ArticlePDF Available

A Novel Method for Quality Evaluation of Gardeniae fructus Praeparatus during Heat Processing Based on Sensory Characteristics and Chemical Compositions

Authors:

Abstract and Figures

The intrinsic chemical components and sensory characteristics of Gardeniae fructus Praeparatus (GFP) directly reflect its quality and subsequently, affect its clinical curative effect. However, there is little research on the correlation between the appearance traits and chemical compositions of GFP during heat processing. In this study, the major components of five typical processed decoction pieces of GFP were determined. With the deepening of processing, the contents of geniposidic acid and 5-HMF gradually increased, while the contents of deacetyl-asperulosidic acid methyl ester, gardenoside, and two pigments declined. Moreover, the electronic eye, electronic tongue, and electronic nose were applied to quantify GFP’s sensory properties. It was found that the chroma values showed a downward trend during the processing of GFP. The results of odor showed that ammonia, alkenes, hydrogen, and aromatic compounds were the material base for aroma characteristics. Complex bitterness in GF was more obvious than that in other GFP processed products. Furthermore, one mathematical model was established to evaluate the correlation between the sensory characteristics and chemical composition of GFP during five different stages. A cluster analysis and neural network analysis contributed to recognizing the processing stage of GFP. This study provided an alternative method for the exterior and interior correlation-based quality evaluation of herbs.
Content may be subject to copyright.
Citation: Zheng, Y.; Wang, Y.; Zhang,
Q.; Liu, W.; Li, K.; Xia, M.; Jia, Z.;
Zhang, C. A Novel Method for
Quality Evaluation of Gardeniae
fructus Praeparatus during Heat
Processing Based on Sensory
Characteristics and Chemical
Compositions. Molecules 2022,27,
3369. https://doi.org/10.3390/
molecules27113369
Academic Editor: Francesco Cacciola
Received: 24 April 2022
Accepted: 23 May 2022
Published: 24 May 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
molecules
Article
A Novel Method for Quality Evaluation of Gardeniae fructus
Praeparatus during Heat Processing Based on Sensory
Characteristics and Chemical Compositions
Yinghao Zheng 1,† , Yun Wang 1,† , Qing Zhang 1,2, Weihong Liu 3, Kai Li 2, Mengyu Xia 1,2, Zhe Jia 1
and Cun Zhang 1, 2,*
1Institute of Chinese Materia Medica, China Academy of Chinese Medical Sciences, Beijing 100700, China;
yinghao200888@163.com (Y.Z.); 15210014020@163.com (Y.W.); 17638706927@163.com (Q.Z.);
xiamengyu1215@163.com (M.X.); jiaz127@163.com (Z.J.)
2College of Pharmacy, Henan University of Chinese Medicine, Zhengzhou 450046, China; cpulikai@163.com
3
The First Affiliated Hospital of Henan University of Traditional Chinese Medicine, Zhengzhou 450046, China;
lwhc@163.com
*Correspondence: zhc95@163.com
These authors contributed equally to this work.
Abstract:
The intrinsic chemical components and sensory characteristics of Gardeniae fructus Praepara-
tus (GFP) directly reflect its quality and subsequently, affect its clinical curative effect. However,
there is little research on the correlation between the appearance traits and chemical compositions of
GFP during heat processing. In this study, the major components of five typical processed decoction
pieces of GFP were determined. With the deepening of processing, the contents of geniposidic acid
and 5-HMF gradually increased, while the contents of deacetyl-asperulosidic acid methyl ester, gar-
denoside, and two pigments declined. Moreover, the electronic eye, electronic tongue, and electronic
nose were applied to quantify GFP’s sensory properties. It was found that the chroma values showed
a downward trend during the processing of GFP. The results of odor showed that ammonia, alkenes,
hydrogen, and aromatic compounds were the material base for aroma characteristics. Complex
bitterness in GF was more obvious than that in other GFP processed products. Furthermore, one
mathematical model was established to evaluate the correlation between the sensory characteristics
and chemical composition of GFP during five different stages. A cluster analysis and neural network
analysis contributed to recognizing the processing stage of GFP. This study provided an alternative
method for the exterior and interior correlation-based quality evaluation of herbs.
Keywords:
Gardeniae fructus Praeparatus; quality assessment; electronic eye; electronic nose; elec-
tronic tongue; correlation analysis
1. Introduction
Chinese medicine processing, a preparation technology (used to make Chinese medicine
decoction pieces more suitable under the guidance of Chinese medicine theory) is an im-
portant step before the clinical application of traditional Chinese medicine (TCM) [
1
]. The
inherited chemical components and sensory characteristics including color, taste, and odor
could reflect the quality of TCM and then affect the clinical curative effect [
2
]. Meanwhile,
the empirical theory of quality assessment from the character identification of TCM would
be interpreted objectively based on the correlational analyses between appearance character-
istics and active ingredients [
3
]. A growing number of researchers focus on the correlation
between sensory features and chemical components, but research on single appearance
signature are relatively common; for example, main active ingredients and quantitative
color values were important bases for the quality evaluation of Gentianae macrophyllae radix
with different drying methods [
4
]. The quality of Corni Fructus was closely related to its
Molecules 2022,27, 3369. https://doi.org/10.3390/molecules27113369 https://www.mdpi.com/journal/molecules
Molecules 2022,27, 3369 2 of 16
color, so it can be identified and graded by exterior color [
5
]. The color values and chemical
components served as the criteria for quality evaluation and processing of the end-point
determination of rhubarb charcoal [
6
]. A good correlation between index composition and
powder color was observed during the microwave processing of Cibotium baronetz which
can be used to monitor processing [
7
]. Odor characteristics that exhibited a significant
correlation with chlorogenic acid provided a reliable basis for the quality control of Lonicera
japonica [
8
]. Quantifying taste information was regarded as a meaningful approach that
controlled the quality of herbal product sage lozenges [
9
]. To investigate the research
further, comprehensive morphological characteristics combining internal ingredients have
attracted attention. A novel quality evaluation method with HPLC, an electronic eye, and
an electronic nose was applied to assess the quality of magnolia bark [
10
]. It was possible
to distinguish Hawthorn and its processed products based on compounds, color, odor, and
flavor [
11
]. Overall, external-interior correlation analysis plays an important role in the
processing of TCM.
Gardeniae fructus (GF) is the dried and ripe fruit of Gardenia jasminoides ellis of the
Rubiaceae family; the fruitcan be processed into Gardeniae fructus Praeparatus (GFP) to
change the medicinal properties and meet the different clinical requirements of syndrome
differentiation in TCM. Currently, some efforts have been made to study the changes in the
processing of GFP. A strategy by UPLC-ESI-QTOF and a multivariate statistical analysis
found that some chemical changes existed between GF and GFP, showing downward
trends of geniposide, genipin-1-O-gentiobioside, 6a-hydroxygeniposide, jasminoside B,
crocin, and mannitol, while mussaenosidic acid increased [
12
]. The study with the aid
of a UPLC characteristic chromatogram and color value demonstrated that there were
significant differences in crocins as well as chroma values between GFP and other processed
products [
13
]. However, equally important qualities such as the odor and flavor of GFP are
ignored, let alone their relationship to intrinsic ingredients. Thus, it is necessary to establish
an objective and accurate quality evaluation system for GFP.
The previous study by our group has preliminarily identified five processed de-
coction pieces of GFP that reflected the different and typical processing stages, namely
pre-processed pieces (GF), middle-processed pieces (GFP-M), near-processed pieces (GFP-
N), the right processed product (GFP) and far-processed pieces (GFP-F) [
14
]. In this study,
a comprehensive evaluation of five processed decoction pieces of GFP was carried out by
UPLC, an electronic eye, electronic nose, and electronic tongue. More importantly, data
fusion from multi-source information was performed to explore the relationships between
intrinsic ingredients and apparent characteristics by correlation analysis. This study could
provide supervision for the processing of GFP based on cluster analysis and artificial neural
network analysis.
2. Results
2.1. Determination of Content in Five GFP Processed Decoction Pieces
We determined the contents of the main effective compositions of GF, GFP-M, GFP-N,
GFP, and GFP-F by UPLC. The chromatograms are shown in Figure 1. Seven iridoid glycosides
were identified at 254 nm, including geniposidic acid, shanzhiside, deacetyl-asperulosidic
acid methyl ester, scandosidemethyl ester, gardenoside, genipin-1-O-gentiobioside (G1), and
geniposide (G2). Detection of crocin-I (C-I) and crocin-II (C-II) at 440 nm showed that their
response values decreased with the deepening of the processing degree and even the C-II
peak was not obvious in the GFP-F sample. Meanwhile, 5-HMF was identified at 283 nm.
Notably, it could not be detected in GF but its response value gradually increased in the
deeper processed samples. The quantitative results are displayed in Table S1 and Figure 2.
