Impact of roasting on the phenolic and volatile compounds in coffee beans

Abstract Phenolic compounds present in coffee beans could generate flavor and bring benefits to health. This study aimed to evaluate the impacts of commercial roasting levels (light, medium, and dark) on phenolic content and antioxidant potential of Arabica coffee beans (Coffea arabica) comprehensively via antioxidant assays. The phenolic compounds in roasted samples were characterized via liquid chromatography–electrospray ionization quadrupole time‐of‐flight mass spectrometry (LC‐ESI‐QTOF‐MS/MS). Furthermore, the coffee volatile compounds were identified and semi‐quantified by headspace/gas chromatography–mass spectrometry (HS‐SPME‐GC‐MS). Generally, for phenolic and antioxidant potential estimation, light roasted samples exhibited the highest TPC (free: 23.97 ± 0.60 mg GAE/g; bound: 19.32 ± 1.29 mg GAE/g), DPPH, and FRAP. The medium roasted beans performed the second high in all assays but the highest ABTS+ radicals scavenging capacity (free: 102.37 ± 8.10 mg TE/g; bound: 69.51 ± 4.20 mg TE/g). Totally, 23 phenolic compounds were tentatively characterized through LC‐ESI‐QTOF‐MS/MS, which is mainly adopted by 15 phenolic acid and 5 other polyphenols. The majority of phenolic compounds were detected in the medium roasted samples, followed by the light. Regarding GC‐MS, a total of 20 volatile compounds were identified and semi‐quantified which exhibited the highest in the dark followed by the medium. Overall, this study confirmed that phenolic compounds in coffee beans would be reduced with intensive roasting, whereas their antioxidant capacity could be maintained or improved. Commercial medium roasted coffee beans exhibit relatively better nutritional value and organoleptic properties. Our results could narrow down previous conflicts and be practical evidence for coffee manufacturing in food industries.


| INTRODUC TI ON
Coffee is gradually becoming one of the main commercial food products and world's most widely consumed beverages (Farah, 2012;Valduga et al., 2019). Coffea arabica (Arabica) is one of the major commercial cultivars which takes about 70% of global coffee markets (Rajesh Banu et al., 2020;Waters et al., 2017). Generally, bioactive compounds in coffee could be divided into three major categories, phenolic compounds, flavonoids, and alkaloids. Based on the antioxidant and anti-inflammatory properties of these present bioactive compounds, numerous researchers pointed that regular drinking of coffee could reduce the risk of some chronic diseases including type II diabetes, cardiovascular and autoimmune diseases, and certain types of cancer (Harumi Kondo & Ikewaki, 2012).
Phenolic compounds are second metabolites that commonly exist in higher plants and beverages of plant origin (Farah & Donangelo, 2006). It is present mostly as free (soluble) in plant cell vacuoles and bound (insoluble) forms bound to the cell wall polymeric molecules by ester and glycoside bonds. Hence, the extraction of bound phenolic compounds should use alkaline or acid hydrolysis rather than aqueous organic solvents directly (Mehari et al., 2020). Lu et al. (2020) indicated that the predominant phenolic compounds present in coffee beans are chlorogenic acids (CGAs), particularly 5-O-caffeoylquinic acid (5-CQA). Chlorogenic acids refer to a class of esters derived from several certain hydroxycinnamic acids and quinic acid, including p-coumaroylquinic, feruloylquinic, and caffeoylquinic acids (Gokcen & Sanlier, 2019). Commonly, the content of chlorogenic acids in Arabica ranges from 4% to 8.4%, meanwhile the concentration is increasing with the maturity of coffee beans (Stalmach, 2012).
Chlorogenic acids contribute to the bitter, acid, and astringent flavor of coffee brew, especially because of caffeoylquinic and feruloylquinic acids (Farah, 2012). It also could improve the nutritional value of coffee brew due to its high antioxidant, antibacterial, antiviral, and chemo-preventive capacity (Gokcen & Sanlier, 2019). However, the decomposition of CGAs could easily occur if the temperature is higher than 80°C during processing owing to the thermal instability of polyphenols (Król et al., 2020). Therefore, the physicochemical characteristics of coffee beans are considerably determined by processing conditions (Cordoba et al., 2020;Farah et al., 2005. Roasting is the most essential procedure during the processing chain to generate aroma and flavor (Baggenstoss et al., 2008).
Aroma is an important attribute for the acceptance of coffee beans. Different varieties of coffee beans, their natural origin and processing, especially roasting, will contribute to a variable volatile composition (Somporn et al., 2011). Commercially, the temperature required for three coffee roasting degrees (light, medium, and dark) should be between 195°C and 245°C (Somporn et al., 2011). In chemical aspects, Maillard reaction, nonenzymatic reaction, browning reaction, and Strecker degradation would take place (Farah, 2012;Somporn et al., 2011). The composition of volatiles in green coffee beans, such as aldehydes, ketones, furans, acetic, propanoic, butanoic acid, and other compounds, could be changed during roasting due to those reactions (Somporn et al., 2011). Meanwhile, the interactions would result in the changes in composition involving the loss of polysaccharides, oligosaccharides, chlorogenic acids, and trigonelline, and the generation of lactones of the chlorogenic acids so that influence the antioxidant activities of coffee beans eventually (Baggenstoss et al., 2008). Additionally, melanoidins would be generated via Maillard reaction between amino acids and reducing sugars (Farah & Donangelo, 2006). Quinic acid would be generated as well (Perez-Burillo et al., 2019). Melanoidins and quinic acid are considered as bioactive compounds which could improve the antioxidant, antibacterial, and metal chelating properties of coffee beans (Farah, 2012).
Therefore, the total antioxidant activities of coffee beans would be partially maintained.
Therefore, this research aimed to assess the impact of commercial roasting degrees (light, medium, and dark)
For the estimation of polyphenols and antioxidant potential, sodium carbonate anhydrous, sodium hydroxide pellets, and hydrogen peroxide (30%) were purchased from Chem-Supply Pty Ltd.

