LWT - Food Science and Technology 161 (2022) 113317

Contents lists available at ScienceDirect

LWT
journal homepage: www.elsevier.com/locate/lwt

Identification of changes in the volatile compounds of robusta coffee beans

during drying based on HS-SPME/GC-MS and E-nose analyses with the aid
of chemometrics
Ke Zhang a, b, Jinhuan Cheng c, 1, Qidi Hong a, Wenjiang Dong a, d, *, Xiaoai Chen a, d,
Guiping Wu a, d, Zhenzhen Zhang b, **
a
Spice and Beverage Research Institute, Chinese Academy of Tropical Agricultural Sciences, Wanning, Hainan, 571533, China
b
College of Food Science and Pharmacy, Xinjiang Agricultural University, Urumqi, Xinjiang, 830052, China
c
Tropical and Subtropical Economic Crops Research Institute, Yunnan Academy of Agricultural Sciences, Baoshan, Yunnan, 678000, China
d
Key Laboratory of Processing Suitability and Quality Control of the Special Tropical Crops of Hainan Province, Wanning, Hainan, 571533, China

A R T I C L E I N F O A B S T R A C T

Keywords: In this study, headspace solid-phase microextraction/gas chromatography-mass spectrometry (HS-SPME/GC-MS)

Robusta coffee and electronic-nose (E-nose) analyses were performed to examine the comprehensive flavor profiles of robusta
HS-SPME/GC-MS coffee during heat air drying. A total of 62 volatile compounds were identified by HS-SPME/GC-MS, the level of
Drying process
acids, aldehydes, alcohols and esters decreased during drying process, however, the relative content of hydro­
Volatile compounds
Chemometrics
carbons and acids increased. Principal component analysis of the E-nose results clearly differentiated the samples
related to drying time, and it was found that samples of initial stages (0 h, 3 h, and 6 h) had strong positive with
different sensors. Partial least squares-discriminant analysis (PLS-DA) indicated that alcohols, ketones, and esters
were the main contributing compounds to the flavour of coffee beans. The VIP score of PLS-DA larger than one
illustrated that T30/1, T70/2, PA/2, and P30/2 were positive with ketones, sulfur compounds, and esters. These
results could provide theoretical evidence about the change rule in volatile compounds of coffee beans during
drying process, and could be applied to other thermally sensitive tropical agricultural products.

1. Introduction can involve a series of steps, such as de-pulping, drying, storaging, and
roasting, which can influence the quality and value of the final product.
Coffee is one of the most popular beverage owing to its unique Fresh coffee fruits are perishable and deteriorative because of their high
flavour and taste. It is widely cultivated in many tropical and subtropical moisture content, and require proper pre-processing steps. Drying
regions worldwide. At present, Coffea canephora (robusta) and Coffea technologies play a major role in food manufacturing and was consid­
arabica (arabica) are the most common species in the coffee market ered as a cheap method to extend the shelf life of agricultural products.
(Cheng et al., 2019). In China, robusta coffee was primarily cultivated in Drying is considered as an efficient process to inhibit microbial
Hainan province, Xinglong coffee is one of the most popular coffee spoilage, delay the enzymes activities prevent chemical deterioration,
brands in China, and protected by the Chinese AQSIQ (the People’s and prolong shelf life. During drying process, many metabolic activities,
Republic of China’s Food and Drug Administration) “Controlled Desig­ including the interconversion of low-molecular-weight sugars and pro­
nation of Origin”. With the increase of the quality of life, coffee con­ tein hydrolysis, can change the volatile compounds. During coffee pro­
sumption has increased rapidly in recent years (Stavridou, Soufleros, cessing, drying can alter the metabolism and formation of important
Bouloumpasi, & Dagkli, 2016), bringing abundant numerous economic aroma precursors, particularly aldehydes, ketones, and alcohol via
efficiency to business and agriculture and promoted local framers’ in­ Maillard reactions (Kulapichitr, Borompichaichartkul, Suppavorasatit,
come (Dong, Zhao, Hu, Dong, & Tan, 2017). After the postharvest, coffee & Cadwallader, 2019; Iriondo-DeHond et al., 2020). Existing literatures

* Corresponding author. Spice and Beverage Research Institute, Chinese Academy of Tropical Agricultural Sciences, Wanning, Hainan, 571533, China.
** Corresponding author.
E-mail addresses: [email protected] (W. Dong), [email protected] (Z. Zhang).
1
This author contributed to the work equally and should be regarded as co-first author.

https://doi.org/10.1016/j.lwt.2022.113317
Received 25 November 2021; Received in revised form 22 January 2022; Accepted 1 March 2022
Available online 18 March 2022
0023-6438/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
K. Zhang et al. LWT 161 (2022) 113317

