon, essential oils (lemon, essential oils (lemon, essential oils (lemon, essential oils (lemon, V VV VValencia or alencia or alencia or alencia or alencia orange ange ange ange ange, and , and , and , and , and C CC CCitrus itrus itrus itrus itrus sudachi sudachi sudachi sudachi sudachi) )) ))
w ww wwer er er er ere selected as the potent char e selected as the potent char e selected as the potent char e selected as the potent char e selected as the potent character acter acter acter acter-impact compounds on the basis of their limited odor unit v -impact compounds on the basis of their limited odor unit v -impact compounds on the basis of their limited odor unit v -impact compounds on the basis of their limited odor unit v -impact compounds on the basis of their limited odor unit values alues alues alues alues, and then , and then , and then , and then , and then
every chemical was cross-matched by sensory test to the 3 oils to attribute each aroma character to 1 of the 3 every chemical was cross-matched by sensory test to the 3 oils to attribute each aroma character to 1 of the 3 every chemical was cross-matched by sensory test to the 3 oils to attribute each aroma character to 1 of the 3 every chemical was cross-matched by sensory test to the 3 oils to attribute each aroma character to 1 of the 3 every chemical was cross-matched by sensory test to the 3 oils to attribute each aroma character to 1 of the 3 Citrus Citrus Citrus Citrus Citrus
oils. The matching-frequency data (ratio data) obtained was subjected to correspondence analysis and graphed on oils. The matching-frequency data (ratio data) obtained was subjected to correspondence analysis and graphed on oils. The matching-frequency data (ratio data) obtained was subjected to correspondence analysis and graphed on oils. The matching-frequency data (ratio data) obtained was subjected to correspondence analysis and graphed on oils. The matching-frequency data (ratio data) obtained was subjected to correspondence analysis and graphed on
a diagr a diagr a diagr a diagr a diagram. C am. C am. C am. C am. Consequently onsequently onsequently onsequently onsequently, it was found that the ar , it was found that the ar , it was found that the ar , it was found that the ar , it was found that the aroma char oma char oma char oma char oma character of lemon oil was mainly r acter of lemon oil was mainly r acter of lemon oil was mainly r acter of lemon oil was mainly r acter of lemon oil was mainly repr epr epr epr epresented b esented b esented b esented b esented by citr y citr y citr y citr y citral, with al, with al, with al, with al, with
a high matching frequency of 0.89 (59 counts out of 66 trials, a high matching frequency of 0.89 (59 counts out of 66 trials, a high matching frequency of 0.89 (59 counts out of 66 trials, a high matching frequency of 0.89 (59 counts out of 66 trials, a high matching frequency of 0.89 (59 counts out of 66 trials,
2 22 22
o oo oo
= 93.36). The orange character consisted mostly of = 93.36). The orange character consisted mostly of = 93.36). The orange character consisted mostly of = 93.36). The orange character consisted mostly of = 93.36). The orange character consisted mostly of
linalool and nonanal. linalool and nonanal. linalool and nonanal. linalool and nonanal. linalool and nonanal. , , , , , -Pinene, -Pinene, -Pinene, -Pinene, -Pinene, -sinensal and myrcene were related to the aroma of -sinensal and myrcene were related to the aroma of -sinensal and myrcene were related to the aroma of -sinensal and myrcene were related to the aroma of -sinensal and myrcene were related to the aroma of C. sudachi C. sudachi C. sudachi C. sudachi C. sudachi oil. The applica- oil. The applica- oil. The applica- oil. The applica- oil. The applica-
tion of the cross-matching test and correspondence analysis in the characterization of food aromas has never been tion of the cross-matching test and correspondence analysis in the characterization of food aromas has never been tion of the cross-matching test and correspondence analysis in the characterization of food aromas has never been tion of the cross-matching test and correspondence analysis in the characterization of food aromas has never been tion of the cross-matching test and correspondence analysis in the characterization of food aromas has never been
r rr rrepor epor epor epor eported in the liter ted in the liter ted in the liter ted in the liter ted in the literatur atur atur atur ature to date e to date e to date e to date e to date, and the v , and the v , and the v , and the v , and the validity of these methods was successfully demonstr alidity of these methods was successfully demonstr alidity of these methods was successfully demonstr alidity of these methods was successfully demonstr alidity of these methods was successfully demonstrated b ated b ated b ated b ated by our study y our study y our study y our study y our study. .. ..