| Sample preparation
Roasted coffee beans samples with light-, medium-, and dark-roasted levels used in the intended research project were purchased from Seven Seeds Company, a local coffee retail in Melbourne, Australia.
Roasted coffee beans were milled into dried powder with a mean particle size by coffee grinder (Russell Hobbs Classic, model DZ-1613, Melbourne, VIC, Australia) and then stored at −20°C in dark area before extraction.

| Extraction of free and bound phenolic compounds
The extraction of free phenolic compounds from coffee samples was performed as per the methods described by Peng et al. (2019) with some modifications. Coffee powder was fully mixed with 70% ethanol at 1:10 (w:w) and homogenized for 30 s at 10,000 rpm by Ultra-Turrax T25 Homogenizer (IKA, Staufen, Germany) followed by 12 h incubation under 4°C at 120 rpm in a shaking incubator (ZWYR-240 incubator shaker, Labwit, Ashwood, VIC, Australia).
Mixture was then centrifuged for 15 min at 5000 rpm under 4°C using Hettich Refrigerated Centrifuge (ROTINA380R, Tuttlingen, BadenWürttemberg, Germany). The supernatant fluid was filtered via 0.45 µm syringe filter (Thermo Fisher Scientific Inc., Waltham, MA, USA) and collected as free phenolic extracts.
Bound phenolic compounds in samples were extracted based on Phan et al. (2019) method with some modifications. The sediment went through alkaline hydrolysis by adding 2 M NaOH and incubating for 1 h at 200 rpm in the shaking incubator. Afterwards, concentrated HCl was added to adjust pH 2.0 for acid hydrolysis and recovered pH to 7.0 with 2 M NaOH. Then, the samples were mixed with 70% ethanol and incubated for 60 min to dissolve the released bound phenolic compounds into the organic solvent phase.
The mixture was centrifuged for 20 min at 8000 rpm under 4°C.
The supernatant fluid was collected and filtered by syringe filter as bound phenolic extracts. Both free and bound phenolic extracts were stored under −20°C and ready for further analysis.

| Quantification of phenolic compounds and antioxidant assays
All estimated analyses for phenolic compounds (TPC, TFC, and TCT), as well as the determination of total antioxidant capacity (DPPH, ABTS, FRAP, · OH-RSA, FICA, and RPA), were modified to adapt to the 96-well plate (Costar, Corning, NY, USA) and spectro- The total content of phenolic compounds in coffee beans was estimated through Folin-Ciocalteu method with some modifications based on Mussatto et al. (2011). Briefly, 25 μl sample extract or standard, 25 μl Folin-Ciocalteu reagent solution and 200 μl water were added into plate followed by 5 min incubation at 25°C.
Subsequently, 25 μl of 10% (w/w) sodium carbonate was added followed by 1 h incubation under the same conditions. Gallic acid (0-200 μg/ml) and water were used as calibration curve and blank, respectively. Absorbance was measured at 765 nm and the results were expressed as mg gallic acid equivalents (GAE) per gram based on dry weight (mg GAE/g) ± standard deviation (SD).