focus on how drying methods impact the volatile compounds of coffee March 2020. All harvested coffee fruits selected as the experimental
beans, for example, the impact of room temperature drying, freeze materials were of commercial maturity, and free from visible blemishes,
drying, hot-air drying, microwave vacuum drying and heat-pump drying and physical damage. After harvesting, the coffee fruits were immedi­
on the colour, fatty composition, bioactive components, and volatiles of ately transported to the laboratory, and separated from the leaves, twigs,
coffee samples (Cheng et al., 2019; Dong, Hu, Chu, Zhao, & Tan, 2017). and other debris. They were mechanically peeled and degummed. Wet
In addition, some studies investigated the major chemical components, coffee beans were kept at 4 ◦ C for a maximum of 24 h before drying.
such as sugars, amino acids, and protein contents in the drying process
(de Melo Pereira et al., 2019). During drying, volatile compounds posed 2.3. Moisture content determination
an important effects on final product. However, there are no reports
about monitoring the changes in volatile compounds during drying. Moisture content was determined by the fast moisture analyser
Therefore, it is necessary to clarify the volatile change rule during drying (MB25; OHAUS, Shanghai, China), and calculated as the weight loss as a
process to regulate the quality of the final product. percentage of the initial weight.
Nowadays, headspace solid-phase microextraction/gas
chromatography-mass spectrometry (HS-SPME/GC-MS) and electronic 2.4. Drying process
nose technology (E-nose) are primarily utilised to analyse food flavour,
and have been widely used in quality control, cultivar discrimination, The green coffee beans were dried via hot air drying, a common
and geographical origin authentication of agricultural products (Liu, method used in food drying. The samples (approximately 2 kg) divided
Deng, Bi, Xu, & Zhang, 2019; Rocchi et al., 2019; Wang, Gao, Liang, Liu, into four parts were uniformly distributed in stainless steel plates with
& Gao, 2020; Wang, Sun, Lassabliere, Yu, &;Liu, 2020). GC-MS is holes forming a single layer. They were dehydrated at 40 ◦ C, 50 ◦ C, and
generally considered a qualitative and quantitative method to analyse 60 ◦ C in an electro-thermal constant-temperature dry box (DHG-9030A,
volatiles, and E-nose is made up of an array of electronic chemicals to Shanghai Yiheng Laboratory Instrument Co., Ltd., Shanghai, China).
imitate human olfactory function. In the previous study, GC-MS and Based on the volatile compound changes in drying kinetics of pre-
E-nose were used to analyse volatile compounds of green coffee beans experiments, moisture content and volatile compounds were deter­
and roasted coffee beans, thereby comparing the influence of different mined every 3 h, to evaluate the variations of volatiles in coffee beans
processing technologies on coffee samples (Dong et al., 2019; Kula­ during drying process. Furthermore, the sampling interval at 40 ◦ C was
pichitr et al., 2019). In addition, volatile compounds of Flammulina set as 0 h, 3 h, 6 h, 9 h, 12 h, 15 h, 18 h, 24 h, 36 h, and 48 h (40‒0 h, 40‒
velutipes detected by GC-MS and E-nose, and a large number of volatile 3 h, 40‒6 h, 40‒9 h, 40‒12 h, 40‒15 h, 40‒18 h, 40‒24 h, 40‒36 h,
compounds, namely, ketones, alcohols, aldehydes, and hydrocarbon 40–48 h) respectively, The sampling interval at drying temperature of
were identified (Yang et al., 2016). HS-SPME/GC-MS coupled with 50 ◦ C was set as 0 h, 3 h, 6 h, 9 h, 12 h, 15 h, 18 h, 21 h (50‒0 h, 50‒3 h,
E-nose were also applied for the analysis of flavor compounds in fruits, 50‒6 h, 50‒9 h, 50‒12 h, 50‒15 h, 50‒18 h, 50‒21 h) respectively,
vegetables, tea, and olive oil (López-López, Cortés-Delgado, de Castro, and the sampling interval at 60 ◦ C was set as 0 h, 3 h, 6 h, 9 h, 12 h, 15 h,
Higinio Sánchez, & Montaño, 2019; Song et al., 2020). However, little 18 h (60‒0 h, 60‒3 h, 60‒6 h, 60‒9 h, 60‒12 h, 60‒15 h, 60‒18 h)
information is available on the drying process of coffee beans, respectively. The final moisture content of the samples was approxi­
HS-SPME/GC-MS coupled with E-nose could be applied to establish mately 11.0 ± 1.0 g/100 g DW, as recommended by the Agricultural
volatile compounds fingerprints in coffee beans during hot air drying Industry Standard of China (NY/T 604-2020) to prevent deterioration
process. before dehulling and roasting. The whole coffee beans dried under
In this study, HS-SPME/GC-MS combined with E-nose techniques different stages were smashed using an ultrafine grinder, and the
were applied to clarify the evolution and formation of volatile com­ resulting homogenates were used for the extraction process of HS-
pounds in coffee beans during hot air drying at different temperatures SPME/GC-MS and E-nose analysis.
(40 ◦ C, 50 ◦ C, and 60 ◦ C), and the flavour fingerprints of coffee beans
were also established. Furthermore, principal component analysis 2.5. HS-SPME/GC-MS analysis
(PCA), partial least squares-discriminant analysis (PLS-DA), and heat
map analysis were used to identify the characteristic compounds among The volatile compounds of coffee samples were extracted via the HS-
different treated samples. This study aimed to provide theoretical evi­ SPME according to our previous study with minor revision (Dong, Tan,
dence to control flavour and quality loss of coffee beans during drying. Zhao, Hu, & Lu, 2015). Briefly, an internal standard was prepared by
The results reported hereIN will could act as a reference guide to pro­ adding 5 μL 3-heptanone into 100 mL methanol (Damao, Tianjing,
duce high-quality coffee beans and improve the flavour quality concern China)The samples (1.0 g each) were put into 10 mL headspace vials,
in coffee conservation. and 2 μL of internal standard solution was added to each coffee sample
and kept for 1 h prior to GC-MS analysis. A 75 μm CAR/PDMS solid
2. Materials and methods phase extraction fibre was inserted into vial bottle to adsorb volatile
compounds at 60 ◦ C for 40 min, and then inserted into the gas chro­
2.1. Chemicals and reagents matographic system inlet to desorb volatile compounds at 250 ◦ C for 3
min. The speed of the agitator was set as 250 rpm,. The volatile com­
High-performance liquid chromatography (HPLC) grade methanol pounds of the coffee samples were analysed using an Agilent 7890A GC
was purchased from Merck (Darmstadt, Germany). The standard of C7‒ 5975C system (Agilent Technologies, Santa Clara, CA) coupled with a
C40 saturated alkanes was obtained from Sigma-Aldrich (St. Louis, MO, quadrupole mass filter. The separation of samples was achieved by a
USA). The 3-heptanone standard (purity>98.0%) was supplied by DB-WAX capillary column (30 m × 0.25 mm × 0.25 μm). The initial
Aladdin Industrial Co. Ltd. (Shanghai, China). SPME fibre (75 μm CAR/ temperature of the column was 40 ◦ C held for 5 min, and raised to 130 ◦ C
PDMS) was purchased from Supelco Co. (Bellefonte, PA, USA). All at 1.5 ◦ C/min, then raised again to 200 ◦ C at 4 ◦ C/min and kept for 5
chemicals were of analytical and HPLC grade. min. The carrier gas was helium and the flow rate was 1.0 mL/min. The
sample was injected at 250 ◦ C in a split-less mode. The electron impact
2.2. Samples and sample preparation ionisation, quadrupole, and transfer line temperatures were set to 250,
150, and 250 ◦ C, respectively. Mass spectra and total ion chromatograms
Fresh coffee fruits (Coffea canephora) were harvested from the were recorded in full-scan mode from 50 m/z to 350 m/z at a rate of 3.06
experimental station in the Spice and Beverage Research Institute of the scan/s. The identified volatile compounds were compared with the
Chinese Academy of Tropical Agricultural Sciences (Hainan, China) in NIST14.0 spectral database and related literature, and retention indices

