Correlation between cup quality and chemical attributes of Brazilian coffee

Abstract

Brazilian arabica coffee is classified for trading according to the quality of the beverage obtained after roasting and brewing. In the present study, Brazilian green and roasted coffee beans were investigated for possible correlations between cup quality and the levels of sucrose, caffeine, trigonelline and chlorogenic acids, determined by HPLC analysis. Trigonelline and 3,4-dicaffeoylquinic acid levels in green and roasted coffee correlated strongly with high quality. To a lesser extent, caffeine levels were also associated with good quality. On the other hand, the amount of defective beans, the levels of caffeoylquinic acids (predominantly 5-caffeoyilquinic acid), feruloylquinic acids, and their oxidation products were associated with poor cup quality and with the Rio-off-flavor. The fact that similar correlations between cup quality and chemical attributes were observed in green and light roasted samples – the latter used for coffee cup classification – indicates that chemical analysis of green beans may be used as an additional tool for coffee quality evaluation.

Full text

Analytical, Nutritional and Clinical Methods
Correlation between cup quality and chemical
attributes of Brazilian coffee
A. Farah
a,*
, M.C. Monteiro
a
, V. Calado
b
, A.S. Franca
c
, L.C. Trugo
a
a
Laborato
´rio de Bioquı
´mica Nutricional e de Alimentos, Instituto de Quı
´mica, Universidade Federal do Rio de Janeiro,
Ilha do Funda
˜o, CEP:21949-900, Rio de Janeiro, RJ, Brazil
b
Escola de Quı
´mica, Universidade Federal do Rio de Janeiro, Ilha do Funda
˜o, CEP:21949-900, Rio de Janeiro, RJ, Brazil
c
Nu
´cleo de Pesquisa e Desenvolvimento em Cafe
´, DEQ/UFMG, R. Espı
´rito Santo, 6

andar, CEP 30160-030, Belo Horizonte, MG, Brazil
Received 4 April 2005; received in revised form 5 July 2005; accepted 30 July 2005
This work is dedicated to the memory of our dear friend, colleague and mentor, Prof. Luiz Carlos Trugo, after his untimely death on November 20,
2004. His example will keep inspiring all who had the privilege of knowing him.
Abstract
Brazilian arabica coffee is classified for trading according to the quality of the beverage obtained after roasting and brewing. In
the present study, Brazilian green and roasted coffee beans were investigated for possible correlations between cup quality and the
levels of sucrose, caffeine, trigonelline and chlorogenic acids, determined by HPLC analysis. Trigonelline and 3,4-dicaffeoylquinic
acid levels in green and roasted coffee correlated strongly with high quality. To a lesser extent, caffeine levels were also associated
with good quality. On the other hand, the amount of defective beans, the levels of caffeoylquinic acids (predominantly 5-caffeoyil-
quinic acid), feruloylquinic acids, and their oxidation products were associated with poor cup quality and with the Rio-off-flavor.
The fact that similar correlations between cup quality and chemical attributes were observed in green and light roasted samples – the
latter used for coffee cup classification – indicates that chemical analysis of green beans may be used as an additional tool for coffee
quality evaluation.
Ó2005 Elsevier Ltd. All rights reserved.
Keywords: Coffee quality; Rio-off-flavor; Chlorogenic acids; Trigonelline; Caffeine
1. Introduction
Flavor is the most important criterion for coffee qual-
ity evaluation, and also one of the major motivations for
consumer preferences (Cantergiani et al., 1999; Clarke,
1987, chap. 2). The assessment of coffee quality by both
buyers and sellers in Brazil is based on the brewing
method of steeping, which consists on pouring boiling
water (150 mL) directly onto roasted and ground cof-
fee (10 g; mild roast; fine grind) contained in a small
cup and performing sensory (smell, flavor) evaluation
after a few minutes (Clarke, 1987, chap. 2; Lingle,
1993). Classification is performed by a trained panel,
and the beverage quality denominations, from best to
worst, are: Strictly soft, Soft, Barely soft, Hard, Rioysh,
Rio and Rio zona (Table 1). This determines the so
called cup quality. The lowest quality coffees (Rioysh,
Rio and Rio zona) are associated with the Rio off-flavor,
usually described as a pungent, medicinal, phenolic or
iodine-like flavor associated with a musty, cellarlike
odor (Lingle, 1993; Spadone, Takeoka, & Liardon,
1990).
The presence of defects is also relevant in establishing
Brazilian coffee quality, since they are associated with
0308-8146/$ - see front matter Ó2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodchem.2005.07.032
*
Corresponding author. Tel.: +55 21 2562 7351; fax: +55 21 2562
8213.