Keywords: Keywords: Keywords: Keywords: Keywords: Citrus lemon Citrus lemon Citrus lemon Citrus lemon Citrus lemon, character-impact compounds, limited odor unit, cross-matching test, correspon- , character-impact compounds, limited odor unit, cross-matching test, correspon- , character-impact compounds, limited odor unit, cross-matching test, correspon- , character-impact compounds, limited odor unit, cross-matching test, correspon- , character-impact compounds, limited odor unit, cross-matching test, correspon-
dence analysis dence analysis dence analysis dence analysis dence analysis
Introduction
T
he characterization of food flavors plays an important role in
the production of high-quality processed foods and in the qual-
ity control of foods. Only human beings, of course, can evaluate aro-
ma quality. Without sensory tests, the evaluation of aroma quality
based solely on instrumental analysis and statistical analysis is im-
possible. So sensory chemical analysis has been applied to find the
flavor components that have the highest sensory impact, such as
character-impact compounds or contributory flavor compounds
(Belitz and Grosch 1986). GC-Olfactometry (GC-O; Schieberle and
Grosch 1990, Roberts and Acree 1996) and the odor unit (Uo) meth-
od (Guadagni and others 1966; Buttery and others 1990; Tamura
and others 1993) are typical and suitable tools for the selection of
potent aroma chemicals, leading to the reconstruction of the model
oils. To bridge the gap between the psychological/sensory effect
and the instrumental analysis data, the omission test (Grosch 2001)
and the odor recognition test (Boonbumrung and others 2003)
have been introduced.
On the other hand, statistical approaches such as chemometrics,
including principal component analysis and cluster analysis, have
been introduced for finding character-impact compounds of es-
sential oils such as those related to fruit and wine aromas (Noble
1978; Forina and others 1982; Mayfield and others 1986; Dimoy and
others 1987; Tamura and others 1990). There, verbal descriptors are
indispensable for the wide-consensus characterization of aromas.
Recently, the cross-matching test, which is 1 of the psychological
procedures, was applied to sensory evaluation by Buchanan and
others (2003). They tested odor memory using 2 different tasks, a
cross-modal odorname matching test and a unimodal odorodor
recognition test. The basic idea was to implement a forced choice
paradigm as a simple nonverbal routine. Syntactic and semantic de-
scriptions are not necessary for this simple evaluation. However, we
were unable to locate descriptions of any other approaches to the
cross-matching test in the literature of the food science field.
Correspondence analysis (CA) has been used for handling mul-
tiple-frequency data (Greenacre 1994). This analysis is similar in
nature to factor analysis (Hoffman and Franke 1986). CA was orig-
inally developed in France by Benzerci in the early 1960s and 1970s
(Panagiotakos and Pitsavos 2004). Since then, it has spread to the
field of social studies (van der Heijden and others 1989; Blasius and
Thiessen 2000). Using the visual configuration obtained by corre-
spondence analysis, Zheng and Yoshino (2003) discussed the cross-
national difference in attitude toward nature and the environment.
In recent genome explorations, CA has been used to identify or-
ganisms or parts of their open reading frame (ORF) products that
are characterized by preferred amino acids or codons (Tekaia and
others 1999; Garcia-Vallv and others 2000). However, there is no
example of a study in aroma science in which the cross-matching
test (psychophysical procedure) and CA (nonparametric data anal-
ysis) are combined. The odor cross-matching test can eliminate the
complexity brought about by the differences between subjective
evaluations, whereas CA can sum up the multiple-contingency
data in a visual configuration.
The aim of this study was to identify character-impact com-
pounds by using the odor cross-matching test and CA for some Cit-
rus species. The potent character-impact compounds are selected
by the limited odor unit method, and the CA diagram can indicate
the actual contributory flavor compounds related to the food aroma.
Materials and Methods
Essential oils Essential oils Essential oils Essential oils Essential oils
Citrus sudachi oil (sudachi oil) and Citrus lemon oil (lemon oil)
were isolated by the modified solvent extraction method reported
in a previous article (Mookdasanit and others 2003). C. sudachi
MS 20050268 Received 5/10/05, Revised 10/10/05, Accepted 10/18/05. The
authors are with Dept. of Biochemistry and Food Science, Kagawa Univ.,
2393 Miki-Cho, Kagawa 761-0795, Japan. Author Chida is with Blending
and Product Development Div., Japan Tobacco Inc., Tokyo, Japan. Direct
inquiries to author Tamura (E-mail: tamura@ag.kagawa-u.ac.jp).
Vol. 71, Nr. 1, 2006JOURNAL OF FOOD SCIENCE S55
S
:

S
e
n
s
o
r
y

&

N
u
t
r
i
t
i
v
e

Q
u
a
l
i
t
i
e
s

o
f

F
o
o
d
URLs and E-mail addresses are active links at www.ift.org
Aroma impact compounds in citrus oils . . .
fruits were supplied by the Horticultural Experiment Station in
Tokushima Prefecture (Japan). C. lemon fruits imported from Sunk-
ist Growers Inc. (Sherman Oaks, Calif., U.S.A.) were purchased and
peeled, and then the essential oil was isolated using the method
described in the same previous article. Cold-pressed oil of Valencia
orange was supplied by Japan Tobacco, Inc. (Tokyo, Japan). The
original sudachi oil contained 2.8% of odorless pentane as the im-
purity. The orange and lemon oils did not contain any solvents. The
concentration of the 3 essential oils was fixed at 200 ppm without
further purification.
Reagents Reagents Reagents Reagents Reagents
23 authentic chemicals were purchased from Fluka Chemika-
BioChemika (Tokyo, Japan), Wako Pure Chemical Industries, Ltd.
(Osaka, Japan) and the Aldrich Chemical Co., Inc. (Milwaukee, Wis.,
U.S.A.). The citral used here was a mixture of geometrical isomers,
because the ratio of neral to geranial in the lemon was approxi-
mately 50:50.
Potent character-impact compounds Potent character-impact compounds Potent character-impact compounds Potent character-impact compounds Potent character-impact compounds
found by the limited odor units found by the limited odor units found by the limited odor units found by the limited odor units found by the limited odor units
The potent character-impact components of lemon, orange, and
sudachi were selected according to their limited odor unit (Lod)
values, in the manner reported previously (Tamura and others
1996). The Lod equation is as follows:
Cr
Limited Odor Unit (Lo) =
____
(1)
Td
where Cr is the concentration of the individual components at the
recognition threshold of the volatile oils (ppm), and Td is the detec-
tion threshold of the individual components (ppm).
Each chemical component was dissolved in methanol and then
gradually diluted with distilled water. The final concentration of the
methanol was less than 200 ppm. The Lod of each constituent in
the 3 oils was calculated by using the odor detection thresholds of
each chemical and the odor recognition thresholds of sudachi, lem-
on, and orange against their individual reference oils. The recogni-
tion thresholds of lemon and orange against iyo-orange (Citrus iyo,
a type of tangerine; it is believed to be a hybrid of tangerine and
orange) were 0.46 ppm and 4.62 ppm, respectively. The recognition
threshold of sudachi oil against lemon oil was 0.18 ppm. More than
20 panel members participated in the determination of the recog-
nition thresholds of the 3 Citrus oils.
Cross-matching test between the Cross-matching test between the Cross-matching test between the Cross-matching test between the Cross-matching test between the
aroma compounds and the 3 aroma compounds and the 3 aroma compounds and the 3 aroma compounds and the 3 aroma compounds and the 3 Citrus Citrus Citrus Citrus Citrus oils oils oils oils oils
Twenty-two subjects participated in the sensory tests. The panel
consisted of 16 females and 6 males with ages ranging from 22 to 47
y (mean age, 23 y). The concentrations of each of the 23 compounds
selected based on the Lod values were 1000-fold higher than their
individual detection thresholds. The subjects sniffed the head-
space gas emitted from a 10-mL sample solution in a 45-mL brown
glass bottle. The subjects initially sniffed the aromas of the lemon,
orange, and sudachi oils and recognized their odor characters.
Then, all panel members sniffed the 23 compounds and matched
these chemicals to the 3 oils according to the odor characters. The
total matching frequencies were 66 collected from 22 subjects (3
times trials in each panel members).
Statistics Statistics Statistics Statistics Statistics
The original data in the cross-matching frequency table were as-
signed to cells according to 2 criteria: items and categories. The con-
tingency data were handled by the reciprocal averaging algorithm of
the correspondence analysis (CA) done by the SAS-JMP
TM
software
(SAS institute, Cary, N.C., U.S.A.). CA computes the sum over all cells
of the chi-square (
2
) of the difference between the theoretical and
the actual value, this square being divided by the theoretical value.
Hence, CA was set as symmetrical standardization (Hoffman and
Franke 1986; Van de Geer 1993), prizing items (23 chemicals), and
categories (3 essential oils). Inertia was defined as the total chi-square
(
2
) for the 2-way frequency table divided by the total sum of all ob-
servation data of the table. The portion of each dimension was de-
fined as the ratio of each inertia against total inertia.
Results and Discussion
Potent character-impact compounds Potent character-impact compounds Potent character-impact compounds Potent character-impact compounds Potent character-impact compounds
The Lod values of the volatile constituents found in the 3 Citrus
species (lemon, Valencia orange, and Citrus sudachi) are shown in
Table 1. The Lod value is essentially the Uo value at the recognition
threshold level of the essential oil against its reference oil, as shown
in Eq. 1 (Tamura and others 1996). The 23 identified compounds
selected as the representative components of the 3 Citrus oils are
shown in Table 2. The chemicals having higher Lod values (Table 2)
consisted of chemicals found in common in the 3 Citrus oils and of
particular chemicals that were found only in 1 or 2 Citrus oils. 1,8-
Cineole, linalool, octanal, limonene, -pinene and (+)-citronellal
Table 1The Lod values for the lemon, sudachi, and orange oils
Nr Lemon Lod
a
Orange Lod
a
Sudachi Lod
a
1 (+)-Limonene 2.60 (-)-Linalool 5.96 (+)-Linalool 1.08
2 -Terpinene 0.20 (+)-Linalool 2.91 (-)-Linalool 0.75
3 Linalool 0.17 (+)-Limonene 2.83 Octanal 0.64
4 Geraniol 0.16 Octanal 0.86 (+)-Limonene 0.09
5 1,8-Cineole 0.15 Dodecanal 0.84 -Pinene 0.04
6 Geranyl acetate 0.08 Myrcene 0.81 p-Cymene 0.03
7 -Pinene 0.08 Decanal 0.72 g-Terpinene 0.03
8 Octanal 0.08 Citronellal 0.16 Myrcene 0.02
9 (-)-Limonene 0.07 Geraniol 0.16 1,8-Cineole 0.02
10 Geranial 0.07 Citronellol 0.11 (+)-Citronellal 0.02
11 -Pinene 0.06 (-)--Pinene 0.11 Decanal 0.02
12 Neral 0.05 1,8-Cineole 0.10 -Phellandrene 0.01
13 Citronellal 0.03 Nonanal 0.08 Dodecanal 0.01
14 Neryl acetate 0.01 Nerol 0.07 (+)-Citronellol 0.007
15 -Terpinene 0.01 -Sinensal 0.05 1-Octanol 0.004
a
The Lod values for the lemon, sudachi, and orange oils are odor unit (Uo) values at the recognition threshold level of the Citrus oils against their reference
oils. The recognition thresholds of lemon and Valencia orange were 0.46 ppm and 4.62 ppm against iyo-orange, and the recognition threshold of sudachi was
0.18 ppm against lemon.
S
:

S
e
n
s
o
r
y

&

N
u
t
r
i
t
i
v
e

Q
u
a
l
i
t
i
e
s

o
f

F
o
o
d
S56 JOURNAL OF FOOD SCIENCEVol. 71, Nr. 1, 2006 URLs and E-mail addresses are active links at www.ift.org
Aroma impact compounds in citrus oils . . .
were common to all 3 Citrus oils. Geranyl acetate, neryl acetate, -
pinene, -terpinene, and citral were chemicals peculiar to lemon.
Nonanal, perillaldehyde (lower Lod value), nerol, and -sinensal
were peculiar to orange. p-Cymene, 1-octanol and -phellandrene
were components found only in sudachi oil.
Cross-matching test between the 3 Cross-matching test between the 3 Cross-matching test between the 3 Cross-matching test between the 3 Cross-matching test between the 3
Citrus Citrus Citrus Citrus Citrus oils and their constituents oils and their constituents oils and their constituents oils and their constituents oils and their constituents
The 2-way frequency data are shown in Table 3. The total frequen-
cy of the chemicals classified under lemon was 543. The numbers for
orange and sudachi were 428 and 547, respectively. Citral showed an
excellent matching result for lemon oil (frequency = 59/66,
2
0
= 53.05
>
2
[DF = 44, = 0.2] = 51.64). Linalool and nonanal also matched
well with orange oil (frequency = 44/66,
2
0
= 34.65; 39/66, 22.35, re-
spectively). In the case of sudachi oil, -pinene, -pinene, and
myrcene had higher matching scores (frequency = 42/66,
2
0
= 13.96;
45/66, 18.93; 44/66, 17.19, respectively). As shown in Table 3, the
independence of the 3 categories (the 3 Citrus oils) was calculated by
Table 2Volatile constituents of the lemon, sudachi, and Valencia orange oils and their concentrations
a
Abbrev- Concentration
c
Components iation Origins Odor description
b
(ppm, v/v)
1,8 Cineole 18c Lemon, orange, and sudachi Fresh, diffusive, camphoraceous-cool* 2
Linalool lin Lemon, orange, and sudachi Light and refreshing, floral-woody* 10
Nonanal non Orange Powerful and diffusive fatty-floral, waxy* 40
Geraniol gol Lemon and orange Mild and sweet, floral rose-type* 8
-Terpinene ate Lemon refreshing, lemony-citrusy* 20
Geranyl acetate gac Lemon sweet, fruity-floral, rosy* 100
1-Octanol oco Sudachi Powerful, fresh, orange-rose-like, waxy and sweet* 250
Dodecanal dod Orange and sudachi Sweet, waxy-herbaceous, very fresh and clean-floral odor* 10
Citronellol col Orange and sudachi Fresh rosy odor* 20
Perillaldehyde per Orange fresh green herbal grassy sweet mint cumin** 60
Decanal dal Orange and sudachi Penetrating and very powerful, sweet-waxy, orange-peel-like* 50
Limonene lim Lemon, orange and sudachi Fresh, light and sweet citrusy* 1200
-Terpinene gte Lemon and sudachi Refreshing, herbaceous-citrusy* 300
-Pinene bpi Lemon Dry-woody, resinous-piney* 3500
neryl Acetate nea Lemon Sweet floral rose soapy citrus dewy grapefruit** 150
-Pinene api Lemon, orange, and sudachi Warm-resinous, refreshing pine-like* 800
Myrcene myr Orange and sudachi Sweet-balsamic-resinous* 700
(+)-Citronellal cal Lemon, orange and sudachi Sweet floral rose waxy citrus green** 150
p-Cymene pcy Sudachi Gassy, kerosene-like* 400
-Phellandrene phl Sudachi Fresh-citrusy and peppery-woody* 400
Octanal oca Lemon, orange, and sudachi Harsh-fatty, penetrating* 6
Citral crl Lemon Powerful lemon-fragrance* 200
-Sinensal sin Orange
a
Fishy, metallic*** 100
a
-Sinensal is not listed in the Table 1, but it was found in orange.
b
Odor descriptions are from *Arctander 1969, ** Good Scents Co 2004, and ***Chisholm and others 2003.
c
Each concentration was set to 1000-times the corresponding odor detection threshold.
Table 3Summary of the cross-matching test
Lemon Orange Sudachi Independence of cat-
egory
Counts
2
0
Counts
2
0
Counts
2
0
Row total
2
0