2
K. Zhang et al. LWT 161 (2022) 113317

Fig. 1. Changes in moisture content of green coffee beans at 40 ◦ C, 50 ◦ C, and 60 ◦ C during drying process.

Fig. 2. Total ion current chromatograms of wet green coffee beans (a), and green coffee beans dried at 40 ◦ C (b), 50 ◦ C (c), and 60 ◦ C (d).

(RI) were calculated using saturated alkane standard C7–C40 injected 2.6. E-nose analysis
into GC-MS under the same chromatographic conditions. The relative
content of each compound was calculated by comparing its peak area The E-nose analyses were carried out by a Fox 3000 electronic nose
with that of the internal standard. The dry weight basis was used to (Alpha MOS., Toulouse, France). The E-nose is composed of pattern
compensate for the differences in drying loss and reached an objective recognition software, sensor array unit, and an HS-100 automatic
judgement between fresh and dried samples. GC-MS analysis were sampler. The sensor array has six metal-oxide semiconductors (T30/1,
completed in triplicate. T70/2, PA/2, P30/2, LY2/AA, and LY2/gCT). Some literatures illustrate
the function of sensors of Alpha M.O.S. E-nose: T30/1, sensitive to Polar
compound, acids compounds; T70/2, sensitive to aromatic compounds;

3
K. Zhang et al. LWT 161 (2022) 113317

Fig. 3. Hierarchical clustering and heat map visualisation of changes in volatile compounds (a), the relative concentration of volatile compounds at 40 ◦ C (b), 50 ◦ C
(c), and 60 ◦ C (d).

PA/2, sensitive to ethanol and ketone compounds. P30/2, sensitive to 3. Result and discussion
organic solvents and light polar molecules; LY2/AA, sensitive to
ammonia and ketones; and LY2/gCT, sensitive to propane, butane, and 3.1. Evaluation of moisture content of coffee beans during drying process
ethanol (Pei et al., 2016). The coffee powder sample (1.0 g) was placed
in a 10 mL sample bottle, and incubated at 60 ◦ C for 5 min. The Fig. 1 shows the moisture contents of green coffee bean dried at
collection delay time was 300 s and the data collection time was 90 s. different temperature. The moisture content reached the final 10%
The sample injection volume was 300 μL. Each sample was determined under different treatments for 48 h, 21 h, and 18 h at 40 ◦ C, 50 ◦ C, and
in triplicate, and the mean value was used for further analysis. 60 ◦ C, respectively. Generally, the initial moisture content of green
coffee beans was 55%–60%, and the final moisture content reached
2.7. Statistical analysis approximately 10%–12% after the drying process. Therefore, the drying
times were 48 h, 21 h, and 18 h at 40 ◦ C, 50 ◦ C, and 60 ◦ C, respectively,
All data of coffee samples were analysed by one-way analysis of and were regarded as the termination of drying process. Generally, heat
variance (ANOVA) and Duncan’s New Multiple Range Test using SPSS transfer and phase changes mainly occurred during drying processing.
22.0 (SPSS Inc., Chicago, IL, USA). The clustering heat map analysis and The different temperature affected the energy of water molecules and
PLS-DA were accomplished by the Metabo Analyst 4.0 software moisture diffusivity, which may explain why the moisture evaporation
(http://www.metaboanalyst.ca/). All statistical tests were two-tailed rate is faster at 60 ◦ C (Li et al., 2019). For different drying processes, the
and p < 0.05 or less was considered as significant. Compounds with moisture content decreased rapidly in the early drying stage, then
variable importance for the projection (VIP) score >1 was identified as a decreased slowly until it reached equilibrium. This result is consistent
key volatile compound. The data were normalized via auto-scaling for with majority of food characteristics (Zhang et al., 2020).
further analysis, and all data were expressed as the mean ± standard
deviation (SD) of triplicate tests. 3.2. Analysis of volatile compounds changes in coffee beans during drying
process

The amount and variety of volatile compounds identified by HS-

SPME/GC-MS was shown in Table 1 and the ion current chromato­
gram of green coffee beans at different drying temperatures were

4
K. Zhang et al. LWT 161 (2022) 113317

Fig. 4. Response intensity of E-nose sensors to wet green coffee beans (a), and green coffee beans dried at 40 ◦ C (b), 50 ◦ C (c), and 60 ◦ C (d).