As shown in Figure 2A, deacetyl-asperulosidic acid methyl ester in GFP-F significantly
decreased as compared to GF (p< 0.05); the contents of gardenoside in GFP-N, GFP, and
GFP-F were significantly different from that of GF (p< 0.05 or p< 0.01). The content of
shanzhiside, scandosidemethyl ester, gardenoside, and G2 were on a downward trend, but
there was no significant difference. The content of geniposidic acid increased gradually
Molecules 2022,27, 3369 3 of 16
during the heating process (p< 0.001). Overall, there was a falling-off in the contents of total
iridoids with no great difference. Pigments such as C-I and C-II decreased sharply with
the deepening of processing and they were significantly reduced in four GFP processed
decoction pieces (p< 0.001 compared to GF). Even C-II could not be detected in GFP-F
(Figure 2B). The contents of the total pigments were reduced significantly (p< 0.05). In
addition, as seen in Figure 2C, 5-HMF was formed during the heating processing and its
content gradually increased (p< 0.05 or p< 0.01).
Molecules 2022, 27, x FOR PEER REVIEW 3 of 16
gardenoside in GFP-N, GFP, and GFP-F were significantly different from that of GF (p <
0.05 or p < 0.01). The content of shanzhiside, scandosidemethyl ester, gardenoside, and G2
were on a downward trend, but there was no significant difference. The content of gen-
iposidic acid increased gradually during the heating process (p < 0.001). Overall, there was
a falling-off in the contents of total iridoids with no great difference. Pigments such as C-
I and C-II decreased sharply with the deepening of processing and they were significantly
reduced in four GFP processed decoction pieces (p < 0.001 compared to GF). Even C-II
could not be detected in GFP-F (Figure 2B). The contents of the total pigments were re-
duced significantly (p < 0.05). In addition, as seen in Figure 2C, 5-HMF was formed during
the heating processing and its content gradually increased (p < 0.05 or p < 0.01).
Figure 1. UPLC chromatogram of five GFP processed decoction pieces. 1—geniposidic acid; 2—
shanzhiside; 3—deacetyl-asperulosidic acid methyl ester; 4—gardenoside; 5—scandosidemethyl es-
ter; 6—G1; 7—G2; 8—C-I; 9—C-II; 10—5-HMF.
Figure 1.
UPLC chromatogram of five GFP processed decoction pieces. 1—geniposidic acid; 2—
shanzhiside; 3—deacetyl-asperulosidic acid methyl ester; 4—gardenoside; 5—scandosidemethyl
ester; 6—G1; 7—G2; 8—C-I; 9—C-II; 10—5-HMF.
Molecules 2022,27, 3369 4 of 16
Molecules 2022, 27, x FOR PEER REVIEW 4 of 16
Figure 2. The changed contents of 10 compounds of GF (pre-processed pieces), GFP-M (middle-
processed pieces), GFP-N (near-processed pieces), GFP (the right processed product) and GFP-F
(far-processed pieces). (A) Iridoids; (B) Pigments; (C) 5-HMF.
2.2. Color Analysis of Five GFP Processed Decoction Pieces
In the heating process of GFP the apparent color of the samples showed intuitive
changes that the color gradually deepened, changing from yellow orange to yellowish-
brown and finally turning tan (Figure 3A). Chroma values for each sample are shown in
Figure 3B. The L*value with the range of 0~100 indicated colors from black to white. It is
found that the L* of decoction pieces during processing went down with the time ex-
tended. In the case of a higher a* value, the color that neared red conversely neared green.
Compared with the GF in zero point, the a* value of samples with deeper processing de-
creased continuously. When the b* value became larger the color turned yellow, other-
wise, it turned blue. The b* value decreased significantly during the processing of GFP.
E*ab showed a downward trend indicating that the color of GFP processed products
changed from red and yellow to brown. As seen in Figure 3C, ΔL*, Δa*, and Δb* all de-
creased, among which Δb* largely declined with the largest slope. Meanwhile, ΔE*ab with
the ranges from 0 to 37.37 showed that the color difference of the samples during contin-
uous processing increased gradually compared with the initial decoction piece.
Figure 2.
The changed contents of 10 compounds of GF (pre-processed pieces), GFP-M (middle-
processed pieces), GFP-N (near-processed pieces), GFP (the right processed product) and GFP-F
(far-processed pieces). (A) Iridoids; (B) Pigments; (C) 5-HMF.
2.2. Color Analysis of Five GFP Processed Decoction Pieces
In the heating process of GFP the apparent color of the samples showed intuitive
changes that the color gradually deepened, changing from yellow orange to yellowish-
brown and finally turning tan (Figure 3A). Chroma values for each sample are shown in
Figure 3B. The L* value with the range of 0~100 indicated colors from black to white. It is
found that the L* of decoction pieces during processing went down with the time extended.
In the case of a higher a* value, the color that neared red conversely neared green. Compared
with the GF in zero point, the a* value of samples with deeper processing decreased
continuously. When the b* value became larger the color turned yellow, otherwise, it turned
blue. The b* value decreased significantly during the processing of GFP. E*ab showed a
downward trend indicating that the color of GFP processed products changed from red and
yellow to brown. As seen in Figure 3C,
L*,
a*, and
b* all decreased, among which
b*
largely declined with the largest slope. Meanwhile,
E*ab with the ranges from 0 to 37.37
showed that the color difference of the samples during continuous processing increased
gradually compared with the initial decoction piece.
Molecules 2022,27, 3369 5 of 16
Molecules 2022, 27, x FOR PEER REVIEW 5 of 16
Figure 3. The result of color analysis of GF (pre−processed pieces), GFP−M (middle−processed
pieces), GFP−N (near−processed pieces), GFP (the right processed product), and GFP−F (far−pro-
cessed pieces). (A) Intuitive colors. (B) The values in the heatmap represent the corresponding L*,
a*, b*, and E*ab values of each decoction piece. The L* value from large to small indicates luminance
from light to dark. The a* value from large to small indicates color from red to green. The b* value
from large to small indicates color from yellow to blue. The E*ab values from large to small mean
color change from GF to GFP-F. (C) The values of ΔL*, Δa*, and Δb* represent the changed chroma
values of L*, a*, and b* compared to the original value of GF, respectively. ΔE*ab represents color
difference values.
2.3. Olfactory Analysis of Five GFP Processed Decoction Pieces
The electronic nose was applied for a quantitative analysis of the aromatic features
of five GFP processed decoction pieces, and the signal responses of sensor arrays informed
that a stable signal output at 70 s was identified as the odor value (Figure S1). The odor
composition of five GFP processed decoction pieces was basically the same but differed
in the response intensity of each odor (Table S2 and Figure 4A). The response values of
W1C, W3C, and W5C from GF were lower than that of the other four processed GFP de-
coction pieces, while the response values of W5S, W6S, W1S, W1W, W2S, W2W, W3S from
GF were higher than other products. The scores scatter plot of the principal component
analysis (PCA) is displayed in Figure 4B. With the first component accounting for 87.80%
of the variables and the second principal component for 10.20%, the accumulative vari-
ance contribution of the two principal components accounted for 98.00%. Five GFP pro-
cessed decoction pieces were distributed in different regions without overlap. Among
them, GFP-M and GFP-N which were intermediate between GF and GFP could be re-
garded as a group as they were assigned to close positions. From this, the electronic nose
could identify different stages of the GFP processed process according to the odor varia-
tion, that is, GF (before processing), GFP-M and GFP-N (processing was not up to the
moderate point), GFP (processing was in a moderate point), and GFP-F (processing is far
away than a moderate point). According to the loadings analysis results, the parameters
of W1C, W3C, and W5C in the first principal component and W6S in the second principal
component had higher scores. These sensors could recognize ammonia, alkenes, hydro-
gen, and aromatic compounds, which might be the important substance bases resulting
in the odor difference of GFP processed decoction pieces (Figure 4C).
Figure 3.
The result of color analysis of GF (pre
processed pieces), GFP
M (middle
processed
pieces), GFP
N (near
processed pieces), GFP (the right processed product), and GFP
F
(far
processed pieces). (
A
) Intuitive colors. (
B
) The values in the heatmap represent the corre-
sponding L*,a*,b*, and E*ab values of each decoction piece. The L* value from large to small indicates
luminance from light to dark. The a* value from large to small indicates color from red to green. The
b* value from large to small indicates color from yellow to blue. The E*ab values from large to small
mean color change from GF to GFP-F. (
C
) The values of
L*,
a*, and
b* represent the changed
chroma values of L*,a*, and b* compared to the original value of GF, respectively.
E*ab represents
color difference values.