a
1,8-Cineole 35 5.50 21 0.31 10 7.99 66 14.27 **
Linalool 14 3.91 44 34.65 8 10.47 66 33.82 **
Nonanal 5 14.67 39 22.34 22 0.13 66 26.27 **
Geraniol 37 7.60 21 0.03 8 10.47 66 19.18 **
-Terpinene 37 7.60 18 0.02 11 6.78 66 16.45 **
Geranyl acetate 24 0.01 18 0.02 24 0.00 66 1.09
Octanal 16 2.45 26 2.94 24 0.00 66 2.55
Dodecanal 15 3.14 25 2.20 26 0.21 66 3.36
Citronellol 31 2.31 21 0.31 14 4.02 66 6.64 *
Perillaldehyde 22 0.11 23 1.04 21 0.33 66 0.09
Decanal 12 5.71 25 2.20 29 1.14 66 7.18 *
Limonene 26 0.24 13 1.69 27 0.44 66 5.55
-Terpinene 27 0.49 12 2.35 27 0.44 66 6.82 *
-Pinene 15 3.14 6 8.54 45 18.93 66 37.91 **
Neryl acetate 39 10.03 11 3.11 16 2.55 66 20.27 **
-Pinene 12 5.71 12 2.35 42 13.95 66 27.27 **
Myrcene 15 3.14 7 7.24 44 17.19 66 34.45 **
(+)-Citronellal 36 6.50 10 3.98 20 0.60 66 15.64 **
p-Cymene 21 0.29 18 0.02 27 0.44 66 1.91
-Phellandrene 28 0.82 14 1.14 24 0.00 66 4.73
1-Octanol 12 5.71 26 2.94 28 0.75 66 6.91 *
Citral 59 53.05 3 13.09 4 16.46 66 93.36 **
-Sinensal 5 14.67 15 0.70 46 20.76 66 41.55 **
Column total 543 428 547 1518
a
Test of independence for category, DF = 2, * = 0.05, ** = 0.01.
Vol. 71, Nr. 1, 2006JOURNAL OF FOOD SCIENCE S57
S
:

S
e
n
s
o
r
y

&

N
u
t
r
i
t
i
v
e

Q
u
a
l
i
t
i
e
s

o
f

F
o
o
d
URLs and E-mail addresses are active links at www.ift.org
Aroma impact compounds in citrus oils . . .
statistical treatment of the frequency scores [
2
0
>
2
(2, 0.05) = 5.99].
Sixteen odor chemicals out of the 23 compounds were found to sig-
nificantly contribute to the characterization of the 3 Citrus aromas.
Correspondence analysis Correspondence analysis Correspondence analysis Correspondence analysis Correspondence analysis
CA summarized the cross-matching data of the 23 selected
chemicals in 2 dimensions (Table 4). The 1st dimension (the C1 axis)
had a singular value of 0.407 and an inertia of 0.165. This dimension
explained 62% of all data. The 2nd dimension (the C2 axis) had a sin-
gular value of 0.318 and an inertia of 0.101. This dimension explained
38% of all data. The category score of lemon oil was plotted at the
point of C1 = 0.525 and C2 = 0.113. That of orange oil was plotted
at 0.141 for C1 and 0.495 for C2. That of sudachi oil was plotted at
0.411 and 0.276, respectively. Furthermore, the item scores of the 23
compounds are summarized in Table 5. The synchronous scatter
plots (CA biplot) between the 3 oils tested here (Table 4) and their 23
constituents (Table 5) are shown in Figure 1. In this case, the expect-
ed frequencies were 23.61 for lemon oil, 18.61 for orange, and 23.78
for sudachi. The frequencies of perillaldehyde were 22 for lemon, 23
for orange and 21 for sudachi. The scatter plot of perillaldehyde (C1
= 0.01, C2 = 0.15 in Table 5) should be in the neighborhood of the
centroid (expected inertia of C1: 0.165; expected inertia of C2: 0.101)
of all of the data. The C1 axis separated the aroma character of lemon
oil from that of the others. C2 axis separated orange oil from the oth-
ers. Many hydrocarbon terpenes are located in the negative area of
C2 and have a direction similar to that of the sudachi vector (C1:
Figure 1Scatter plots for the 3 Citrus oils and their con-
stituents. The abbreviations for the components are listed
in Table 2. Chemicals within the dotted circle showed
o
2
of category data less than 5.99. = Citrus oils; = chemi-
cals.
Table 4Summary of the correspondence analysis
Category score
Dimension Singular value Inertia Portion Lemon Orange Sudachi
C1 0.407 0.165 0.621 0.525 0.141 0.411
C2 0.318 0.101 0.379 0.113 0.495 0.276
0.411, C2: 0.276). The aliphatic alcohols and aldehydes have the
same direction as the orange vector (C1: 0.141, C2: 0.495). Oxygenat-
ed terpenes such as alcohols, aldehydes, and esters are mainly scat-
tered in the same region as lemon oil. The dotted circle in Figure 1
includes 7 compounds that had
2
o
values lower than
2
(2, 0.05) =
5.99. They were geranyl acetate, 1-octanol, dodecanal, perillalde-
hyde, limonene, p-cymene, and -phellandrene. Those compounds
are neutral and do not make specific contributions to any of the 3
Citrus oils with respect to aroma quality.
Correspondence analysis of lemon. Correspondence analysis of lemon. Correspondence analysis of lemon. Correspondence analysis of lemon. Correspondence analysis of lemon. For the lemon aroma, Lod
picked up limonene (fresh, light, and sweet-citrusy), -terpinene (re-
freshing, herbaceous-citrusy), linalool (light and refreshing, floral-
woody), geraniol (mild and sweet, floral rose-type), 1,8-cineol (fresh,
diffusive, camphoraceous-cool), and other compounds listed in Table
1 as the potent character-impact compounds (see odor description in
Table 2). These compounds are indeed essential for the preparation
of an aroma oil similar to lemon oil. On the other hand, CA with cross-
matched frequency data selected citral, neryl acetate, -terpinene,
citronellal, geraniol, and citronellol as possible chemicals related to the
characteristic aroma of lemon. Some of the chemicals were selected by
both the Lod and the CA method. These common chemicals may be
some of the base odorants for the lemon aroma.
Citral had the highest frequent matching scores (59 counts out of
66 trials) to lemon oil. Both the coordinates of lemon oil and those of
citral were located in the same region on the diagram (Figure 1). The
higher Lod and Uo values of some chemicals indicate a higher prob-
ability that they are potent character-impact compounds. Individual
chemicals selected based on their Lod and Uo values do not always
constitute an image similar to that of the original aroma. The high
Table 5Summary of the items
Item score
Aroma chemicals C1 C2
1,8-Cineole 0.42 0.18
-Pinene 0.47 0.33
-Terpinene 0.46 0.08
-Pinene 0.43 0.52
Citral 1.08 0.30
Citronellol 0.28 0.15
(+)-Citronellal 0.35 0.22
Dodecanal 0.24 0.17
Decanal 0.34 0.14
Geranyl acetate 0.01 0.02
Geraniol 0.49 0.19
-Terpinene 0.05 0.22
Limonene 0.03 0.19
Linalool 0.08 0.86
Myrcene 0.42 0.49
Neryl acetate 0.46 0.16
Nonanal 0.44 0.60
Octanal 0.19 0.21
1-Octanol 0.33 0.18
p-Cymene 0.10 0.04
Perillaldehyde 0.01 0.15
-Phellandrene 0.11 0.14
-Sinensal 0.69 0.28
S
:

S
e
n
s
o
r
y

&

N
u
t
r
i
t
i
v
e

Q
u
a
l
i
t
i
e
s

o
f

F
o
o
d
S58 JOURNAL OF FOOD SCIENCEVol. 71, Nr. 1, 2006 URLs and E-mail addresses are active links at www.ift.org
Aroma impact compounds in citrus oils . . .
frequency of cross-matching between citral and lemon, especially,
indicates that citral created a lemon-like odor impression in all panel
members by itself, as the key compound of lemon oil. CA can theo-
retically treat and resolve the mass category data, but the category
data cannot be subjected to principal component analysis. A combi-
nation of Uo, Lod, GC-O (we did not try this yet), and CA with the
cross-matching test may resolve the question as to which chemicals
are the real key aroma compounds in foods.
Correspondence analysis of orange. Correspondence analysis of orange. Correspondence analysis of orange. Correspondence analysis of orange. Correspondence analysis of orange. Based on the Lod values,
linalool (refreshing floral-woody), limonene, octanal (harsh-fatty),
dodecanal (sweet, waxy-herbaceous and floral), myrcene (sweet-
balsamic-resinous), decanal (orange-like aroma), nonanal (power-
ful and diffusive fatty-floral, waxy), and other compounds were
selected as character impact compounds of Valencia orange oil (see
odor description in Table 2). The CA biplots of nonanal and linalool
in Figure 1 show an aroma similar to that of orange oil. Nonanal,
found only in orange, had a moderate Lod value (0.08) and contrib-
uted to the fatty-floral aroma. (+)-Linalool and (-)-linalool were
found in orange, and had higher Lod values in the oil. They signif-
icantly contributed to the refreshing floral-woody aroma in orange,
according to the sensory cross-matching test and the Lod method.
Other chemicals, such as perillaldehyde and -sinensal, were found
in orange oil as well, but those compounds did not have CA biplots
in the region of the coordinates of orange oil. -Sinensal has a me-
tallic-fishy note (Chisholm and others 2003) and a low odor thresh-
old, whereas -sinensal possesses an orange-like aroma (GSC
2004). Overall, the sweet aroma derived from linalool and nonanal
might contribute to the orange-like aroma.
Correspondence analysis of C. sudachi. Correspondence analysis of C. sudachi. Correspondence analysis of C. sudachi. Correspondence analysis of C. sudachi. Correspondence analysis of C. sudachi. Citrus sudachi is 1 of the
Citrus species grown in Shikoku, Japan. It has a peculiar odor quality.
Padrayuttawat and others (1997) reported that the character impact
compounds are linalool (light and refreshing, floral-woody), octanal
(harsh-fatty, penetrating), limonene (fresh, light and sweet-citrusy),
-pinene (warm-resinous, refreshing pine-like), p-cymene (gassy,
kerosene-like), -terpinene (refreshing, herbaceous-citrusy), 1,8-
cineole (fresh, diffusive, camphoraceous-cool), (+)-citronellal (sweet,
floral-rose, waxy, green-citrusy), and decanal (penetrating and very
powerful, sweet-waxy, orange-peel-like), based on Lod values (see
odor description in Table 2). Two unknown components, which had
spicy and butter-like aromas, were also found in C. sudachi through
GC-Olfactometry. Using GC-O and Lod, citronellal, citronellol, and
carvone as well as the 2 unidentified compounds were proposed as
contributors to the aroma of C. sudachi (Tamura and others 1993). On
the other hand, CA characterized sudachi oil as containing -sinen-
sal, -pinene, -pinene and myrcene. These components are char-
acteristically related to piney, spicy, and phenolic odors. The 2 un-
known chemicals, which have spicy and butter-like aromas, may be
indispensable for the C. sudachi aroma. In the cross-matching test,
the subjects found sudachi oil to have a citrusy, woody, resinous, and
harsh aroma. This result of the CA may give us a hint regarding the
aroma character of the 2 unknown compounds which greatly con-
tribute to the C. sudachi aroma.
Conclusions
T
he cross-matching procedure used in this study for the charac-
terization of the Citrus aromas of lemon, Valencia orange, and
Citrus sudachi was based on a nonverbal method and reflex human
decision-making. The subjects classified the 23 chemicals under 3
categories (three Citrus oils). Altogether, 66 frequencies were col-
lected to get precise sensory data. The frequencies reflect the prob-
ability of classification of the chemicals under the 3 Citrus species.
By correspondence analysis (CA) of the category data (frequency
data), the contingency table was summarized as a CA biplot, and
the 23 chemicals were attributed to the 3 Citrus oils as potent char-
acter-impact compounds. Whenever a chemical with a higher Lod
value had a greater frequency score for 1 of the fruits, it was consid-
ered to be a key aroma compound of the fruit. This simple and
quick sensory test using CA with cross-matching may replace and
reduce the number of model chemicals needed to maintain an aro-
ma quality similar to that of the original oil.
References