displayed in Fig. 2a–d. Sixty two volatile compounds were identified and 3.2.2. Effect of the different drying processes on the aldehydes compounds
classified by their chemical group, including 7 acids, 12 aldehydes, 13 in wet coffee beans
hydrocarbons, 2 ketones, 7 esters, 4 furans, 9 alcohols, 3 phenols, 2 A series of aldehydes consisting of hexanal, 2-hexenal, octanal,
sulfur compounds, and 3 others. Previous studies illustrated various nonanal and benzaldehyde characterising green and fruity odors, were
volatile compounds during drying coffee beans in terms of the esters, found to be the major aldehydes present in fresh coffee beans (Poyraz,
aldehydes, alcohols, acids, hydrocarbons and ketones (Dong, Hu, et al., Öztürk, Kiyan, & Demirci, 2016). Their contents were found to be
2017). Heat map analysis of the volatile compounds of drying process changed during drying process, and the formation of aldehydes was
was exhibited in Fig. 3a, it ievidently showed that benzaldehyde, methyl mainly originated from lipid oxidation degradation and Maillard reac­
salicylate, D-Limonene, styrene and benzeneacetaldehyde had relatively tion (Zhang et al., 2019; Zhou et al., 2019). In the drying process, the
higher contents, followed by benzyl alcohol, acetic acid, hexanal, and aldehydes were the most abundant volatile compounds in the early stage
2-heptanol. of drying which was observed in other foods (Yang et al., 2016). The
total concentration of aldehydes decreased from 45.98 to 0.66 μg/g DW,
3.2.1. Effect of the different drying processes on the acids compounds in wet 45.98 to 1.21 μg/g DW and 45.98 to 1.35 μg/g DW in 40 ◦ C, 50 ◦ C, and
coffee beans 60 ◦ C, respectively, especially for benzaldehyde and benzeneacetalde­
The total concentration of volatile acids decreased during drying hyde, which was similar to other study that the aldehydes of muskmelon
process at different temperatures, from 2.62 to 0.69 μg/g DW, 2.62 to puree were lost during heating (Priyanka, Sindhoora, Vijayanand, Kul­
1.01 μg/g DW, and 2.62 to 1.03 μg/g DW at 40 ◦ C, 50 ◦ C, and 60 ◦ C, karni, & Nagarajan, 2015). And the aldehydes was the most abundant
respectively. Acetic acid and 3-methyl-2-butenoic acid profiles were volatile compounds in the early stage of drying, this was consistent with
higher than other acids in the drying process, this result was accordance the drying processing of Flammulina velutipes (Yang et al., 2016). Drying
with the heat map analysis results in Fig. 3a. Previous studies have processing significantly decreased the concentration of benzaldehyde
identified acetic acid in the fermentation and drying process of green and benzeneacetaldehyde at different temperatures. In addition, the
coffee bean, and it also as the major volatile compounds in the drying hexanal was also identified in our study, imparting grassy and fatty
stage (Martín et al., 2017). Acetic acid was the main volatile compound notes for green coffee beans, and previous studies indicated it was the
in coffee beans, which was produced by carbohydrate catabolism most predominant in Turkish green coffee bean (Poyraz et al., 2016).
(Korkmaz, Atasoy, & Hayaloglu, 2020). Acetic acid is known for its
pungent, sour and vinegar-like odour found among different tempera­ 3.2.3. Effect of the different drying processes on the hydrocarbons
tures. Acetic acid content increased and showed higher content in 50 ◦ C compounds in wet coffee beans
and 60 ◦ C at the end of process (Abouelenein et al., 2021). Hence, high Hydrocarbons are the part of the volatile compounds in green coffee
temperature promoted of acetic acid retention in dry coffee beans. bean that are observed during drying process (de Melo Pereira et al.,
Moreover 3-methylbutanoic acid was reported in Hawaiian green coffee, 2019). This result indicated that the total concentration of hydrocarbons
was found in different process, however, there was no significant dif­ decreased at 40 ◦ C. In contrast, the number of hydrocarbons exhibited
ference across the three drying temperatures. By contrast, the relative an increasing trend in drying temperature at 50 ◦ C and 60 ◦ C. Although
concentration of acids showed an increasing trend in different temper­ previous studies reported that some hydrocarbons are susceptible to
atures from 2.33% to 17.41%, 2.33%–12.85%, and 2.33%–12.33% at evaporation losses due to their high partition coefficient, in this study,
40 ◦ C, 50 ◦ C, and 60 ◦ C, respectively (Fig. 3b–d). Overall, the volatile the D-Limonene and o-Cymene, which imparted the citrus and lemon
acids contents decreased during drying process except for 3-methyl-2-­ odors, showed the significantly increased (de Melo Pereira et al., 2019;
butenoic acid. Polat, Guclu, Kelebek, Keskin, & Selli, 2022; Rajkumar, Shanmugam,

5
K. Zhang et al. LWT 161 (2022) 113317

3.2.4. Effect of the different drying processes on the esters compounds in

wet coffee beans
Esters are formed by the oxidation of unsaturated fatty acids in coffee
beans and are typically considered as a pleasant flavor, imparting the
notes of fruity. The change rule of esters was recently described in roast
processing (Caporaso, Whitworth, Cui, & Fisk, 2018). As shown in
Table 1, the total ester contents decreased significantly during drying
process with the loss of moisture content, from 42.84 to 0.54 μg/g DW,
42.84 to 1.72 μg/g DW, and 42.84 to 1.18 μg/g DW respectively. The
total esters content of samples at 50 ◦ C and 60 ◦ C were higher than that
of samples at 40 ◦ C at the end of drying, those proved that esters were
preserved in high temperature during drying process (Ge et al., 2020).
Methyl salicylate and ethyl 2-hydroxybenzoate were the dominant vol­
atile compounds in the drying processing with the loss of moisture, this
phenonemon was found when the C8 compounds decreased in the
drying process of button mushroom, and was also found in peppers
during hot air drying (Ge et al., 2020; Pei et al., 2016). In Fig. 3b–d, the
initial relative ester concentrations was relatively higher, which then
decreased in the drying process. Coffee beans possessed abundant un­
saturated fatty acid, and esters were mainly generated by their oxida­
tion, demonstrating the increase in esters during the early stage of the
drying process.