E-mail address: afarah@iq.ufrj.br (A. Farah).
www.elsevier.com/locate/foodchem
Food Chemistry 98 (2006) 373–380
Food
Chemistry

problems during harvesting and pre-processing opera-
tions. The term defect is used in reference to the presence
of defective (black, sour or brown, immature, immature-
black, bored, broken, etc.) beans and also of extraneous
matter (husks, twigs, stones, etc.) in a given coffee sam-
ple (Clarke, 1987; Franca, Oliveira, Mendonc¸a, & Silva,
2005; Franca, Mendonc¸a, & Oliveira, 2005; Mazzafera,
1999). Black beans are those from over-ripened fruits.
Sour beans are from fruits that are fermented on the
ground or due to improper processing conditions.
Immature beans come from immature fruits, imma-
ture-black are beans from immature fruits in which the
skin is oxidized, and bored beans are those damaged
by insect action. Even though defects are known to neg-
atively affect coffee flavor, the total counting of defects
alone cannot be used to accurately predict cup quality
(Smith, 1985, (chap. 1)).
The chemistry of flavor development during coffee
roasting is highly complex and not completely under-
stood. Even though roasting process appears to be sim-
ple in terms of processing conditions, it is quite complex
from a chemistry point of view, since hundreds of chem-
ical reactions take place simultaneously. Examples in-
clude Maillard and Strecker reactions, degradation of
proteins, polysaccharides, trigonelline and chlorogenic
acids (De Maria, Trugo, Aquino Neto, Moreira, & Alvi-
ano, 1996). Sugars, particularly sucrose as the most
abundant, will act as aroma precursors, originating sev-
eral substances (furans, aldehydes, carboxylic acids, etc.)
that will affect both flavor and aroma of the beverage.
Trigonelline is a pyridine derivative, known to contrib-
ute indirectly to the formation of desirable aromas dur-
ing roasting (Ky, 2001; Macrae, 1985, (chap. 4)).
Caffeine, a xantine derivative, presents a characteristic
bitter taste reported to be important to coffee flavor
(Trugo, 1984). This compound has also been the subject
of several investigations in view of its pharmacological
effects (Azam, Hadi, Khan, & Hadi, 2003; Barone &
Roberts, 1996; Macrae, 1985, Ribeiro-Alves, Trugo, &
Donangelo (chap. 4); Ribeiro-Alves et al., 2003). Chlor-
ogenic acids (CGA), a group of phenolic compounds
that represent 6–12% of coffee constituents in mass (Far-
ah, De Paulis, Trugo, & Martin, 2005), are known to be
responsible for coffee pigmentation, aroma formation,
and astringency (De Maria, Trugo, Moreira, & Petr-
acco, 1995; Trugo, 1984). Furthermore, thermal degra-
dation of chlorogenic acids during roasting will result
in phenolic substances that contribute to bitterness (Clif-
ford, 1985, chap. 5). The major CGA subgroups in cof-
fee are the caffeoylquinic acids (CQA), feruloylquinic
acids (FQA) and dicaffeoylquinic acids (diCQA) (Clif-
ford & Wight, 1976; Trugo & Macrae, 1984). These
compounds have received much attention lately due to
various pharmacological activities observed in vitro
and in animals (Farah et al., 2005).
Even though more than eight hundred volatile and
non-volatile compounds have been already identified
in coffee, the question of which constituents are the most
relevant contributors to low cup quality coffee is contro-
versial and far from being completely answered, espe-
cially in regard to the Rio-off-flavor. According to
Spadone et al. (1990), 2,4,6-trichloroanisole and 2,4,6-
trichlorophenol were identified as two of the compo-
nents responsible for the Rio-off-flavor. While Amorim,
Basso, Crocomo, and Teixeira (1977) have not observed
a correlation between cup quality and the levels of poly-
amines, Oliveira, Franca, Glo
´ria, and Borges (2005)
found higher levels of amines in lower quality coffee
samples, comparing to those of good quality. Mazzafera
(1999) and Franca et al. (2005) associated higher acidity
with low cup quality, possibly due to the presence of
defective coffee beans, specifically the ones that had
undergone fermentation. Chagas (1994) observed a
positive association between the levels of reducing and
non-reducing sugars and cup quality. 2-Methylbutyral-
dehyde and 3-methylbutyraldehyde were described as
two of the volatile compounds characteristic of green
defective beans and low cup quality coffee blends
(Cunha, 2005).
In view of the above, a more extensive investigation
of chemical attributes of Brazilian coffees of different
cup qualities is needed. Therefore, the objective of the
present study was to investigate the existence of a possi-
ble correlation between cup quality and the content of
some of the most important compounds in coffee: su-
crose, trigonelline, caffeine and chlorogenic acids.