| Determination of total flavonoid compounds (TFC)
The total flavonoids content of roasted coffee beans was determined according to Ali et al. (2021). Briefly, 80 μl sample extract, 80 μl 2% aluminum chloride, and 120 µl 50 g/L sodium acetate solution were added into plate in sequence, followed by 2.5 h incubation in the dark at 25°C. Quercetin (0-50 μg/ml) and water were used for standard curve and blank, respectively. Absorbance was measured at 440 nm and the final content of flavonoids in samples was expressed as mg quercetin equivalents (QE) per dry weight (mg QE/g) ± SD.

| Determination of total condensed tannins (TCT)
The total content of condensed tannins in coffee beans was quantified through the vanillin sulfuric acid method according to Ali et al. (2021). Briefly, 25 μl sample extract, 150 μl vanillin solution, and 25 μl 32% sulfuric acid were injected into plate and incubated for 15 min in the dark under 25°C. Catechin (0-1 mg/ml) and water were used for standard curve and blank, respectively. The absorbance was measured at 500 nm and converted into the final content of condensed tannins in coffee beans as mg catechin equivalents (CE) per dry weight (mg CE/g) ± SD.

| 2,2'-diphenyl-2-picryl-hydrazyl (DPPH) antioxidant assay
Modified DPPH assay based on the method of Nebesny and Budryn (2003) was used as a preliminary test for the evaluation of free radical scavenging activity of coffee beans with the change in color from purplish to yellowish. Briefly, 40 μl sample extract or standard and 260 μl 0.1 mM DPPH solution were added into plate and incubated for 30 min at 25°C. Trolox (0-200 μg/ml) and water were used for standard curve and blank, respectively. The absorbance was measured at 517 nm and the results were expressed as mg Trolox equivalents (TE) per dry weight (mg TE/g) ± SD.
2.4.5 | 2,2'-azinobis-(3-ethylbenzothiazoline-6sulfonic acid) (ABTS) assay Modified ABTS + decolorization assay according to the method of Re et al. (1999) was also conducted to evaluate coffee beans antioxidant capacity. ABTS + dye solution was prepared by mixing 5 ml 7 mM ABTS + solution and 88 μl 140 mM potassium persulfate followed by 16 h incubation in the darkroom. Subsequently, 10 μl sample extract and 290 μl dye solution were added into plate and incubated for 6 min at 25°C. Trolox (0-500 μg/ml) and water were used for calibration curve and blank, respectively. The absorbance was measured at 734 nm and the results were expressed as mg TE/g ± SD.
2.4.6 | Ferric reducing antioxidant power (FRAP) assay FRAP assay was conducted according to the method of Benzie and Strain (1996) with modifications. The FRAP dye solution was prepared in the dark by mixing 300 mM sodium acetate solution, 10 mM TPTZ solution, and 20 mM Fe [III] solution at a ratio of 10:1: Briefly, 20 μl sample extract and 280 μl dye solution were added into plate and incubated for 10 min at 37°C. Trolox (0-200 μg/ml) and water were used for calibration curve and blank, respectively. The absorbance was measured at 593 nm and the results were expressed as mg TE/g ± SD.
2.4.7 | Estimation of hydroxyl radical scavenging activity ( · OH-RSA) Modified Fenton-type reaction method according to the method of Smirnoff and Cumbes (1989) was used for the evaluation of hydroxyl radical scavenging activity of coffee beans. Briefly, 50 μl sample extract, 50 μl 6 mM ferrous sulfate heptahydrate, and 50 μl 6 mM hydrogen peroxide were injected into plate and incubated for 10 min under about 25°C. Subsequently, 50 μl 6 mM 3-hydroxybenzoic acid was added. Trolox (0-400 μg/ml) and water were used for calibration and blank, respectively. The absorbance was measured at 510 nm and the results were expressed as mg TE/g ± SD.
2.4.8 | Estimation of ferrous ion chelating activity (FICA) Modified FICA assay was performed according to the method of Dinis et al. (1994) with some modifications. Briefly, 15 μl sample extract or EDTA standard, 85 μl water, 50 μl 2 mM ferrous chloride, and 50 μl 5 mM ferrozine were injected into plate and incubated for 10 min in the dark at 25°C. EDTA (0-50 μg/ml) and water were used for calibration curve and blank, respectively. The absorbance was measured at 562 nm and the results were expressed as mg EDTA equivalents per dry weight (mg EE/g) ± SD.