3.2.5. Effect of the different drying processes on the alcohols compounds in

wet coffee beans
The coffee beans exhibited a relatively higher concentration of al­
cohols, particularly 2-heptanol, 1-octen-3-ol and benzyl alcohol. Phe­
nylethyl alcohol was identified in a higher content in Hawaiian green
coffee (de Melo Pereira et al., 2019). In the present study, the most of
alcohols were the 2-heptanol, 1-octen-3-ol, benzyl alcohol, and phe­
nylethyl alcohol, which was accordance with the previous studies
(Dong, Hu, et al., 2017). The alcohols of green coffee beans showed a
dramatically decreasing trend from 10.55 to 0.53 μg/g DW, 10.55 to
0.85 μg/g DW, and 10.55 to 0.70 μg/g DW at 40 ◦ C, 50 ◦ C, and 60 ◦ C,
respectively. Similar evolutionary trends were also observed in a pre­
vious study that reported the alcohols contents in mushroom signifi­
cantly decreased after the drying process (Zhang et al., 2021). However,
the relative concentration of alcohols fluctuanted during drying process,
and the alcohol concentration of samples dried at 40 ◦ C was slightly
higher than that of samples dried at 50 ◦ C and 60 ◦ C. This effect was
primarily caused by the loss of other flavour compounds in relatively
higher temperature (Fig. 3b–d). The benzyl alcohol and phenylethyl
alcohol impart a floral note for green coffee beans. It was found that the
content of benzyl alcohol and phenylethyl alcohol also showed an
fluctuanting trends during drying process. However, but the actual
content of dried samples was approximately 7–10 fold compared to wet
samples. These results demonstrated that the drying process induced
volatile alcohols loss.

3.2.6. Effect of the different drying processes on the other volatile

compounds in wet coffee beans
Fig. 5. Changes of response values of sensors from E-nose at 40 ◦ C (b), 50 ◦ C The phenols, sulfur compounds, and ketones were also identified in
(c), and 60 ◦ C (d) during hot air drying. green coffee beans, the present study was accordance with their reported
result (de MeloPereira et al., 2019). In this study, 2-pentylfuran,
Galvâo, Leite Neta et al., 2017). In addition, in terms of relatively con­ methional and caffeine made up most of the remaining of volatile
centration of hydrocarbons, all of samples showed an increasing trends compounds. The drying process reduced its concentration to lower than
along with drying process, the relatively concentration was the most 0.1 μg/g DW by its completion, this was accordance with the pervious
abundant volatile compounds at the end of drying, reaching to the study of coffee bean process using drying (Kulapichitr et al., 2019).
36.46%, 36.02%, and 42.36%, respectively (Fig. 3b–d). The content of
D-limonene, styrene, and o-cymene significant increased with the rise of 3.3. E-nose response and PCA of coffee beans
drying temperature which formed the basis of coffee flavor in subse­
quent process. E-nose was considered as a better method to differentiate the dif­
ference of flavours. Hence, it has been applied to many agricultural
products, such as nuts, vegetable, and fruits (Ali, Hashim, Aziz, &
Lasekan, 2020). The analysis of flavor using E-nose was performed on
the roast coffee beans and dried coffee beans, to date, there is no

6
K. Zhang et al. LWT 161 (2022) 113317

Fig. 6. Multivariate analysis of volatile compounds using PLS-DA during hot air drying of green coffee beans. PLS-DA score plot (a) and volatile compounds (b)
ranked by VIP scores at 40 ◦ C, PLS-DA score plot (c) and volatile compounds (d) ranked by VIP scores at 50 ◦ C, PLS-DA score plot (e) and volatile compounds (f)
ranked by VIP scores at 60 ◦ C.