2. Material and methods
2.1. Samples
Arabica coffee samples classified as Soft, Hard,
Rioysh, Rio and Rio zona, (Pinhal, Sa
˜o Paulo, Brazil)
were provided by the Brazilian Association of Coffee
Industry (ABIC). Samples of randomly selected beans
were separated from each lot and the defective (black,
sour, immature, bored) and non-defective beans were
manually separated and weighted in order to determine
the mass composition of defective beans for each lot.
Table 1
Official classification of Brazilian coffee beverage (Bartholo & Gui-
mara
˜es, 1997)
Classification Characteristics
Strictly soft Very smooth flavor; slightly sweet; low acidity
Soft Smooth flavor; slightly sweet
Barely soft Smooth flavor, but with slight astringency
Hard Astringent flavor; rough taste; lacks sweetness
Rioysh Slight taste of iodoform or phenic acid
Rio Strong unpleasant taste, reminding iodoform or
phenic acid
Rio zona Intolerable taste and smell
374 A. Farah et al. / Food Chemistry 98 (2006) 373–380

Since it is very difficult to distinguish sour from black-
immature beans and the latest constitute a minor defect,
comparing to the first, black-immature and sour beans
were counted as sour beans. Three hundred grams of
each coffee sample were roasted at 200 °C, in an electric
labscale roaster (CAEL LTDA, Brazil), for 8, 12, 27 and
45 min, resulting in light, medium, dark and very dark
roasting degrees, respectively. Roasting degrees were
established according to the reference color system used
by ABIC. Disks #75; #55; #35 and #25, from the Roast
Color Classification System (Agtron-SCAA, 1995), were
used as standards for light; medium; dark and very dark
roasting degrees. Temperature variation inside the roast-
er was monitored by a thermometer during the roasting
process. Roasted coffee beans were then grounded to
pass through a 0.75 mm sieve and the color was deter-
mined by reflectance measurements (Color Test II –
Neuhaus Neotec, Germany), as an average of three
determinations for each sample. A sample of C.cane-
phora cv Conillon or Robusta coffee (Espı
´rito Santo,
Brazil), known to be of inferior quality (Martin, Pablo,
& Gonsalez, 1998), was also roasted for comparison
with the arabica samples.
2.1.1. Extraction
For sucrose evaluation, coffee samples were extracted
with 80 °C distilled water and clarified with activated
charcoal (Trugo, Farah, & Cabral, 1995). Caffeine and
trigonelline were also extracted with hot distilled water.
The extracts were clarified with Carrez reagents for caf-
feine analysis, and with lead acetate (60%) for trigonel-
line analysis (Trugo, 1984). CGA were extracted with
aqueous methanol (40%) and clarified with Carrez re-
agents (Trugo & Macrae, 1984). All extractions were
performed in triplicates.
2.1.2. HPLC analysis
Isocratic systems consisting of Shimadzu pump and
integrator (Japan) and a 20 lL Rheodyne fixed loop
injector (USA) were used for all analyses. Sucrose was
determined by a RI detector (Waters, USA), using a
Spherisorb NH
2
column (Supelco, USA), and acetoni-
trile 80% as mobile phase, at 1 mL/min (Trugo et al.,
1995). For caffeine and trigonelline analysis, a Knauer
UV detector (Germany) was employed at 272 nm for
caffeine and at 264 nm for trigonelline. A reverse-phase
column (ODS C-18 Merck, Germany) was used with
methanol as mobile phase at 40% for caffeine and 5%
for trigonelline. Flow rate was 1 mL/min (Trugo,
1984). For CGA analysis, a reverse-phase column (Rex-
chrom, Regis, USA) and a Shimadzu UV detector at
325 nm were used. The mobile phase was 10 mM triso-
dium citrate: methanol (65:35 v/v), adjusted to pH 2.5
with HCl, at 1 mL/min. CGA were identified with the
use of non-commercial standards, as previously de-
scribed (Farah et al., 2005). The concentrations of the
compounds were calculated using the peak areas of
commercial standards (Aldrich Chem., USA) as refer-
ences. Because the only commercially available CGA
standard is 5-CQA, the quantification of the CGA iso-
mers took into account the area of 5-CQA standard
and the area and molar extinction coefficients of each
CGA (Farah et al., 2005; Ruback, 1969). All reagents
employed for HPLC analysis were of HPLC grade and
the others were of analytical grade.
2.1.3. Water content
In order to express the results in dry matter (dm) the
water content of each sample was determined according
to AOAC procedures (AOAC, 2000).