2.3. Olfactory Analysis of Five GFP Processed Decoction Pieces
The electronic nose was applied for a quantitative analysis of the aromatic features of
five GFP processed decoction pieces, and the signal responses of sensor arrays informed
that a stable signal output at 70 s was identified as the odor value (Figure S1). The odor
composition of five GFP processed decoction pieces was basically the same but differed
in the response intensity of each odor (Table S2 and Figure 4A). The response values of
W1C, W3C, and W5C from GF were lower than that of the other four processed GFP
decoction pieces, while the response values of W5S, W6S, W1S, W1W, W2S, W2W, W3S
from GF were higher than other products. The scores scatter plot of the principal component
analysis (PCA) is displayed in Figure 4B. With the first component accounting for 87.80%
of the variables and the second principal component for 10.20%, the accumulative variance
contribution of the two principal components accounted for 98.00%. Five GFP processed
decoction pieces were distributed in different regions without overlap. Among them,
GFP-M and GFP-N which were intermediate between GF and GFP could be regarded as
a group as they were assigned to close positions. From this, the electronic nose could
identify different stages of the GFP processed process according to the odor variation, that
is, GF (before processing), GFP-M and GFP-N (processing was not up to the moderate
point), GFP (processing was in a moderate point), and GFP-F (processing is far away than a
moderate point). According to the loadings analysis results, the parameters of W1C, W3C,
and W5C in the first principal component and W6S in the second principal component had
higher scores. These sensors could recognize ammonia, alkenes, hydrogen, and aromatic
compounds, which might be the important substance bases resulting in the odor difference
of GFP processed decoction pieces (Figure 4C).
Molecules 2022,27, 3369 6 of 16
Molecules 2022, 27, x FOR PEER REVIEW 6 of 16
Figure 4. The odor information of five GFP processed decoction pieces. (A) Radar map of response
value. (B) PCA analysis of electronic nose data. (C) Loadings analysis of electronic nose data.
2.4. Gustatory Analysis of Five GFP Processed Decoction Pieces
The taste–response value of the reference solution was bench-marked so the initial
values of sourness and saltiness were identified as −13 and −6, respectively, and the initial
values of other tastes were 0. The quantitative taste results of five GFP processed decoc-
tion pieces are shown in Table S3 and Figure 5A. The taste difference between GF and
another four GFP processed decoction pieces was significant and there was little differ-
ence in the four GFP products. As shown, five GFP processed decoction pieces did not
contain sourness and H-bitterness as their responses were lower than the initial value. The
sweetness could be recognized in all samples with almost the same response value and
the response of umami in GF was slightly higher than that of the other four types of GFP.
The richness of the umami aftertaste derived from GF but might be lost during processing.
The saltiness, astringency, and aftertaste-A were higher in GF, decreased successively in
GFP-M and GFP-N, then gradually increased in GFP and GFP-F. Significantly, the re-
sponse values of bitter taste, acidic bitter taste, and alkaline bitter taste increased with the
processing degree, indicating that a large number of bitter ingredients could be produced
during the processing of GFP. A PCA analysis showed that the first and second principal
components encompassed 71.10% and 27.40% of the contributing rate which covered the
main information of five GFP processed decoction pieces with a 98.8% cumulative contri-
bution of variance (Figure 5B). Each decoction piece was distributed in different regions
in the PCA scores scatter plot. It is worth noting that GFP-M was close to GFP-N in that
they had similar flavor features. Overall, the electronic tongue could figure out the differ-
ent processed products of GFP according to the smell messages. Concretely, GFP-M and
GFP-N were classified into one group while GF, GFP and GFP-F were the other three
groups. A loadings analysis was performed to determine the flavors that differentiated
the five GFP processed decoction pieces. As shown in Figure 5C, bitterness, aftertaste-B,
and B-bitterness2 contributed significantly to the first and second principal components;
saltiness had the highest loading parameters in the second principal component, followed
by sweetness, astringency, and aftertaste-A which contributed to the second component.
Thus, seven flavors including bitterness, aftertaste-B, B-bitterness2, saltiness, sweetness,
astringency, and aftertaste-A were regarded as primary indicators for distinguishing the
difference in flavor profile among GFP processed decoction pieces.
Figure 4.
The odor information of five GFP processed decoction pieces. (
A
) Radar map of response
value. (B) PCA analysis of electronic nose data. (C) Loadings analysis of electronic nose data.
2.4. Gustatory Analysis of Five GFP Processed Decoction Pieces
The taste–response value of the reference solution was bench-marked so the initial
values of sourness and saltiness were identified as
13 and
6, respectively, and the
initial values of other tastes were 0. The quantitative taste results of five GFP processed
decoction pieces are shown in Table S3 and Figure 5A. The taste difference between GF
and another four GFP processed decoction pieces was significant and there was little
difference in the four GFP products. As shown, five GFP processed decoction pieces did
not contain sourness and H-bitterness as their responses were lower than the initial value.
The sweetness could be recognized in all samples with almost the same response value and
the response of umami in GF was slightly higher than that of the other four types of GFP.
The richness of the umami aftertaste derived from GF but might be lost during processing.
The saltiness, astringency, and aftertaste-A were higher in GF, decreased successively
in GFP-M and GFP-N, then gradually increased in GFP and GFP-F. Significantly, the
response values of bitter taste, acidic bitter taste, and alkaline bitter taste increased with the
processing degree, indicating that a large number of bitter ingredients could be produced
during the processing of GFP. A PCA analysis showed that the first and second principal
components encompassed 71.10% and 27.40% of the contributing rate which covered
the main information of five GFP processed decoction pieces with a 98.8% cumulative
contribution of variance (Figure 5B). Each decoction piece was distributed in different
regions in the PCA scores scatter plot. It is worth noting that GFP-M was close to GFP-N in
that they had similar flavor features. Overall, the electronic tongue could figure out the
different processed products of GFP according to the smell messages. Concretely, GFP-M
and GFP-N were classified into one group while GF, GFP and GFP-F were the other three
groups. A loadings analysis was performed to determine the flavors that differentiated
the five GFP processed decoction pieces. As shown in Figure 5C, bitterness, aftertaste-B,
and B-bitterness2 contributed significantly to the first and second principal components;
saltiness had the highest loading parameters in the second principal component, followed
by sweetness, astringency, and aftertaste-A which contributed to the second component.
Thus, seven flavors including bitterness, aftertaste-B, B-bitterness2, saltiness, sweetness,
Molecules 2022,27, 3369 7 of 16
astringency, and aftertaste-A were regarded as primary indicators for distinguishing the
difference in flavor profile among GFP processed decoction pieces.
Molecules 2022, 27, x FOR PEER REVIEW 7 of 16
Figure 5. The taste information of five GFP processed decoction pieces. (A) Radar map of response
value. (B) PCA analysis of electronic tongue data. (C) Loadings analysis of electronic tongue data.
2.5. Integration Analysis of External Characteristics and Internal Components
2.5.1. Correlation Analysis
A correlation analysis was performed between the main active ingredients and the
quantified sensory indexes of five GFP processed decoction pieces. The first consideration
was to assess the normal distribution of the variables and homogeneity of variance. As a
result, geniposidic acid, b*, E*ab, sweetness, saltiness, and bitterness simultaneously satis-
fied the two requirements (Table S4) which were performed with a Pearson correlation
analysis. Other factors were analyzed by a Spearman correlation test. As shown in Table
S5, except for G1, the other nine components had different degrees of correlation with the
appearance indexes. The chroma values including L*, a*, b*, and E*ab were positively cor-
related with shanzhiside, deacetyl-asperulosidic acid methyl ester, gardenoside, scando-
sidemethyl ester, G2, C-I, and C-II, and were negatively correlated with geniposidic acid
and 5-HMF. As for the relationship between aroma and components, W5S, W6S, W1S,
W1W, W2S, W2W, and W3S had a significant positive association with shanzhiside,
deacetyl-asperulosidic acid methyl ester, gardenoside, scandosidemethyl ester, G2, C-I,
and C-II, while they were inversely correlated with geniposidic acid and 5-HMF. W3C
and W5C were associated with a positive value of geniposidic acid and 5-HMF. W3C was
negatively correlated with deacetyl-asperulosidic acid methyl ester, gardenoside, scando-
sidemethyl ester, G2, C-I, and C-II, and we also found that there was a significant negative
correlation between W5C and these six compounds and shanzhiside. In the correlation
analysis of flavor and composition, richness was significantly and negatively correlated
with geniposidic acid but positively correlated with G1 and C-I. Saltiness showed a strong
correlation with geniposidic acid. There was a strong correlation between aftertaste-A and
G1. Further, a strong relationship with p < 0.01 appeared between three kinds of bitter
taste (bitterness, aftertaste-B, and B-bitterness2) and nine main ingredients (geniposidic
acid, shanzhiside, deacetyl-asperulosidic acid methyl ester, gardenoside, scandoside-
methyl ester, G2, C-I, C-II, and 5-HMF). Specifically, the correlation coefficient between
B-bitterness2 and nine components was the highest, regardless of positive or negative cor-
relation, compared to bitterness and aftertaste-B. The relationships between the main ac-
tive ingredients that were significantly related to the indicators of color, odor, and taste
Figure 5.
The taste information of five GFP processed decoction pieces. (
A
) Radar map of response
value. (B) PCA analysis of electronic tongue data. (C) Loadings analysis of electronic tongue data.
2.5. Integration Analysis of External Characteristics and Internal Components
2.5.1. Correlation Analysis
A correlation analysis was performed between the main active ingredients and the
quantified sensory indexes of five GFP processed decoction pieces. The first consideration
was to assess the normal distribution of the variables and homogeneity of variance. As a re-
sult, geniposidic acid, b*,E*ab, sweetness, saltiness, and bitterness simultaneously satisfied
the two requirements (Table S4) which were performed with a Pearson correlation analysis.