7
K. Zhang et al. LWT 161 (2022) 113317

research related to the drying process of green coffee beans (Dong, Hu, and 2-heptanol in the drying process, consistent with the PLS-DA result.
et al., 2017; Dong et al., 2019). In this study, the flavor changes of green
coffee bean were comprehensively analysed by E-nose during drying
process. Fig. 4 displayed that the dynamic intensity changes of green 3.5. The correlation analysis of E-nose and volatile compounds of coffee
coffee bean at different temperatures during drying. The significant beans
variance of sensors in the drying process symbolized the formation of
new flavour compounds and the difference in volatile compounds (Cui, Based on the results of above analysis, GC-MS and E-nose could
Wang, Yang, Wu, & Wang, 2015). According to the signal response of the distinguish the differences in green coffee beans dried at different
E-nose in Fig. 4a–d, the response value of P30/2, PA/2, T70/2, and temperature. As we known, GC-MS primarily focuses on the identifica­
T30/1 increased. When it reached the peak value, the P30/2, PA/2, tion and quantification of a single substance. E-nose rapidly screens and
T70/2, and T30/1 response values declined sharply and then remained identifies tested samples, however, it typically has low sensitivity and
stable. Conversely, the LY2/AA response value decreased first, then accuracy. Therefore, combining these techniques can enhance their
increased until reaching a stable equilibrium, while the LY2/gCT ability to analyse the tested samples (Barbosa-Pereira, Rojo-Poveda,
remained stable. Notably, with the increasing drying temperature, the Ferrocino, Giordano, & Zeppa, 2019). The correlation of analysis was
PA/2 peak value decreased. As shown in Fig. 5a–c, the signal responses visualized between e-nose and volatile compounds with VIP >1 in
drastically declined during process, particularly the PA/2, from 366.16 Fig. S2a-c, which indicated that the responses of LY2/AA and LY2/gCT
to 8.25, 366.16 to 7.27 and 366.16 to 5.37, at 40 ◦ C, 50 ◦ C, and 60 ◦ C, sensors had negative with the hydrocarbons during drying process at
respectively. At the end of drying, the PA/2 response was still the 40 ◦ C. In contrast, T30/1, T70/2, PA/2, and P30/2 sensors exhibited
highest than that of other sensors, and the response of LY2/gCT was the strong positive with ketones. When the drying temperature was 50 ◦ C,
lowest among all the sensors. LY2/AA, LY2/gCT and P30/2 had negative correlations with acids,
PCA was applied to the E-nose data matrix from of coffee samples whileas T30/1, T70/2, PA/2, and P30/2 sensors had significant positive
with different drying processes. The first two principal components (PC1 relationships with sulfur compounds and ketones. With the increasing
and PC2) contributed to the total variance by 98.1%, 98.5%, and 99.6%, temperature of hot air drying, the results showed that the significant
respectively, demonstrating that PC1 and PC2 could explain the overall difference of T30/1, T70/2, PA/2, and P30/2 sensors had strong positive
features of the tested samples. As shown in Fig. S1a-c, the coffee samples correlations with sulfur compounds and esters. However, LY2/AA and
from the initial drying process stage were distributed in the PC1 positive LY2/gCT were positively correlated with aldehydes. The correlation
result indicated that E-nose could distinguish green coffee beans in
axis, which mainly were the samples dried for 0 h, 3 h, and, 6 h in
different processes. The coffee samples from the rest drying stages were different drying processes by sensors responses.
located at PC1 negative axis. The 40-3 h and 40-6 h samples were
distinguished from the other samples dried by hot air at 40 by the PA/2, 4. Conclusions
P30/2, and T30/1 sensors. The samples dryied at 50 ◦ C had similar
distribution trends to those dried at 40 ◦ C. With all the samples projected This study illustrated the concentrations and contents of volatile
on the PC1 axis, they can be differentiated according to drying time, and compounds of robusta coffee beans during hot air drying for the first
the 50-0 h and 50-3 h samples were located at the right of the Y-axis, the time. Observations demonstrated that the benzaldehyde, methyl salic­
50-6 h, 50-9 h, 50- 12 h, 50-15 h, 50-18 h, and 50-21 h samples were all ylate, D-limonene, styrene, and benzeneacetaldehyde showed a relative
located at the left of Y-axis, there was a slightly overlapping between 50- higher content during drying. Alcohol, ketones, and esters were the
18 h and 50-21 h samples. Similarly, coffee samples dried for different main volatile compounds during drying. PCA combined with E-nose
drying times at 60 ◦ C were distributed from right to left along the PC1 could effectively distinguish coffee beans treated under different drying
axis in the PC1-PC2 bio-plot. times. PLS-DA and heat map coupled with HS-SPME/GC-MS suggested
that ketones and esters greatly contributed greatly to the discrimination
3.4. PLS-DA analysis of volatile compounds of coffee beans treated by of coffee samples, which was in consist with PCA results. Hot air dying at
different drying process 50 ◦ C could be an alternative to traditional drying methods owing to its
relatively high efficiency and favourable volatile compounds conserva­
PLS-DA has been widely applied to analyse the interrelation of tion. Furthermore, our findings have established the content and con­
different volatile compounds (Yu et al., 2020). In this study, a 200 centration patterns of volatile compounds during hot air drying. They
permutation test was performed to verify the fit of the PLS-DA model, supply theoretical evidence for further mechanistic research into the of
and results indicated that the model was not overfitting. In addition, to volatile compounds of coffee beans, and more reference guidance for
evaluate the robustness and prediction performance of PLS-DA, leave improving the quality coffee beans. It also provides an excellent example
one out cross-validation (LOOCV) was implemented based on the for better understanding the evolutionary mechanisms of volatile com­
experimental datasets. PLS-DA was performed on the data matrix (33, positions in other aromatic food products.
24, and 21 samples, respectively x 10 types of volatile compounds) of
volatile compounds from coffee beans. The principal component 1 Conflict of interests
(PLS1) and principal component 2 (PLS2) explained the total variance of
49.6% and 40.8%, 77.2% and 7.6%, and 78.1% and 13.5% at 40 ◦ C, The authors declare that they have no known competing financial
50 ◦ C, and 60 ◦ C, respectively, demonstrating that PLS1 and PLS2 could interests or personal relationships that could have appeared to influence
explain the difference of volatile compounds (Fig. 6a–f). VIP ≥1 was the work reported in this paper.
considered to significantly contribute to the differentiation of the
various samples. Therefore, the volatile compounds accounting for the CRediT authorship contribution statement
main contributors were hydrocarbons, phenols, and esters at 40 ◦ C
during hot air drying process (Fig. 6a and b). At 50 ◦ C, the main volatile Ke Zhang: Methodology, Validation, Writing – original draft. Jin­
compounds contributors were ketones, esters, others, and acids (Fig. 6c huan Cheng: Conceptualization, Supervision. Qidi Hong: Methodol­
and d), and Fig. 6e and f demonstrated that ketones, others, esters, and ogy, Software. Wenjiang Dong: Methodology, Software, Formal
sulfur compounds contributed greatly to the samples at 60 ◦ C. In com­ analysis, Investigation, Writing – original draft, Writing – review &
bination with the heat map analysis results, the relatively higher content editing, Visualization, Funding acquisition. Xiaoai Chen: Investigation,
of volatile compounds comprised benzaldehyde, methyl salicylate, Supervision. Guiping Wu: Software, Supervision. Zhenzhen Zhang:
D-limonene, styrene, benzeneacetaldehyde, benzyl alcohol, acetic acid Supervision, Project administration.