2.1.4. Statistical analysis
The HPLC results were tested for correlations with
the Statistica
Ò
software, version 6.0, using the least
square difference method and considered significant
when p< 0.05. The correlations with cup quality were
obtained using data from all samples and the correla-
tions with ‘‘Rio-off-flavor’’ excluded data from the Hard
sample. Because sample is a qualitative variable, we
found inappropriate to propose an equation to describe
the correlations between the dependent and independent
variables.
3. Results and discussion
The arabica coffee samples used in the present study
were classified by cup quality as Soft, Hard, Rioysh,
Rio and Rio zona. The sample classified as soft pre-
sented a soft, sweet note and lacked off-notes. The Hard
one presented a distinct astringent note, and the three
last types of beverage showed an increasing perception
of the Rio-off-flavor.
3.1. Defective beans
Average results for the distribution of defective and
non-defective beans in the coffee samples are shown in
Fig. 1. The highest quality sample (Soft) consisted of
100% non-defective beans. The percentage of defective
beans increased as cup quality decreased. Franca, Men-
donc¸a, et al. (2005) also reported higher percentage of
defects in Rioysh and Rio samples, when compared to
Soft and Hard samples. The percentage of each type
of defect (immature, bored and sour) also increased as
cup quality decreased. Black beans were only found in
the Rio zona sample. The prevailing defect in all samples
consisted of sour beans, followed by immature beans.
This could be an indication that the percentage of sour
and immature beans could present a positive correlation
with low cup quality and/or with the Rio-off-flavor.
Further studies employing a wide variety of arabica
A. Farah et al. / Food Chemistry 98 (2006) 373–380 375

coffee samples from different crops and locations are
needed in order to verify this possibility.
3.2. Color of the beans and roasting temperature
The color intensity of the green arabica samples in-
creased significantly as their quality decreased (Fig. 2).
Such behavior could be attributed to the increasing pres-
ence of defective beans as cup quality decreased (Fig. 1)
and, indirectly, to the action of the enzyme polyphenol
oxidase over phenolic acids (see below). Franca, Oli-
veira, et al. (2005) showed that both black and sour
defective beans presented lower luminosity and color
saturation values in comparison to non-defective ones.
Even though no significant changes in temperature were
recorded inside the roaster for the same time points,
there was a small but significant difference in the color-
imetric values of arabica samples roasted to the same
roasting degrees. This difference did not correspond to
the differences in colorimetric values observed in the
green samples (Fig. 3).
3.3. Trigonelline
Derivatives of trigonelline are known to be important
to the coffee aroma (Trugo, 1984). As quality worsened,
the levels of trigonelline in the green beans decreased
from 1.34 ± 0.05 (dry matter-dm) to 0.96 ± 0.03 g/
Immature 3.2%
Sour 4.3%
Bored 1.0%
Non-defective
91.6%
Immature 0.4% Sour 1.7%
Bored 0.8%
Non-defective
97.1%
Immature 11.6%
Sour 13.7%
Bored 11.8%
Non-defective
61.9%
Black 1.0%
Immature 7.1%
Sour 9.4%
Bored 3.1%
Non-defective
80.3%
ab
cd
Fig. 1. Distribution of defective beans in Brazilian arabica coffee samples classified by cup quality: (a) Hard; (b) Rioysh; (c) Rio; (d) Rio zona.
Soft Hard Rioysh Rio Rio Zona
Type of Beverage
162
164
166
168
170
172
174
176
178
180
182
184
186
Degree of Luminosity
Fig. 2. Correlation between Brazilian arabica green samples classified
by cup quality and color, measured by intensity of luminosity. The
negative correlation indicates that as the samples worsen, they get
darker (r=0.96).
Soft
Hard
Rioysh
Rio
Rio Zona
Robusta
Green Light Medium Dark Very dark
Roasting degree
60
80
100
120
140
160
180
200
Degree of luminosity
Fig. 3. Colorimetric results obtained for green and roasted Brazilian
arabica coffee classified by cup quality and for Robusta coffee beans.
376 A. Farah et al. / Food Chemistry 98 (2006) 373–380

100 g (dm), as shown in Fig. 4, resulting in a strong neg-
ative correlation with poor quality (r=0.93) and with
the Rio-off-flavor (r=0.94) (Fig. 5). These levels are
within the range previously reported for samples from
different sources (Ky, 2001; Trugo, 1984). Both higher
(Martin et al., 1998; Mazzafera, 1991) and lower (Fran-
ca, Mendonc¸a, et al. (2005)) amounts of trigonelline
have also been reported for Brazilian coffee samples,
which could be attributed to the use of different analyt-
ical methods. The roasting process caused a significant
decrease in trigonelline content, as expected (Amorim
et al., 1975; Franca, Oliveira, et al. (2005); Franca, Men-
donc¸a, et al. (2005); Trugo, 1984)(Fig. 4). The average
loss of trigonelline from the green beans to the dark
roasted beans was 90%. Trigonelline losses of 50–80%
after roasting have been previously reported (Franca,
Oliveira, et al. (2005); Franca, Mendonc¸a, et al.