Other factors were analyzed by a Spearman correlation test. As shown in Table S5, except
for G1, the other nine components had different degrees of correlation with the appearance
indexes. The chroma values including L*,a*,b*, and E*ab were positively correlated with
shanzhiside, deacetyl-asperulosidic acid methyl ester, gardenoside, scandosidemethyl ester,
G2, C-I, and C-II, and were negatively correlated with geniposidic acid and 5-HMF. As
for the relationship between aroma and components, W5S, W6S, W1S, W1W, W2S, W2W,
and W3S had a significant positive association with shanzhiside, deacetyl-asperulosidic
acid methyl ester, gardenoside, scandosidemethyl ester, G2, C-I, and C-II, while they were
inversely correlated with geniposidic acid and 5-HMF. W3C and W5C were associated
with a positive value of geniposidic acid and 5-HMF. W3C was negatively correlated with
deacetyl-asperulosidic acid methyl ester, gardenoside, scandosidemethyl ester, G2, C-I,
and C-II, and we also found that there was a significant negative correlation between
W5C and these six compounds and shanzhiside. In the correlation analysis of flavor and
composition, richness was significantly and negatively correlated with geniposidic acid
but positively correlated with G1 and C-I. Saltiness showed a strong correlation with
geniposidic acid. There was a strong correlation between aftertaste-A and G1. Further, a
strong relationship with p< 0.01 appeared between three kinds of bitter taste (bitterness,
aftertaste-B, and B-bitterness2) and nine main ingredients (geniposidic acid, shanzhiside,
Molecules 2022,27, 3369 8 of 16
deacetyl-asperulosidic acid methyl ester, gardenoside, scandosidemethyl ester, G2, C-I,
C-II, and 5-HMF). Specifically, the correlation coefficient between B-bitterness2 and nine
components was the highest, regardless of positive or negative correlation, compared to
bitterness and aftertaste-B. The relationships between the main active ingredients that were
significantly related to the indicators of color, odor, and taste are visualized in Figure 6. As
shown, the more connections to other nodes, the larger the node. The edges connecting
two nodes represent the correlation between them, and thicker edges mean stronger con-
nections. Meanwhile, solid orange lines indicate a positive correlation between the two
factors, while the dashed gray ones present a negative correlation. In the whole network,
four color indexes, nine sensor arrays, and six taste indexes were correlated with internal
substance components.
Molecules 2022, 27, x FOR PEER REVIEW 8 of 16
are visualized in Figure 6. As shown, the more connections to other nodes, the larger the
node. The edges connecting two nodes represent the correlation between them, and
thicker edges mean stronger connections. Meanwhile, solid orange lines indicate a posi-
tive correlation between the two factors, while the dashed gray ones present a negative
correlation. In the whole network, four color indexes, nine sensor arrays, and six taste
indexes were correlated with internal substance components.
Figure 6. The relationship network presents significant correlations between the components and
the indexes of external features. The a* indicates color from red to green, b* represents color from
yellow to blue, and E*ab means total chromatic aberration.
2.5.2. Cluster Heatmap Analysis
Data fusion of external characteristics and internal components was successfully ap-
plied to estimate the processing stage of GFP. The cluster heatmap was plotted for a com-
prehensive analysis of the indexes, including ingredients, color, flavor, and taste (Figure
7). It was obvious that GF was significantly different from the other four pieces and was
regarded as a separate group. GFP-M and GFP-N were combined into a group because of
the weaker difference between their internal and external features. Further, GFP came
close to GFP-F, so they were considered as one class. Overall, gradual changes occurred
when GF processed decoction pieces was produced in a more deeply processed GFP,
whether with internal components or external features.
Figure 6.
The relationship network presents significant correlations between the components and the
indexes of external features. The a* indicates color from red to green, b* represents color from yellow
to blue, and E*ab means total chromatic aberration.
2.5.2. Cluster Heatmap Analysis
Data fusion of external characteristics and internal components was successfully
applied to estimate the processing stage of GFP. The cluster heatmap was plotted for
a comprehensive analysis of the indexes, including ingredients, color, flavor, and taste
(Figure 7). It was obvious that GF was significantly different from the other four pieces
and was regarded as a separate group. GFP-M and GFP-N were combined into a group
because of the weaker difference between their internal and external features. Further,
GFP came close to GFP-F, so they were considered as one class. Overall, gradual changes
occurred when GF processed decoction pieces was produced in a more deeply processed
GFP, whether with internal components or external features.
Molecules 2022,27, 3369 9 of 16
Molecules 2022, 27, x FOR PEER REVIEW 9 of 16
Figure 7. Cluster heatmap for estimating the processing stage of GFP. The scale bar of the heatmap
represents Z-score normalized.
2.5.3. Artificial Neural Network Analysis
To classify GFP processed samples more accurately we constructed an artificial neu-
ral network as the machine-learning classifier. Thirty-five nodes representing sensory
characteristics and chemical compositions were fed into the input layers; eight units were
in the hidden layer and five GFP processed decoction pieces were used as outputs. The
different importance of the variables for the neural network model is displayed in Figure
S2. The network structure is presented in Figure 8 and focuses on the top ten most im-
portant predictors, such as W6S, bitterness, scandosidemethyl ester, sweetness, G2, gar-
denoside, a*, W5S, shanzhiside, and 5-HMF. The connection lines in the diagram are col-
ored according to an estimate of the synaptic weight, and greater line width corresponds
to greater importance. The neural network had a high prediction rate of 100% (Figure S3),
suggesting the validity of classification.
Figure 7.
Cluster heatmap for estimating the processing stage of GFP. The scale bar of the heatmap
represents Z-score normalized.
2.5.3. Artificial Neural Network Analysis
To classify GFP processed samples more accurately we constructed an artificial neural
network as the machine-learning classifier. Thirty-five nodes representing sensory charac-
teristics and chemical compositions were fed into the input layers; eight units were in the
hidden layer and five GFP processed decoction pieces were used as outputs. The different
importance of the variables for the neural network model is displayed in Figure S2. The
network structure is presented in Figure 8and focuses on the top ten most important predic-
tors, such as W6S, bitterness, scandosidemethyl ester, sweetness, G2, gardenoside, a*, W5S,
shanzhiside, and 5-HMF. The connection lines in the diagram are colored according to an
estimate of the synaptic weight, and greater line width corresponds to greater importance.
The neural network had a high prediction rate of 100% (Figure S3), suggesting the validity
of classification.
Molecules 2022,27, 3369 10 of 16
Figure 8. Artificial neural network for classification of five GFP processed decoction pieces.
3. Discussion
As intelligent bionic systems mature, the electronic eye, electronic nose, and electronic
tongue combined with analytical instruments are applied for the quality discrimination
of TCM processed products. For instance, Xu et al. [
15
] successfully applied intelligent
sensory technology and chromatographic analysis technology in the identification of Semen
arecae and its processed products. The researchers measured the color values of crude and
processed Leonuri fructus using the intelligent sensor technology of a colorimeter to assess
their quality [
16
]. The analysis of the sensory index of Chinese herbal decoction pieces
followed by a correlation analysis of the appearance and internal chemical composition of
the decoction pieces can better reveal the correlation between the appearance and internal
quality of the decoction pieces. It is an inevitable trend to establish a scientific and practical
quality evaluation system of TCM decoction pieces for reflecting the characteristics of
TCM. Here, an attempt in the present work has been made to explore the relationships
between the internal components and external characteristics of GFP by Pearson and
Spearman correlations. Satisfyingly, color characters had a good correlation with the main
ingredients, including geniposidic acid, shanzhiside, deacetyl-asperulosidic acid methyl
ester, gardenoside, scandosidemethyl ester, G2, C-I C-II, and 5-HMF. It is reported that
the higher the contents of total crocins and C-I, the redder GF [
17
]. With the deepening
of GFP processing, the contents of C-I and C-II dropped significantly due to the breaking
of bonds during the heating process. As a result, a* value indicated the gradual decline
in red property, identifying a positive correlation with C-I and C-II. As reported in the
literature, the Maillard reaction occurred during heat processing when the product 5-HMF
was formed [
18
]. The content of 5-HMF that was produced by the Maillard reaction rose
with increased heating time and temperature [
19
]. Thus, the content of 5-HMF continued to
grow during the heat processing of GFP. In addition, 5-HMF played a role in browning [
20
],
and this work found the intuitive phenomena associated with it. 5-HMF increased and
L* value decreased successively in deeper processed products, indicating that the 5-HMF
might be related to a color change from bright to dark. Nine electronic nose sensors (W2S
W5S, W3C, W6S, W5C, W1S, W1W, W2S, W2W, and W3S) had correlations with geniposidic
acid, shanzhiside, deacetyl-asperulosidic acid methyl ester, gardenoside, scandosidemethyl
Molecules 2022,27, 3369 11 of 16
ester, G2, C-I, C-II, and 5-HMF in varying degrees, indicating that odor characteristics
were closely related to the quality of GFP. A burnt smell could be produced during the
processing of GFP which may be related to the aromatic components that were identified
by these sensors. Five GFP processed decoction pieces had a complex taste of bitterness, an
acidic–bitter aftertaste, and alkalescent bitterness. The bitterness of GF was mainly derived
from geniposide [
21
]. However, deeper GFP processed products with a lower content of
geniposide showed stronger bitterness, suggesting that other bitter substances would be
produced during the processing of GFP. Most likely, the Maillard reaction promoted the
formation of various products with significant bitterness, such as 5-HMF [
22
]. Therefore, the
tastes of bitterness, acidic–bitter aftertaste, and alkalescent bitterness negatively correlated
to most iridoid glycosides, as well as showing high positive correlations with 5-HMF.