8
K. Zhang et al. LWT 161 (2022) 113317

Acknowledgements Liu, X., Deng, J. K., Bi, J. F., Wu, X. Y., & Zhang, B. (2019). Cultivar classification of
cloudy apple juices from substandard fruits in China based on aroma profile
analyzed by HS-SPME/GC-MS. Lebensmittel-Wissenschaft und -Technologie- Food
We thank the Hainan Provincial Natural Science Foundation of China Science and Technology, 102, 304–309.
(2019CXTD416), National Natural Science Foundation of China López-López, A., Cortés-Delgado, A., de Castro, A., Sánchez, A. H., & Montaño, A. (2019).
(31872888), China Central Public-Interest Scientific Institution Basal Changes in volatile composition during processing and storage of black ripe olives.
Food Research International, 125, 108568.
Research Fund (1630142022010), and Yunnan Provincial Rural Revi­ Martín, A., Hernández, A., Aranda, E., Casquete, R., Velázquez, R., Bartolomé, T., et al.
talization of Science and Technology Project of China (2017). Impact of volatile composition on the sensorial attributes of dried paprikas.
(202104BI090015) for their financial support. Food Research International, 100, 691–697.
de Melo Pereira, G. V., de Carvalho Neto, D. P., Magalhães Júnior, A. I., Vásquez, Z. S.,
Medeiros, A. B. P., Vandenberghe, L. P. S., et al. (2019). Exploring the impacts of
Appendix A. Supplementary data postharvest processing on the aroma formation of coffee beans - a review. Food
Chemistry, 272, 441–452.
Pei, F., Yang, W. J., Ma, N., Fang, Y., Zhao, L. Y., An, X. X., et al. (2016). Effect of the two
Supplementary data related to this article can be found at htt drying approaches on the volatile profiles of button mushroom (Agaricus bisporus) by
ps://doi.org/10.1016/j.lwt.2022.113317. headspace GC–MS and electronic nose. Lebensmittel-Wissenschaft und -Technologie-
Food Science and Technology, 72, 343–350.
Polat, S., Guclu, G., Kelebek, H., Keskin, M., & Selli, S. (2022). Comparative elucidation
References of colour, volatile and phenolic profiles of black carrot (Daucus carota L.) pomace
and powders prepared by five different drying methods. Food Chemistry, 369,
Abouelenein, D., Mustafa, A. M., Angeloni, S., Borsetta, G., Vittori, S., Maggi, F., et al. 130941.
(2021). Influence of freezing and different drying methods on volatile profiles of Poyraz, I. E., Öztürk, N., Kiyan, H. T., & Demirci, B. (2016). Volatile compounds of Coffea
strawberry and analysis of volatile compounds of strawberry commercial jams. arabica L. green and roasted beans. Anadolu University Journal of Science and
Molecules, 26, 4153. Technology, 5, 31–35.
Ali, M. M., Hashim, N., Aziz, S. A., & Lasekan, O. (2020). Principles and recent advances Priyanka, D., Sindhoora, S., Vijayanand, P., Kulkarni, S. G., & Nagarajan, S. (2015).
in electronic nose for quality inspection of agricultural and food products. Trends in Influence of thermal processing on the volatile constituents of muskmelon puree.
Food Science & Technology, 99, 1–10. Journal of Food Science & Technology, 52, 3111–3116.
Barbosa-Pereira, L., Rojo-Poveda, O., Ferrocino, I., Giordano, M., & Zeppa, G. (2019). Rajkumar, G., Shanmugam, S., de Sousa Galvâo, M., Neta, M. T. S. L., Sandes, R. D. D.,
Assessment of volatile fingerprint by HS-SPME/GC-qMS and E-nose for the Mujumdar, A. S., et al. (2017). Comparative evaluation of physical properties and
classification of cocoa bean shells using chemometrics. Food Research International, aroma profile of carrot slices subjected to hot air and freeze drying. Drying
123, 684–696. Technology, 35, 699–708.
Caporaso, N., Whitworth, M. B., Cui, C. H., & Fisk, I. D. (2018). Variability of single bean Rocchi, R., Mascini, M., Faberi, A., Sergi, M., Compagnone, D., Martino, V. D., et al.
coffee volatile compounds of arabica and robusta roasted coffees analysed by SPME- (2019). Comparison of IRMS, GC-MS and E-nose data for the discrimination of
GC-MS. Food Research International, 108, 628–640. saffron samples with different origin, process and age. Food Control, 106, 106736.
Cheng, K., Dong, W. J., Long, Y. Z., Zhao, J. P., Hu, R. S., Zhang, Y. J., et al. (2019). Song, J. X., Chen, Q. Q., Bi, J. F., Meng, X. J., Wu, X. Y., Qiao, Y. N., et al. (2020). GC/MS
Evaluation of the impact of different drying methods on the phenolic compounds, coupled with MOS E-nose and flash GC E-nose for volatile characterization of
antioxidant activity, and in vitro digestion of green coffee beans. Food Sciences and Chinese jujubes as affected by different drying methods. Food Chemistry, 331,
Nutrition, 7, 1084–1095. 127201.
Cui, S. Q., Wang, J., Yang, L. C., Wu, J. F., & Wang, X. L. (2015). Qualitative and Stavridou, K., Soufleros, E. H., Bouloumpasi, E., & Dagkli, V. (2016). The phenolic