(2005); Trugo, 1984). These differences may be attrib-
uted to distinct roasting conditions, which include differ-
ences in colorimetric standards, since trigonelline
degradation was reported to be strongly dependent
upon the degree of roast (Borges, Mendonc¸a, Franca,
& Oliveira, 2004; Casal, Oliveira, & Ferreira, 2000; Tru-
go, 1984). The negative correlations observed in the
green beans were higher for the light roasted beans for
both quality and the Rio-off-flavor (Fig. 5). The con-
tents in darker roasting degrees did not show any signif-
icant correlations with cup quality. Fig. 4 also presents
the content of trigonelline in a Robusta sample, known
to be of lower quality than arabica coffee (Martin et al.,
1998; Trugo & Macrae, 1984). Trigonelline levels were
lower for green Robusta beans in comparison to all
arabica beans, but quite similar to the levels found in
the worst quality arabica sample, Rio zona.
3.4. Sucrose
The highest sucrose level in green beans was found in
the Rioysh sample (7.85 ± 0.26 g/100 g, dm) and the
lowest level, in the Rio zona sample (4.88 ± 0.10 g/
100 g dm). There was a drastic decrease in the sucrose
Fig. 4. Trigonelline content in green and roasted Brazilian arabica coffee samples classified by cup quality and in Robusta coffee beans.
-0.93
-0.97
0.67
0.51
-0.74
-1.5 -1 -0.5 0 0.5 1
Trigonelline
Cafeine
Sucrose
Correlation Coeficient
Green
a
b
Light roast Medium roast
-0.94
-0.64
-0.95
-0.87
0.74
0.5
-0.63
-1.5 -1 -0.5 0 0.5 1
Trigonelline
Cafeine
Sucrose
Correlation Coeficient
Green Light roast Medium roast
Fig. 5. Correlation coefficients between: (a) cup quality or (b) Rio-off-
flavor and the content of sucrose, caffeine, trigonelline. The right side
of the x-axis indicates: (a) poor cup quality or (b) presence of Rio-off-
flavor.
A. Farah et al. / Food Chemistry 98 (2006) 373–380 377

content of all samples during the roasting process (aver-
age loss of 98% for the dark roasted samples), as a con-
sequence of caramelization and Maillard reactions.
Although Mazzafera (1999) had found lower sucrose
levels in defective green beans in comparison with
non-defective (good quality) green beans, in the present
work, the content of sucrose in green and roasted beans
correlated neither with quality nor with the Rio-off-fla-
vor (Fig. 5).
3.5. Caffeine
The caffeine content in the green beans showed a
small but significant difference. The highest content
was observed for the Soft sample (1.23 ± 0.06 g/100 g
dm) and the lowest content, for the Hard sample
(0.96 ± 0.01 g/100 g dm). As expected, roasting did not
affect the content of caffeine other than causing a slight
relative increase due to the loss of other components.
Caffeine correlated negatively with the Rio-off-flavor
only for the light roasted samples (r=0.87, Fig. 5).
No other correlations were of significance. Franca,
Mendonc¸a, et al. (2005) have also reported higher caf-
feine content for their highest quality sample (Soft),
compared to other arabica samples.
3.6. Chlorogenic acids
Eight CGA were identified. The caffeoylquinic acids
accounted for about 83% of the total CGA in green
beans. The highest content of total CGA in the green
beans were observed for the worst quality sample, Rio
zona (7.02 ± 0.17 g/100 g dm), and the lowest content,
for the best quality sample, Soft (5.78 ± 0.09 g/100
dm) (Fig. 6). CGA levels decreased gradually during
roasting, as previously reported (Farah et al., 2005; Tru-
go, 1984). The average loss of total CGA from green to
dark roasted beans was 93%. One of the descriptions of
the Rio-off-flavor is a phenolic note (Spadone et al.,
1990). Strong positive correlations were found between
the levels of most of CGA monoesters and low cup qual-
ity. Fig. 7 shows the correlation coefficients between
individual CGA and cup quality (a) or Rio-off-flavor
(b). In the green beans, the levels of 5-CQA and 5-
FQA correlated strongly with poor cup quality
(r= 0.90). The levels of 4-CQA, 5-CQA, 4-FQA and
5-FQA correlated strongly with the Rio-off-flavor
(r= 0.93, 0.94, 0.82 and 0.90, respectively). In contrast,
the levels of 3,4-diCQA were negatively correlated with
poor cup quality (r=0.88) and with the Rio-off-flavor
(r=0.83). To a lesser extent, 3,5-diCQA levels in the
green beans also correlated negatively with the Rio-off-
flavor (r=0.75). As displayed in Fig. 7, even though
not all the dicaffeoylquinic acids have shown high corre-
lations with cup quality and the Rio-off-flavor, in the
green samples, the correlation coefficients for all three
dicaffeoylquinic acids showed a tendency towards the
left side of the xaxis, which indicates good quality. Con-
versely, Ohiokpehai, Brumen, and Clifford (1982) re-
ported that the addition of dicaffeoyilquinic acids
conferred a disagreeable flavor to coffee beverage, which
disappeared on subsequent addition of monocaffeoyil-
quinic acid.