Together, these results of correlation analysis might facilitate our understanding of the
scientific connotation of the morphological identification of GFP. It is an important feature of
this study to establish a holistic quality evaluation system of GFP by combining quantitative
sensory attributes with an internal material basis.
Importantly, these obtained results enable judgment on the processing stage of GFP. It
was satisfying to determine whether electronic nose or electronic tongue technology could
be used for monitoring the processing of GFP, separately. Five GFP processed decoction
pieces were independent of each other, among which GFP-M was close to GFP-N as they
had a certain common flavor and taste. Further, the overall cluster analysis of data fusion,
including ingredients, color, aroma, and taste proposed the division of stages during the
processing of GFP. Not only were previous phases approaching GFP (namely GFP-M
and GFP-N) classified as one group, but GFP and GFP-F were also closely associated.
The artificial neural network was better at classifying complex data with efficiency and
accuracy [
23
] which showed an ideal classification performance on GFP processed decoction
pieces with a prediction rate of 100%. In aggregate, the integration strategy based on the
exterior and interior characteristics is expected to provide a reference for the overall quality
evaluation of other decoction pieces and monitoring dynamic processes.
4. Materials and Methods
4.1. Materials and Chemicals
The typical processed decoction pieces of GFP were identified in the previous re-
port [
14
]. Briefly, an electromagnetic stir-frying machine (CYJ 900, Beijing Hualin Ruikong
Technology Co., Ltd., Beijng, China) was used to prepare GFP processed decoction pieces
and GFP. The processed decoction pieces were sampled continuously every 0.5 or 1 min
for 15 min in the processing of GFP, among which the sample at 12.5 min was judged as
qualified GFP by an experienced pharmacist. As recorded in the Chinese Pharmacopoeia (2020
edition), the surface was brown or black, and the inner surface and seed surface of pericarp
was yellowish brown or tan [
24
]. Subsequently, all samples were clustered based on their
component contents and chroma values in order to screen five representative decoction
pieces spanning from initiation to far exceeding endpoint: GF (0 min), GFP-M (6 min),
GFP-N (10 min), GFP (12.5 min), and GFP-F (15 min). Geniposidic acid, shanzhiside, and
C-II were purchased from Chengdu Chroma-Biotechnology Co., Ltd. (Chengdu, China). C-I
and 5-HMF were bought from Chengdu Must Bio-Technology Co., Ltd. (Chengdu, China).
Scandosidemethyl ester was offered by Chengdu Lemetian Medicine Technology Co., Ltd.
(Chengdu, China). The purities of these commercially available reference substances were
above 98% as determined by HPLC analysis. Gardenoside, deacetyl-asperulosidic acid
methyl ester, G1, and G2 were prepared by our laboratory with purity over 98% by HPLC
analysis. Methanol, formic acid, absolute alcohol, and concentrated hydrochloric acid
were provided by Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Potassium
chloride (KCl), potassium hydroxide (KOH), and DL-tartaric acid were acquired from
Tianjin Kemiou Chemical Reagent Co., Ltd. (Tianjin, China). Silver chloride (AgCl) was
obtained from Shanghai Macklin Biochemical Co., Ltd. (Shanghai, China).
Molecules 2022,27, 3369 12 of 16
4.2. Preparation of Samples and Reference Substance
The sample solution was prepared following the reported protocol [
14
]. Briefly, GF,
GFP-M, GFP-N, GFP, and GFP-F were crushed into powder using a pulverizer for 2 min and
passed through a 40
µ
m mesh sieve, then 0.5 g of each sample was accurately weighed and
placed into a 50 mL conical flask. After adding 25 mL of 50% methanol, the samples were
extracted by ultrasound for 30 min (Kun Shan Ultrasonic Instruments Co., Ltd., Jiangsu,
China). The sample solution was cooled down to room temperature and weighed again,
and then the lost weight was restored to the original weight with 50% methanol. The
continuous filtrate was collected, followed by filtering with 0.22
µ
m of organic membranes.
Samples were prepared in triplicate for UPLC analysis.
The mixed reference substance was dissolved with 50% methanol which contained
5.75
µ
g/mL of geniposidic acid; 64.00
µ
g/mL of shanzhiside; 15.90
µ
g/mL of deacetyl-
asperulosidic acid methyl ester; 31.80
µ
g/mL of gardenoside; 8.80
µ
g/mL of scandosidemethyl
ester; 157.00
µ
g/mL of G1; 279.00
µ
g/mL of G2; 45.50
µ
g/mL of C-I; 10.10
µ
g/mL of C-II;
and 4.95 µg/mL of 5-HMF.
4.3. UPLC Analysis
Five typical processed decoction pieces of GFP and the mixed reference substances
were determined according to the reported method. [
14
] Briefly, UPLC (Nexera XR LC-
20AD XR, Shimadzu, Kyoto, Japan) was operated with a Waters Acquity UPLC BEH C
18
column (2.1 mm
×
100 mm, 1.7
µ
m) at the temperature of 35
C. The gradient elution with
0.5% formic acid (A) and methanol (B) was performed as follows: 0–6 min, 6% B; 6–11 min,
6–14% B; 11–19 min, 14–40% B; 19–24 min, 40~45% B; 24–29 min, 45–65% B; 29–30 min,
65–100% B; 30–36 min, 100% B. The flow rate was set as 1.0 mL/min, and the injection
volume was 1 µL. The samples were detected at multiple wavelengths of 254 nm, 283 nm,
and 440 nm.
4.4. Electronic Eye Analysis
Five kinds of powder from GFP processed decoction pieces were put into a dish,
making the surface flat. First, an electronic vision analyzer (VA400, Alpha M.O.S, Toulouse,
France) equipped with a CMOS lens needed to be calibrated with the color card, and then
it was used to gather color information from the samples. The chroma values were con-
verted into the parameters of CIELAB color space such as L* (Luminosity); a* (red/green);
and b* (blue/yellow). E*ab (total color value) was calculated in the following formula:
E*ab = (L*2+a*2+b*2)1/2
, while
E*ab = [(L*
L
0
*)
2
+ (a*
a
0
*)
2
+ (b*
b
0
*)
2
]
1/2
, among
which L0,a0, and b0were determined based on GF.
4.5. Electronic Nose Analysis
The aroma of five GFP processed decoction pieces was discriminated by the elec-
tronic nose (PEN3, Airsense Analytics, Schwerin, Germany). Five samples were put into a
headspace vial and sealed with plastic wrap, respectively. A needle injection probe was in-
serted into the sealed samples to acquire odor characteristics by direct headspace aspiration.
The self-cleaning time of the sensors was 100 s and the sample preparation time was 5 s. The
injection flow was 400 mL/min with a sampling time of 80 s. At 70 s, the sensor signal value
was stable, so the data of this time were considered as the output value. The electronic nose
had ten sensor arrays each responding to different sensitive substances. Specifically, W1C
was sensitive to aromatic compounds; W5S recognized nitrogen oxides; W3C presented
ammonia and aromatic compounds; W6S detected hydrogen; W5C responded to alkenes
and aromatic compounds; W1S identified methane; W1W reflected sulfides compounds;
W2S was related to alcohols and partially aromatic compounds; W2W indicated aromatic
compounds and organic sulfides; and W3S corresponded to alkenes. The measured data
were imported into Simca 14.1 software for a PCA and loadings analysis. The main idea
of PCA was to reduce the dimensionality of data, transferring the original n-dimensional
features to k-dimensional features (k < n) with little loss of whole-data information [
25
].
Molecules 2022,27, 3369 13 of 16
The PCA analysis was exploited in order to identify the differences or similarities between
samples in an unsupervised mode to judge the degree of GFP processing. The loadings
analysis would evaluate the influence of the variables on the components based on the
distance from the variable to the origin [
26
], and it was used to identify better indexes that
distinguished differences between the samples.
4.6. Electronic Tongue Analysis
An electronic tongue (TS-5000Z, Insent Company, Atsugi-Shi, Japan) containing multi-
ple taste sensors simulates a flavor-recognition system which can distinguish bitterness,
aftertaste-B, B-bitterness2, H-bitterness, astringency, aftertaste-A, umami, richness, sweet-
ness, sourness, and saltiness. Five sample solutions were prepared by adding 3 g of sample
to 100 mL of water, followed by an ultrasound for 10 min at 37
C. The cleaning liquid for
the positive electrode was prepared in a mixture with a total volume of 1000 mL consisting
of 7.46 g KCl, 0.56 g KOH, and 300 mL absolute alcohol. As for the cleaning liquid of
negative electrodes, mix 300 mL of anhydrous ethanol with 500 mL of deionized water,
then add 8.3 mL of concentrated hydrochloric acid, and finally replenish to 1000 mL. Next,
0.18 g DL-tartaric acid and 8.946 g KCl were dissolved with 4000 mL pure water to prepare
the reference fluid.