Arctander, S. 1969. Perfume and flavor compounds. Vol. I, II. Montclair, N.J.: Det
Hoffensbergske.
Belitz HD, Grosch W. 1986. Aroma substances. In: Food chemistry. Berlin, Heidel-
berg: Springer Verlag. 258 p.
Blasius J, Thiessen V. 2000. Methodological artifacts in measures of political
efficacy and trust: a multiple correspondence analysis. Pol Anal 9:120.
Boonbumrung S, Tamura H, Mookdasanit J, Nakamoto H, Ishihara M, Yoshizawa
T, Varanyanond W. 2001. Characteristic aroma components of the volatile oil of
yellow Keaw mango fruits determined by limited odor unit method. Food Sci
Technol Res 7:2006.
Buchanan TW, Tranel D, Adolphs R. 2003. A specific role for the human amygdala
in olfactory memory. Learn Mem 10:31925.
Buttery RG, Teranishi R, Ling LC, Turnbaugh. 1990. Quantitative and sensory
studies on tomato paste volatile. J Agric Food Chem 38:33640.
Chisholm MG, Jell JA, Cass DM. 2003.Characterization of the major odorants
found in the peel oil of Citrus reticulata Blanco cv. Clementine using gas chro-
matography-olfactometry. Flavour Fragr J 18:27581.
Dimov NA, Tsoutsoulova A, Stojanov E. 1987. Pattern recognition methods for
discrimination of essential oils (rose oils) by their gas chromatograms. Per-
fum Flav 12:458.
Greenacre MJ. 1994. Correspondence analysis and its interpretation. In: Green-
acre MJ, Blasius J, editors. Correspondence analysis and the social sciences.
London: Academic Press. p 322.
Hoffman D, Franke G. 1986. Correspondence analysis: graphical representa-
tion of categorical data in marketing research. J Market Res 23:21327.
Forina M, Tiscorina E. 1982. Pattern recognition methods in the prediction of
Italian olive oils by their fatty acid content. Ann Chim 72:1436.
Garcia-Vallv S, Romeu A, Palau J. 2000. Horizontal gene transfer in bacterial
and archaeal complete genomes. Genome Res 10:171925.
[GSC] Good Scents Co. 2004. Information for perfumery. Flavor Row Materials.
Available from: www.thegoodscentscompany.com. Accessed Dec 5, 2005.
Grosch W. 2001. Evaluation of the key odorants of foods by dilution experiments
aroma models and omission. Chem Senses 26: 53345.
Guadagni DG, Buttery RG, Harris J. 1966. Odor intensities of hop oil compo-
nents. J Sci Food Agric 17:1424.
Guadagni DG, Okano S, Buttery RG, Burr HK. 1966. Correlation of sensory and
gas-liquid chromatographic measurements of apple volatiles. Food Technol
20:1669.
Hoffman D, Franke G. 1986. Correspondence analysis: graphical representa-
tion of categorical data in marketing research. J Market Res 23:21327.
Mookdasanit J, Tamura H, Yoshizawa T, Tokunaga K. 2003. Trace volatile compo-
nents in essential oil of Citrus sudachi by means of modified solvent extrac-
tion method. Food Sci Technol Res 9(1):5461.
Mayfield HT, Bertsch W, Mar T, Staroscik JA. 1986. Application of chemometrics
to the classification of orange oil varieties by GLC. J High Resol Chromatogr
Comm 9:7883.
Noble AC. 1978. Sensory and instrumental evaluation of wine aroma. In:
Charalambous G, editor. Analysis of foods and beverages. New York: Academic
Press. p 20328.
Padrayuttawat A, Yoshizawa T, Tamura H, Tokunaga T. 1997. Optical isomers and
thresholds of volatile constituents in Citrus sudachi. Food Sci Technol Int To-
kyo 3:4028.
Panagiotakos DB, Pitsavos C. 2004. Interpretation of epidemiological data using
multiple correspondence analysis and log-linear models. J Data Sci 2:7586.
Roberts DD, Acree TE. 1996. Effects of heating and cream addition on fresh rasp-
berry aroma using retronasal aroma simulator and gas chromatography. J
Food Sci 44:391925.
Schieberle P, Grosch W. 1990. Evaluation of potent odorants in cucumbers (Cucu-
mis sativus) and muskmelons (Cucumis melo) by aroma extract dilution anal-
ysis. J Food Sci 55:1935.
Tekaia F, Lazcano A, Dujon B. 1999. The genomic tree as revealed from whole
proteome comparisons. Genome Res 9:5507.
Tamura H, Fukuda Y, Padrayuttawat A. 1996. Characterization of Citrus aroma
quality by odor threshold values. In: Takeoka GR, Teranishi R, Williams PJ,
Kobayashi A, editors. Biotechnology for improved foods and flavors. Washing-
ton, D.C.: American Chemical Society. ACS symposium series 637:28294.
Tamura H, Yang RH, Sugisawa H. 1993. Aroma profiles of peel oils of acid Citrus.
In: Teranishi R, Buttery RG, and Sugisawa H, editors. Bioactive volatile com-
pounds from plants. Washington, D.C.: American Chemical Society. ACS sym-
posium series 52512136.
van de Geer JP. 1993. Multivariate analysis of categorical data. In: van de Geer
JP, editor. Advanced quantitative techniques social science. London: SAGE pub-
lications. Vol 3.
van der Heijden PGM, de Falguerolles A, Jan de Leeuw. 1989. A combined ap-
proach to contingency table analysis using correspondence analysis and log-
linear analysis. Appl Statist 38:24992.
Zheng Y, Yoshino R. 2003. Diversity patterns of attitude toward nature and envi-
ronment in Japan, USA and European nations. Behaviormetrika 30(1):2137.

Anda mungkin juga menyukai