Menezes (1994) have observed an inverse association
of the levels of CQA and coffee fruits maturation. We
have also observed that immature and immature-black
defective beans contain significantly higher levels of all
CGA, but mostly CQA and FQA, comparing to healthy
and black defective beans (unpublished data). Accord-
ing to Mazzafera (1999) immature and immature-black
Fig. 6. 5-CQA content in green and roasted Brazilian arabica coffee samples classified by cup quality and in Robusta coffee beans.
378 A. Farah et al. / Food Chemistry 98 (2006) 373–380

beans are critical defects for coffee quality. The fact that
in the present study as cup quality decayed the percent-
age of immature beans and the levels of 5-CQA in-
creased, reiterates the association of 5-CQA with poor
quality.
For the light roasted beans, only 3-CQA, 4-CQA and
5-CQA levels presented positive correlations with both
poor cup quality (r= 0.87, 0.81, 0.9, respectively) and
the Rio-off-flavor (r= 0.93, 0.87, 0.86, respectively).
For the beans roasted to medium degree, the levels of
5-CQA, 4-FQA and 5-FQA were correlated with poor
cup quality (r= 0.83; r= 0.96; r= 0.81, respectively),
and the levels of 3-CQA, 4-CQA, 5-CQA, 4-FQA, and
5-FQA were positively correlated with the Rio-off-flavor
(r= 0.75, 0.85, 0.93, 0.95, respectively). Correlations
were not significant for darker roasting degrees.
3.7. Chlorogenic acids and color
Similar to that observed for 5-CQA, as sample qual-
ity decreased, the color intensity of green arabica beans
increased significantly (r=0.96). A high correlation
was found between intensity of color and the content
of 5-CQA and total CGA (r= 0.90, 0.80, respectively).
When considering only the samples with Rio-off-flavor,
the correlations were even higher for 5-CQA and total
CGA (r= 0.91, 0.93, respectively). Because 5-CQA ac-
counts for 61.8% of the total CGA, there may be an
important contribution of this compound to color inten-
sity, augmented by other CGA. 5-CQA is very likely to
be the major substrate for the enzyme polyphenol oxi-
dase in coffee (Mazzafera & Robinson, 2000). Therefore,
another possibility is that ortho-quinones, formed by the
action of polyphenol oxidase on 5-CQA, cause darken-
ing of the grains. Moreover, Amorim, Cruz, Dias, Mel-
lo, and Teixeira (1977) related the action of polyphenol
oxidase, triggered by structural changes of bean cell
membranes, as a possible cause of the Rio-off-flavor.
This hypothesis is in accordance with our results and
indicates that oxidation products of 5-CQA may con-
tribute to the Rio-off-flavor.
4. Conclusions
In the present study, trigonelline and 3,4-dicaffeoilqui-
nic acid and, to a lesser extent, caffeine, showed associa-
tion with good cup quality, for both green and light
roasted coffee. In contrast, along with the amount of
defective beans, higher levels of caffeoylquinic acids (pre-
dominantly 5-CQA), feruloylquinic acids (to a lesser ex-
tent), and their oxidation products were associated with
poor cup quality and with the Rio-off-flavor. The fact
that similar correlations between cup quality and chem-
ical attributes were observed in green and light roasted
samples – the latter used for coffee cup classification –
indicates that chemical analysis of green beans may be
used as an additional tool for evaluating coffee quality.
Acknowledgements
The authors greatly acknowledge the financial sup-
port of FUNCAFE-EMBRAPA, Conselho Nacional
de Pesquisa (CNPq) and Fundaca
˜o Carlos Chagas Filho
de Amparo a
`Pesquisa (FAPERJ), Brazil. We also
would like to thank Monica C. Pinto from ABIC – Bra-
zil, for supplying us with the coffee samples and advice
on national roasting standards, Juliana Marquito, from
CAFE
´TOKO Ltda, Brazil, for assistance with colori-
metric data, Juliana C.F. Mendonc¸a (NPDC/UFMG)
for assistance with sample classification and Carmen
Marino Donangelo for assistance with revision.