After washing in positive and negative electrode cleaning solutions for 90 s and the
reference solution for 120 s twice, the sensor returned to zero at the equilibrium position for
30 s. The sample was tested for 30 s, outputting the first taste value. The sensor was cleaned
and immediately inserted into a new reference solution to test the aftertaste for 30 s. Each
sample was tested four times among which the first cycle was removed and the next three
cycles were valid. The average data were retained as the test result. Similarly, the sweetness
was tested for five cycles, retaining the data of the middle three times. Furthermore, the
response values of the odor characteristics were analyzed by PCA and loadings analysis
for preliminary identification of the different processed samples and important variables.
4.7. Multi-Source Data Integration
The multi-source information from internal components and external characteristics
in terms of color, smell, and taste were integrated. These results were imported into
SPSS 26 software for correlation analysis. Data were examined for normal distribution
via the Shapiro-Wilk test and homogeneity of variance using the Levene test. Data that
simultaneously satisfied both the normality and homogeneity of variance were analyzed
using Pearson’s coefficient correlation, whereas variables were performed by Spearman’s
correlation. When p< 0.05, a significant correlation was affirmed. Moreover, the relationship
among correlational indicators (geniposidic acid, shanzhiside, deacetyl-asperulosidic acid
methyl ester, gardenoside, scandosidemethyl ester, G1, G2, C-I, C-II, 5-HMF, L*,a*,b*,E*ab,
W5S, W3C, W6S, W5C, W1S, W1W, W2S, W2W, and W3S, saltiness richness, aftertaste-A,
bitterness, aftertaste-B, and B-bitterness2) was visualized in network form by Cytoscape
3.7.0 software (http://www.cytoscape.org, accessed on 23 April 2022). A cluster heatmap
was performed on a free online platform (http://www.bioinformatics.com.cn, accessed
on 23 April 2022) to judge the processing degree of GFP. The values of all indicators were
normalized with a Z-score standardization method. By calculating each row of raw data
into the normalized Z-score one by one, the data of different dimensions were normalized to
the same dimension which helped to demonstrate the distribution of data. The calculation
formula was as follows: Z-score = (raw value
mean)/standard deviation [
27
]. Then, a
cluster analysis was conducted using Euclidean distance for two samples and a complete
linkage method for two clusters. A neural network model with multilayer perceptron
(MLP) was established by SPSS Modeler 1.0.0.430 software which consisted of an input
layer, a hidden layer with eight nodes, and an output layer. The original datasets including
five GFP processed decoction pieces in triplicate were split into a 70% training set and a
30% test set.
Molecules 2022,27, 3369 14 of 16
4.8. Statistical Analysis
The variance among groups was calculated using SPSS 26 software. The statistical
significance of GF and other processed GFP products was assessed by one-way analysis
of variance (ANOVA). An LSD test was used when the variance was homogeneous while
Dunnett’s T3 test was used when the variance was non-homogeneous. A p< 0.05 was
considered to be significant.
5. Conclusions
In summary, here, UPLC combined with intelligent sensory technologies including
the electronic eye, electronic nose and electronic tongue were applied to analyze five
representative GFP decoction pieces, followed by an integrated analysis of data from these
instruments which was considered as a novel strategy for the quality evaluation of GFP and
its processed products. Based on this, ten active ingredients were key indexes to evaluate
the quality of GFP processed products and had definite correlations with digital chroma
values. Aromatic compounds might be relevant to the burnt odor that occurs during heat
processing. Bitterness, acidic–bitter aftertaste, and alkalescent bitterness were defining
flavor features of GFP. Further, various processed GFP decoction pieces could be better
identified by cluster analysis and artificial neural network analysis. Hence, the proposed
method based on sensory characteristics and chemical compositions owns the advantages
of objective evaluation, fast and simple operation, and is expected to succeed in assessing
the quality and monitoring procedures of GFP during processing.
Supplementary Materials:
The following supporting information can be downloaded at: https:
//www.mdpi.com/article/10.3390/molecules27113369/s1, Table S1: The contents of 10 compounds
among 5 processed products of GFP. Table S2: The response values of sensors of the electronic nose
among five processed products of GFP. Table S3: The response values of sensors of the electronic
tongue among five processed products of GFP. Table S4: Results of the normal distribution test of
variables and homogeneity of variance. Table S5: Correlation between the components and the
sensory characteristics (color, odor, and taste). Figure S1: Response curve of sensors of an electronic
nose. Figure S2: Variable importance of neural network model. Figure S3: Overall percentage
correction of the neural network model.
Author Contributions:
Conceptualization, Y.W. and C.Z.; Methodology, Y.Z.; Software, Q.Z.; Val-
idation, M.X.; Formal Analysis, W.L.; Investigation, K.L.; Resources, Y.Z.; Data Curation, Y.Z.;
Writing—Original Draft Preparation, Y.Z. and Y.W.; Writing—Review & Editing, Z.J. and Y.W.; Visual-
ization, Y.Z.; Supervision, Y.W. and C.Z.; Project Administration, Y.W. and C.Z.; Funding Acquisition,
C.Z. All authors have read and agreed to the published version of the manuscript.
Funding:
This work was supported by the scientific and technological innovation project of the China
Academy of Chinese Medical Sciences (No. CI2021A04204); the National Natural Science Foundation
of China (No. 81873010, 82173979, 81703708); and the Fundamental Research Funds for the Central
public welfare research institutes of the China Academy of Chinese Medical Sciences (No. zz13-019,
zz13-YQ-050).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement:
The data presented in this study are available in Supplementary Material.
Conflicts of Interest: The authors declare no conflict of interest.
Sample Availability: Samples of the compounds are available from the authors.
Molecules 2022,27, 3369 15 of 16
References
1.
Zhao, Z.; Liang, Z.; Chan, K.; Lu, G.; Lai Mei Lee, E.; Chen, H.; Li, L. A unique issue in the standardization of Chinese materia
medica: Processing. Planta Med. 2010,76, 1975–1986. [CrossRef] [PubMed]
2.
Miao, X.; Cui, Q.; Wu, H.; Qiao, Y.; Zheng, Y.; Wu, Z. New sensor technologies in quality evaluation of Chinese materia medica:
2010–2015. Acta Pharm. Sin. B 2017,7, 137–145. [CrossRef] [PubMed]
3.
Li, J.; Wei, X.; Wan, G.; Yang, X.; Shi, J. Historical evolution and modern research progress of quality evaluation based on character
identification of traditional Chinese medicinal materials. Chin. J. Exp. Tradit. Med. Formulae 2021,27, 189–196. [CrossRef]
4.
Chen, Q.; Shi, Z.; Wang, Y.; Zhang, R.; Hu, H. Study on effects of sweating treatment and different drying methods on the quality
of Gentianae macrophyllae Radix.Sci. Rep. 2021,11, 9717. [CrossRef]
5.
Fei, C.; Ren, C.; Wang, Y.; Li, L.; Li, W.; Yin, F.; Lu, T.; Yin, W. Identification of the raw and processed Crataegi Fructus based on the
electronic nose coupled with chemometric methods. Sci. Rep. 2021,11, 1849. [CrossRef]
6.
Yang, L.; Gong, Y.; Xu, M.; Yang, L.; Chen, J.; Dong, L. Research on relationship between color and composition changes in
processing of rhubarb charcoal based on “external and internal correlation”. Chin. Tradit. Herb. Drugs
2020
,51, 5705–5713.
[CrossRef]
7.
Liang, Z.; Huang, G.; Cui, T.; Zhang, Z.; Sun, D. Analysis on the correlation of index composition and powder color during
microwave processing of Cibotium baronetz.Chin. Tradit. Patent Med. 2020,42, 1548–1552. [CrossRef]
8.
Xiong, Y.; Xiao, X.; Yang, X.; Yan, D.; Zhang, C.; Zou, H.; Lin, H.; Peng, L.; Xiao, X.; Yan, Y. Quality control of Lonicera japonica
stored for different months by electronic nose. J. Pharmaceut. Biomed. 2014,91, 68–72. [CrossRef]
9.
Eckert, C.; Lutz, C.; Breitkreutz, J.; Woertz, K. Quality control of oral herbal products by an electronic tongue—Case study on
sage lozenges. Sens. Actuat. B-Chem. 2011,156, 204–212. [CrossRef]
10.
Li, J.; Shao, Y.; Yao, Y.; Yu, Y.; Cao, G.; Zou, H.; Yan, Y. A novel quality evaluation method for magnolia bark using electronic nose
and colorimeter data with multiple statistical algorithms. J. Tradit. Chin. Med. Sci. 2020,7, 221–227. [CrossRef]
11.