quantitative analysis on aroma characteristics of ginseng at different ages using E- potential of wines from French grape varieties Cabernet Sauvignon, Merlot and
nose and GC–MS combined with chemometrics. Journal of Pharmaceutical and Syrah cultivated in the region of Thessaloniki (Northern Greece) and its evolution
Biomedical Analysis, 102, 64–77. during aging. Food and Nutrition Sciences, 7, 122–137.
Dong, W. J., Hu, R. S., Chu, Z., Zhao, J. P., & Tan, L. H. (2017). Effect of different drying Wang, Y., Gao, Y. Y., Liang, W., & Gao, H. Y. (2020). Identification and analysis of the
techniques on bioactive components, fatty acid composition, and volatile profile of flavor characteristics of unfermented stinky tofu brine during fermentation using
robusta coffee beans. Food Chemistry, 234, 121–130. SPMS-GC-MS, E-nose, and sensory evaluation. Journal of Food Measurement and
Dong, W. J., Hu, R. S., Long, Y. Z., Li, H. H., Zhang, Y. J., Zhu, K. X., et al. (2019). Characterization, 14, 597–612.
Comparative evaluation of the volatile profiles and taste properties of roasted coffee Wang, C. H., Sun, J. C., Lassabliere, B., Yu, B., & Liu, S. Q. (2020). Coffee flavor
beans as affected by drying method and detected by electronic nose, electronic modification through controlled fermentation of green coffee beans by
tongue, and HS-SPME-GC-MS. Food Chemistry, 272, 723–731. Saccharomyces cerevisiae and Pichia kluyveri: Part I. Effects from individual yeasts.
Dong, W. J., Tan, L. H., Zhao, J. P., Hu, R. S., & Lu, M. Q. (2015). Characterization of Food Research International, 136, 109588.
fatty acid, amino acid and volatile compound compositions and bioactive Yang, W. J., Yu, J., Pei, F., Mariga, A. M., Ma, N., Fang, Y., et al. (2016). Effect of hot air
components of seven coffee (Coffea robusta) cultivars grown in Hainan Province, drying on volatile compounds of Flammulina velutipes detected by HS-SPME-GC-MS
China. Molecules, 20, 16687–16708. and electronic nose. Food Chemistry, 196, 860–866.
Dong, W. J., Zhao, J. P., Hu, R. S., Dong, Y. P., & Tan, L. H. (2017). Differentiation of Yu, M. M., Gang, K. Q., Li, C., Wang, J. H., Liu, Y. X., Zhou, D. Y., et al. (2020). Change of
Chinese robusta coffees according to species, using a combined electronic nose and lipids in whelks (Neptunea arthritica cumingi Crosse and Neverita didyma) during cold
tongue, with the aid of chemometrics. Food Chemistry, 229, 743–751. storage. Food Research International, 136, 109330.
Ge, S., Chen, Y., Ding, S., Zhou, H., Jiang, L., Yi, Y., et al. (2020). Changes in volatile Zhang, J. H., Cao, J., Pei, Z. S., Wei, P. Y., Xiang, D., Cao, X. Y., et al. (2019). Volatile
flavor compounds of peppers during hot air drying process based on headspace-gas flavour components and the mechanisms underlying their production in golden
chromatography-ion mobility spectrometry (HS-GC-IMS). Journal of the Science of pompano (Trachinotus blochii) fillets subjected to different drying methods: A
Food and Agriculture, 100, 3087–3098. comparative study using an electronic nose, an electronic tongue and SDE-GC-MS.
Iriondo-DeHond, A., Elizondo, A. S., Iriondo-DeHond, M., Ríos, M. B., Mufari, R., Food Research International, 123, 217–225.
Mendiola, J. A., et al. (2020). Assessment of healthy and harmful maillard reaction Zhang, L. J., Dong, X. B., Feng, X., Ibrahim, S. A., Huang, W., & Liu, Y. (2021). Effects of
products in a novel coffee cascara beverage: Melanoidins and acrylamide. Foods, 9, drying process on the volatile and non-volatile flavor compounds of Lentinula edodes.
620. Foods, 10, 2836.
Korkmaz, A., Atasoy, A. F., & Hayaloglu, A. A. (2020). Changes in volatile compounds, Zhang, B., Li, M., Qiao, Y., Gao, P., Li, L., & Zheng, Z. (2020). Potential use of low-field
sugars and organic acids of different spices of peppers (Capsicum annuum L.) during nuclear magnetic resonance to determine the drying characteristics and quality of
storage. Food Chemistry, 311, 125910. Arctium lappa L. in hot-blast air. Lebensmittel-Wissenschaft und -Technologie- Food
Kulapichitr, F., Borompichaichartkul, C., Suppavorasatit, I., & Cadwallader, K. R. (2019). Science and Technology, 132, 109829.
Impact of drying process on chemical composition and key aroma components of Zhou, Q., Jia, X., Yao, Y. Z., Wang, B., Wei, C. Q., Zhang, M., et al. (2019).
Arabica coffee. Food Chemistry, 291, 49–58. Characterization of the aroma-active compounds in commercial fragrant rapeseed
Li, M., Chen, Y. N., Wang, X., Cheng, S. P., Liu, F., & Huang, L. Q. (2019). Determination oils via monolithic material sorptive extraction. Journal of Agricultural and Food
of drying kinetics and quality changes of Panax quinquefolium L. dried in hot-blast Chemistry, 67, 11454–11463.
air. Lebensmittel-Wissenschaft und -Technologie- Food Science and Technology, 116,
108563.