References
Amorim, H. V., Teixeira, A. A., Melo, M., Cruz, V. F., & Malavolta,
E. (1975). Chemistry of Brazilian green coffee and quality of
0.56
0.66
0.9
0.9
-0.88
0.82
0.81
0.88
-
0.96
0.8
0.83
-1 -0.5 0 0.5 1 1.5
3-CQA
4-CQA
5-CQA
4-FQA
5-FQA
3,4-diCQA
3,5-diCQA
4,5-di
a
b
CQA
Correlation Coeficient
Green Light roast Medium roast
0.73
0.93
0.94
0.82
0.9
-0.83
-0.75
0.93
0.87
0.86
-0.48
0.51
0.75
0.87
0.93
0.95
0.67
0.58
-
1 -0.5 0 0.5 1 1.
5
3-CQA
4-CQA
5-CQA
4-FQA
5-FQA
3,4-diCQA
3,5-diCQA
4,5-diCQA
Correlation Coeficient
Green Li
g
ht roast Medium roast
Fig. 7. (a) Correlation coefficients between each of eight chlorogenic
acids and (a) cup quality or (b) presence of Rio-off-flavor. The right
side of the xaxis indicates (a) poor cup quality or (b) presence of Rio-
off-flavor.
A. Farah et al. / Food Chemistry 98 (2006) 373–380 379

beverage. A electrophoresis of proteins in agar–agar gel and its
interaction with chlorogenic acid. Turrialba, 25, 18–24.
Amorim, H. V., Basso, L. C., Crocomo, O. J., & Teixeira, A. A.
(1977). Polyamines in green and roasted coffee. Journal of
Agricultural and Food Chemistry, 25(4), 957–958.
Amorim, H. V., Cruz, A. R., St. Angelo, A. J., Dias, R. M., Mello, M.,
& Teixeira, A. A., et al. (1977). Biochemical, physical and
organoleptical changes during raw coffee quality deterioration.
Proceedings of 8th International Colloq Coffee. Abidjan: ASIC.
AOAC. (2000). Official methods of analysis of the association of official
analytical chemists (17th ed.). Gaithersburg, MD, USA: Associa-
tion of official Analytical chemists.
Azam, S., Hadi, N., Khan, N. U., & Hadi, S. M. (2003). Antioxidant
and prooxidant properties of caffeine, theobromine and xanthine.
Med Sci Monit, 9(9), BR330–BR335.
Barone, J. J., & Roberts, H. R. (1996). Caffeine consumption. Food and
Chemical Toxicology, 34(1), 119–129.
Bartholo, G. F., & Guimara
˜es, P. T. G. (1997). Cuidados na colheita e
no preparo do cafe
´.Informe Agropecua
´rio, 18(187), 33–42 (in
Portuguese).
Borges, M. L. A., Mendonc¸a, J. C. F., Franca, A. S., & Oliveira, L. S.
(2004). Perfis de trigonelina, acido 5-cafeoilquı
´nico e cafeı
´na em
cafe
´s de diferentes qualidades durante a torrac¸a
˜o. Revista Brasile-
ira de Armazenamento, Especial Cafe
´(8), 14–18 (in Portuguese).
Cantergiani, E., Brevar, H., Amado, R., Krebs, V., Feria-Morales, A.,
& Yeetizian, C. (1999). Proceedings of 18th International Colloq
Coffee. Helsinki.
Casal, S., Oliveira, M. B., & Ferreira, M. A. (2000). HPLC/diod-array
applied to the thermal degradation of trigonelline, nicotinic acid
and the caffeine in coffee. Food Chemistry, 68, 481–485.
Chagas, S. J. R. (1994). Caracterizac¸a
˜o quı
´mica e quantitativo de cafe
´s
de alguns municı
´pios de tre
ˆs regio
˜es produtoras de Minas Gerais,
MSc. Thesis. Universidade Federal de Lavras, Brazil. (in
Portuguese).
Clarke, R. J. (1987). Grading, storage, pre-treatments and blending. In
R. J. Clarke & R. Macrae (Eds.), Coffee volume 2: technology
(pp. 35–58). Amsterdam: Elsevier Applied Science.
Clifford, M. N. (1985). Chlorogenic acids. In R. J. Clarke & R. Macrae
(Eds.), Coffee volume 1: chemistry (pp. 153–202). London: Elsevier
Applied Science.
Clifford, M. N., & Wight, J. (1976). The measurements of feruloyl-
quinic acids and caffeoylquinic acids in coffee beans. Development
of the technique and its preliminary application to green coffee
beans. Journal of the Science of Food and Agriculture, 27, 73–84.