Xie, D.; Peng, W.; Chen, J.; Li, L.; Zhao, C.-B.; Yang, S.; Xu, M.; Wu, C.; Ai, L. A novel method for the discrimination of Hawthorn
and its processed products using an intelligent sensory system and artificial neural networks. Food Sci. Biotechnol.
2016
,25,
1545–1550. [CrossRef] [PubMed]
12.
Li, H.; Yang, H.; Xue, X.; Liu, X.; Tian, F.; Poh, Y.; Cai, H.; Lee, Y.H.; Yu, H.; Ong, S.P.; et al. A metabolomics approach to study the
dual modulation by characterization of chemical alteration during processing of Gardeniae Fructus using UPLC-ESI-QTOF. Anal.
Methods 2016,8, 3629–3635. [CrossRef]
13.
Liu, L.; Gan, L.; Zeng, Z.; Li, G.; Li, Y. Optimization of processing technology and establishment of quality standard of Gardenia
jasminoides.J. Chin. Med. Mater. 2019,42, 1527–1532. [CrossRef]
14.
Xia, M.; Wang, Y.; Zheng, Y.; Song, Y.; Liu, T.; Zhang, C. Comparising the quality changes of different fried forms of Gardeniae
Fructus Praeparatus during processing based on color-component correlation analysis. China J. Chin. Mater. Med.
2021
,46, 10.
[CrossRef]
15.
Xu, M.; Yang, S.; Peng, W.; Liu, Y.; Xie, D.; Li, X.; Wu, C. A novel method for the discrimination of Semen Arecae and its processed
products by using computer vision, electronic nose, and electronic tongue. Evid.-Based Complentary Altern. Med.
2015
,2015,
753942. [CrossRef] [PubMed]
16.
Liu, X.; Yan, D.; Deng, X.; Zhao, B.; Xue, X.; Wang, S.; Zhang, Y.; Meng, J. Quality assessment of crude and processed Leonuri
Fructus by chemical and color analysis combined with chemometric method. Chin. Herb. Med. 2018,10, 388–395. [CrossRef]
17.
Fu, X.; Yang, C.; Wu, Z.; Liu, J.; Wu, J.; Pei, J.; Huang, X.; Wang, F. Correlation between color and content of eight components of
Gardeniae Fructus at different harvest time. China J. Chin. Mater. Med. 2020,45, 3191–3202. [CrossRef]
18. Huang, Y.; Li, C.; Hu, H.; Wang, Y.; Shen, M.; Nie, S.; Chen, J.; Zeng, M.; Xie, M. Simultaneous determination of acrylamide and
5-hydroxymethylfurfural in heat-processed foods employing enhanced matrix removal-lipid as a new dispersive solid-phase
extraction sorbent followed by liquid chromatography-tandem mass spectrometry. J. Agr. Food Chem.
2019
,67, 5017–5025.
[CrossRef]
19.
Woo, K.S.; Hwang, I.G.; Kim, H.Y.; Jang, K.I.; Lee, J.; Kang, T.S.; Jeong, H.S. Thermal degradation characteristics and antioxidant
activity of fructose solution with heating temperature and time. J. Med. Food. 2011,14, 167. [CrossRef]
20.
Hsieh, H.J.; Lin, J.A.; Chen, K.T.; Cheng, K.C.; Hsieh, C.W. Thermal treatment enhances the
α
-glucosidase inhibitory activity
of bitter melon (Momordica charantia) by increasing the free form of phenolic compounds and the contents of Maillard reaction
products. J. Food Sci. 2021,86, 3109–3121. [CrossRef]
21.
Guan, D.; Fang, Z.; Pan, Z.; Lu, W.; Liu, X.; Liang, C.; Wu, Z.; Gou, W. Establishing a standard verification method for Five Flavors
of Chinese medicine. Chin. J. Basic Med. Tradit. Chin. Med. 2011,17, 1333–1335.
22.
Li, H.; Zhang, W.; Tang, X.; Wu, C.; Yu, S.; Zhao, Z. Identification of bitter-taste compounds in class-III caramel colours. Flavour
Fragr. J. 2021,36, 404–411. [CrossRef]
23.
Summerson, V.; Gonzalez Viejo, C.; Szeto, C.; Wilkinson, K.L.; Torrico, D.D.; Pang, A.; De Bei, R.; Fuentes, S. Classification of
smoke contaminated Cabernet Sauvignon berries and leaves based on chemical fingerprinting and machine learning algorithms.
Sensors 2020,20, 5099. [CrossRef] [PubMed]
24.
Commission, C.P. Pharmacopeia of the People’s Republic of China, Volume 1; China Medical Science and Technology Press: Beijing,
China, 2020; p. 260.
Molecules 2022,27, 3369 16 of 16
25.
Jiang, H.; Huang, Y.; You, Z. SAEROF: An ensemble approach for large-scale drug-disease association prediction by incorporating
rotation forest and sparse autoencoder deep neural network. Sci. Rep. 2020,10, 4972. [CrossRef]
26.
Yeo, M.T.Y.; Yeo, P.L.Q.; Bi, X.; Henry, C.J. Energy density and nutrient contents of selective Chinese new year snacks. Foods
2020
,
9, 1137. [CrossRef]
27.
Lee, H.H.; Lee, A.J.; Park, W.S.; Lee, J.; Park, J.; Park, B.; Joung, J.Y.; Lee, K.H.; Hong, D.; Kim, S.H. Epithelial splicing regulatory
protein (ESPR1) expression in an unfavorable prognostic factor in prostate cancer patients. Front. Oncol.
2020
,10, 556650.
[CrossRef]
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
To provide basis for improving the quality of Gentianae macrophyllae Radix and optimising the field processing method. Gentianae macrophyllae Radix samples were collected from Long County. Five main active iridoids were determined by using HPLC. The HPLC fingerprints were measured and analysed by the traditional Chinese medicine chromatographic fingerprint similarity evaluation system. Water, total ash, acid insoluble ash and ethanol extract contents were determined in accordance with Chinese Pharmacopoeia (2015). Colour of the surface and powdered samples were determined by using a colorimeter. Data were analysed by SPSS11.0. Notable differences were found among samples from different sweating and drying processes, in spite of the relatively consistent overall chemical characteristics. Sweating and drying methods influenced the contents of some active ingredients and colour very significantly; moreover, they also showed significant effects on the water, ash, and ethanol extract contents. The necessity of sweating treatment in the field processing of Gentianae macrophyllae Radix may be insufficient. Drying in the shade and oven drying are more profitable for preserving the active constituents. Gentianae macrophyllae Radix should be dried directly without sweating, and drying in the shade and oven drying should be adopted preferentially to improve the quality.
Article
Full-text available
Crataegi Fructus (CF) is widely used as a medicinal and edible material around the world. Currently, different types of processed CF products are commonly found in the market. Quality evaluation of them mainly relies on chemical content determination, which is time and money consuming. To rapidly and nondestructively discriminate different types of processed CF products, an electronic nose coupled with chemometrics was developed. The odour detection method of CF was first established by single-factor investigation. Then, the sensor array was optimised by a stepwise discriminant analysis (SDA) and analysis of variance (ANOVA). Based on the best-optimised sensor array, the digital and mode standard were established, realizing the odour quality control of samples. Meanwhile, mathematical prediction models including the discriminant formula and back-propagation neural network (BPNN) model exhibited good evaluation with a high accuracy rate. These results suggest that the developed electronic nose system could be an alternative way for evaluating the odour of different types of processed CF products.
Article
Full-text available
Background: To evaluate the role of epithelial splicing regulatory protein 1 (ESRP1) expression in survival prognoses and disease progression for prostate cancer (PC) using The Cancer Genome Atlas (TCGA) dataset and to validate it using patients' prostatectomy specimens. Methods: A preliminary investigation into the clinical significance of ESRP1 in PC was conducted using TCGA PC PRAD dataset and then using immunohistochemistry in 514 PC patients' tissue microarrays of radical prostatectomy specimens. The interpretation of immunohistochemistry was done using its intensity (high vs. low) or the semi-quantitative expression value (H-score, 0-300). The prognostic significance of ESRP1 expression was analyzed for biochemical recurrence (BCR), recurrence-free survival (RFS), overall survival (OS) and cancer-specific survival (CSS) using the Cox proportional-hazards model (p < 0.05). Results: In the publicly available prostate adenocarcinoma dataset, ESRP1 expression was significantly higher in the tumor samples compared to the normal samples (p < 0.001). Survival analysis showed that the tumor samples in the ESRP1-high group had significantly worse BCR-free survival and RFS compared to the ESRP1-low group (p < 0.05), whereas OS was not (p=0.08). These results were largely consistent with the 514 patients' clinical data during a median 91.2 months of follow-up. After adjusting for significant prognostic clinicopathological factors, the multivariable models showed that the ESRP1 was a significantly risk factor for CSS (Hazard ratio 3.37, p = 0.034) and for BCR (HR 1.34, p=0.049) without any significance for OS (p=0.464). Conclusions: The higher ESRP1 expression appeared increased risk of disease progression and cancer-specific death in PC.
Article
Full-text available
Wildfires are an increasing problem worldwide, with their number and intensity predicted to rise due to climate change. When fires occur close to vineyards, this can result in grapevine smoke contamination, and subsequently, the development of smoke taint in wine. Currently, there are no in-field detection systems that growers can use to assess whether their grapevines have been contaminated by smoke. This study evaluated the use of near-infrared (NIR) spectroscopy as a chemical fingerprinting tool, coupled with machine learning, to create a rapid, non-destructive in-field detection system for assessing grapevine smoke contamination. Two artificial neural network models were developed using grapevine leaf spectra (Model 1) and grape spectra (Model 2) as inputs, and smoke treatments as targets. Both models displayed high overall accuracies in classifying the spectral readings according to the smoking treatments (Model 1: 98.00%; Model 2: 97.40%). Ultraviolet to visible spectroscopy was also used to assess the physiological performance and senescence of leaves, and the degree of ripening and anthocyanin content of grapes. The results showed that chemical fingerprinting and machine learning might offer a rapid, in-field detection system for grapevine smoke contamination that will enable growers to make timely decisions following a bushfire event, e.g., avoiding harvest of heavily contaminated grapes for winemaking or assisting with a sample collection of grapes for chemical analysis of smoke taint markers.
Article
Full-text available
Background: In this study, the energy density and nutrient contents of thirty Chinese New Year (CNY) snacks commonly consumed in the Asian region during the CNY festive season were measured. Methods: Calorie Answer™, Gas Chromatography-Mass Spectrometry (GC-MS) and Inductively coupled plasma-mass spectrometry (ICP-MS) were the main methods applied in this study. Results: All snacks showed high energy density (kJ/100 g) and the nutrient content, including macro-and micronutrients of these snacks were remarkably different. The most abundant minerals in these snacks include sodium, magnesium, potassium, and calcium. Palmitic (C16:0), oleic (C18:1) and linoleic (C18:2) acids were the main fatty acids and trans-fat was only identified in Butter cookies. Conclusions: This study provides a large database on macro- and micronutrient contents in CNY snacks consumed in the Asian region that has not been previously reported. Our results indicate that the snacks were generally energy-dense and nutrient-poor. This study provides necessary information to enable the reformulation of snacks with lower saturated fat and sodium content. It is also a source of information for consumers to select healthier snacks.
Article
Full-text available
Background Magnolia bark (Magnolia Officinalis REHD. & WILS. and Magnolia officinalis REHD. & WILS. VAR. biloba REHD. & WILS, Hou Po in Chinese), is widely applied in clinical prescriptions and Chinese patent medicines. Origin place is a crucial factor affecting the quality of Hou Po, and chemical composition is an important index for evaluating its quality, which is closely related to its clinical efficacy. This study aims to develop a novel method for rapidly, accurately and comprehensively identifying the origin places of Hou Po and predicting the contents of its important chemical components. Methods High performance liquid chromatography was used to analyze the contents of magnolol and honokiol and ultra-performance liquid chromatography the contents of magnocurarine and magnoflorine. The cold soak method was used to determine the contents of water-soluble extracts. The E-nose and colorimeter were used to determine the odor and color characteristics, respectively, of the collected Hou Po samples. Results Using several statistical algorithms, different discriminant models based on the E-nose and colorimeter data were established to distinguish the habitat of Hou-Po and predict the chemical components of honokiol, magnolol, magnocurarine, magnoflorine and water-soluble extracts. The results showed that the Random Forest classifier combined with the ten-fold cross-validation method provided the highest classification accuracy for origin place, accounting for 99.53% among these models. The correlation coefficients between predicted and experimental values of the five chemical components were all higher than 0.96. Conclusion This study has indicated that the electronic nose and colorimeter are promising methods for evaluating the quality of Chinese herbal medicines both qualitatively and quantitatively.
Article
Inhibition of α-glucosidase can slow carbohydrate metabolism, which is known as an effective strategy for diabetes treatment. The aim of this study is to evaluate the effect of thermal treatment (50, 60, and 70℃) for 15 days on the α-glucosidase inhibitory activity of bitter melon. The results show that the bitter melon heated at 70℃ for 12 days had the best α-glucosidase inhibitory effect. However, the amount of free polyphenols, 5-hydroxymethyl-2-furfural (5-HMF), and the browning degree of bitter melon generally increased with the time (15 days) and temperature of the thermal treatment, which is positively related to their antioxidant and α-glucosidase inhibitory activities. In conclusion, aged bitter melon shows great α-glucosidase inhibitory activity, which may be related to the increased free form of the involved phenolic compounds and Maillard reaction products. This suggests that thermal processing may be a good way to enhance the application of bitter melon for diabetes treatment. Practical Application The thermal processing of bitter melon provides an application for diabetes treatment. This study demonstrated that heat-treated bitter melon can lower the blood glucose level; therefore, it can be used as a potential anti-hyperglycemic and functional food.
Article
The chromatic values of the broken-fried and single-fried Gardeniae Fructus Praeparatus(GFP) were measured by the color analyzer to analyze the color variation rule, and the contents of 10 main components were determined by ultra-high performance liquid chromatography(UPLC). The multivariate statistical analysis, Pearson correlation analysis, and discriminant analysis were conducted to investigate the color and components of GFP samples. The experimental results revealed that L~*, a~*, b~*, and E~*ab decreased continuously during processing, and the color of samples gradually deepened. The trend and range of chromatic values during broken-frying and single-frying processes were basically identical. Gardenoside, crocin-Ⅰ(C-Ⅰ), and crocin-Ⅱ(C-Ⅱ) showed an obviously downward trend, while the contents of geniposidic acid and 5-hydroxymethylfurfural(5-HMF) increased significantly. Shanzhiside, deacetyl-asperulosidic acid methyl ester, and geniposide(G2) showed a downward trend. Scandoside methyl ester rose first and fell later. Genipin-1-O-gentiobioside(G1) went through a decrease-increase-decrease trend. The change trends of component contents during broken-frying and single-frying processes were generally consistent, but the change range was different. Among all the components, scandoside methyl ester and G1 showed obvious change. Because of different stir-frying time, the change rate of each component content in the process of broken-frying was higher than that in single-frying process. Additionally, geniposidic acid, gardenoside, scandoside methyl ester, C-Ⅰ, C-Ⅱ, and 5-HMF exhibited a higher correlation with apparent color. On the basis of above findings, the discriminant function of two frying processes was established, which could be applied to the discrimination of broken-fried and single-fried samples. This study analyzed the dynamic quality change rule of GFP during broken-frying and single-frying processes based on color-component correlation analysis, and found the two methods showed consistent change trend, yet with slight difference in the quality of samples. This study can provide data support for the processing of GFP.
Article
Caramel colours become one of the most widely used colorings in various foods and beverages due to their total miscibility with water and great coloring properties by containing colloidal aggregates. Bitterness in caramel colours can negatively influence their acceptability. The sequential solvent extraction (SSE) and semi-preparative liquid chromatography (SP-LC) combined with the taste dilution analysis were applied to identify the most intense bitter-tasting compounds in class-III caramel colours. Four main compounds with the highest bitterness intensity were isolated and structurally elucidated based on high resolution-electrospray ionization-mass spectrum (HR-ESI- MS), nuclear magnetic resonance (NMR) spectroscopy and gas chromatography-mass spectrum (GC-MS): 2,3-dihydro-3,5-dihydroxy-6-methyl-4(H)-pyran-4-one (DDMP), 5-(hydroxymethyl)furfural (HMF), furfural and 5-methylfurfural (MF). The threshold concentrations of DDMP, HMF, furfural and MF were 14.3, 30.0, 4.4 and 17.5μmol L−1, respectively. Furthermore, two volatile bitter compounds were identified: [5-(di-methoxymethyl)furan-2-yl]methanol (DMFM) and 2,3,4,4-tetramethyl-pentane-1,3-diol (TMP). Based on the composition of the caramel colours and the structures of these compounds, Maillard reaction was supported to generate these bitter compounds in caramel colour-III.
Article
This research is to establish an HPLC method for determination of geniposidic acid, genipin-1-β-D-gentiobioside, geniposide, p-trans-coumaroylgenipin gentiobioside, chlorogenic acid, crocin-Ⅰ, crocin-Ⅱ and crocin-Ⅲ in Gardeniae Fructus at different harvest time. The detection wavelength was 238, 320 and 440 nm. Principal component analysis(PCA), correlation analysis, regression analysis and partial least squares(PLS) analysis were used to explore the relationship of color and content of eight components in Gardeniae Fructus. The result showed that the trend of the eight components in Gardeniae Fructus at harvest time in different three years was varied similarly. According to the variation of eight components at different harvest time, the mature and immaturate Gardeniae Fructus were discriminated. The content of crocin-Ⅰwas correlated positively with a~* of color significance. The redder color of Gardeniae Fructus showed the higher value of a~* and content of crocin-Ⅰ, indicating the better quality of Gardeniae Fructus. This method provided reference for justifying the color and quality of Gardeniae Fructus and scientific evidence for "assessing quality by distinguishing color".