Cunha, R. D. C. (2005). Identificac¸a
˜o de compostos vola
´teis de cafe
´
sadio e com defeitos por cromatografia gasosa e ana
´lise estatı
´stica
multivariada. MSc. Thesis. Universidade Federal do Rio de
Janeiro, Instituto de Quı
´mica, Brazil. (in Portuguese).
De Maria, C. A. B., Trugo, L. C., Aquino Neto, F. R., Moreira, R. F.
A., & Alviano, C. S. (1996). Composition of green coffee water-
soluble fractions and identification of volatiles formed during
roasting. Food Chemistry, 55(3), 203–207.
De Maria, C. A. B., Trugo, L. C., Moreira, R. S. A., & Petracco, M.
(1995). Simultaneous determination of total chlorogenic acids,
trigonelline and caffeine in green coffee samples by high perfor-
mance gel filtration chromatography. Food Chemistry, 52(4),
447–449.
Farah, A. F., De Paulis, T., Trugo, L. C., & Martin, P. R. (2005).
Effect of roasting on the formation and degradation of chlorogenic
acid lactones in roasted coffee. Journal of Agricultural and Food
Chemistry, 53(5), 1505–1513.
Franca, A. S., Oliveira, L. S., Mendonc¸a, J. C. F., & Silva, X. A.
(2005). Physical and chemical attributes of defective crude and
roasted coffee beans. Food Chemistry, 90(1–2), 84–89.
Franca, A. S., Mendonc¸a, J. C. F., & Oliveira, S. D. (2005).
Composition of green and roasted coffees of different cup
qualities. Lebensmittel-Wissenschaft und -Technologie, 38, 709–
715.
Ky, C. L. (2001). Caffeine, trigonelline, chlorogenic acids and sucrose
diversity in wild Coffea arabica L. and C. canephora P. accessions.
Food Chemistry, 75, 223–230.
Lingle, T. R. (1993). The basics of cupping coffee. Long Beach:
Specialty Coffee Association of America.
Macrae, R. (1985). Nitrogenous compounds. In R. J. Clarke & R.
Macrae (Eds.), Coffee, vol 1: chemistry (pp. 115–152). Amsterdam:
Elsevier Applied Science.
Martin, M. J., Pablo, S., & Gonsalez, G. A. (1998). Discrimination
between arabica and Robusta green coffee varieties according to
their chemical composition. Talanta, 46, 1259–1264.
Mazzafera, P. (1991). Trigonelline in coffee. Phytochemistry, 30,
2309–2310.
Mazzafera, P. (1999). Chemical composition of defective coffee beans.
Food Chemistry, 64, 547–554.
Mazzafera, P., & Robinson, S. P. (2000). Characterization of
polyphenol oxidase in coffee. Phytochemistry, 55, 285–296.
Menezes, H. C. (1994). The relationship between the state of maturity
of raw coffee beans and the isomers of caffeoyilquinic acids. Food
Chemistry, 50, 293–296.
Oliveira, S. D., Franca, A. S., Glo
´ria, M. B. A., & Borges, M. L. A.
(2005). The effect of roasting on the presence of bioactive amines in
coffees of different qualities. Food Chemistry, 90(1–2), 287–
291.
Ohiokpehai, O., Brumen, G., & Clifford, M. N. (1982). The content of
chlorogenic acid in some peculiar green coffee beans and the
implications for beverage quality. IX Colloque Scientifique Inter-
national sur la Chemie du Cafe
´(pp. 177–185). Salvador, Brazil.
Ribeiro-Alves, M., Trugo, L. C., & Donangelo, C. M. (2003). Use of
oral contraceptives blunts the calciuric effect of caffeine in young
adult women. Journal of Nutrition, 133, 393–398.
Ruback, K. (1969). Beitrag zur Analytik der dihidroxyzimtsaureester
des Kaffes. Dissertation. Technische Universitat, Berlin.
Smith, A. W. (1985). Introduction. In R. J. Clarke & R. Macrae (Eds.),
Coffee volume 1: chemistry (pp. 1–41). London: Elsevier Applied
Science.
Spadone, J. C., Takeoka, G., & Liardon, R. (1990). Analytical
Investigation of Rio Off-Flavor in Green Coffee. Journal of
Agricultural and Food Chemistry, 38, 226–233.
Trugo, L. C. (1984). HPLC in Coffee Analysis – PhD Thesis.
University of Reading, England.
Trugo, L. C., & Macrae, R. A. (1984). Chlorogenic acid composition
of instant coffee. Analyst, 109, 263–266.
Trugo, L. C., Farah, A., & Cabral, L. (1995). Oligosacharides
distribution in Brazilian soybean cultivars. Food Chemistry, 52(4),
385–387.
380 A. Farah et al. / Food Chemistry 98 (2006) 373–380