| Estimation of reducing power (RPA)
Modified RPA assay according to the method of Ferreira et al. (2007) was used for the evaluation of the reducing power of coffee beans with the color changing from yellow to green. Briefly, 10 μl sample extract, 25 μl 0.2 M phosphate buffer (pH 6.6), and 25 μl 1% potassium ferricyanide (III) solution were injected into plate followed by 20 min incubation at 25°C. Subsequently, 25 μl 10% trichloroacetic acid was added to stop the reaction followed by the addition of 85 μl water and 8.5 μl 0.1% ferric chloride solution, and 15 min incubation at 25°C. Trolox (0-500 μg/ml) and water were used for calibration curve and blank, respectively. The absorbance was measured at 750 nm and the results were expressed as mg TE/g ± SD.

| Characterization of phenolic compounds via LC-ESI-Q TOF-MS/MS
The characterization of phenolic compounds is analyzed via the LC-

| Identification and quantification of volatile compounds by headspace/gas chromatography-mass spectrometry (HS-SPME-GC-MS)
Volatile compounds in coffee ground samples were analyzed by HS-SPME-GC-MS according to the method of Rocchetti et al. (2020).
GC-MS analysis was conducted via a gas chromatograph (6850 series II Network GC System, Agilent Technologies, USA) coupled to an HS-SPME system (PAL RSI I20, Switzerland) and a mass spectrometer (5973Network Mass Selective Detector, Agilent Technologies, USA). Aldrich, USA). The carrier gas was helium with 60 kPa column head pressure. Samples were incubated for 15 min at 60°C, then 15 min extraction and 6 min desorption. The GC oven program was set as follow: 40°C for 5 min followed by an increase to 190°C with the rate of 5°C/min for 8 min; subsequently, the temperature reached 240°C at a rate of 10°C/min and maintained for 10 min. The acquisition was in SCAN mode (35-350 m/z). The solvent delay time was 2 min.
One gram ground coffee sample mixed with 20 μl 100 mg/L 4-Octanol as internal standard was added into vials and then injected as the temperature gradient program above. The linear retention index (LRI) was calculated by alkane standard (C 7 -C 20 ) as the following equation that compares the retention time of one target compound (RT x ) with those of n-alkanes with n and n + 1 carbon eluted before and after the target compound (RT n ): LRI and mass spectrum of volatile compounds detected in coffee samples were compared to the data in the NIST Chemistry WebBook spectrum library (NIST2017) and NIST mass spectra database, respectively. Semi-quantification was conducted by comparing the response area of the target compound and a closely eluted compound with known concentration after LRI and compound MS confirmed.

| Statistical analysis
All results were pure results subtracted by blanking or control values and expressed as mean ± standard deviations (SD) of triple independent analyses. All the statistical analysis was conducted by Minitab 19 (Minitab ® for Windows Release 19, Minitab Inc., Chicago) and GraphPad Prism 9. One-way analysis of variance (ANOVA) and Tukey's honestly significant differences (HSD) were used to verify and analysis the significant differences among samples.

| Phenolic content estimation (TPC, TFC, and tannins content)
The results of TPC, TFC, and TCT for the estimation of phenolic content in the coffee beans were analyzed as shown in Table 1. Overall, all values of the free phenolic compounds were higher than that of the bound, except TCT. There were significant differences (p < .05) shown in the phenolic content of coffee beans with different roasted degrees.
In terms of TPC, the light-roasted coffee beans possessed the highest value with 23.97 ± 0.60 mg GAE/g, followed by medium (22.41 ± 0.58 mg GAE/g) and dark roasted (20.14 ± 0.72 mg GAE/g). It is consistent with previous research that the decreased tendency of the total content of phenolic compounds is along with the intensification of roasting (Cho et al., 2014;Król et al., 2020;Somporn et al., 2011). Polyphenolic compounds, especially chlorogenic acids in coffee beans, performing highly thermal instability, could be directly decomposed with a temperature higher than 80°C which cause the TPC reduction after intensive roasting (Hecimovic et al., 2011;Król et al., 2020). Partial bound phenolic compounds existing in the plant matrix could be liberated during thermal processing by disrupting cellulose constituents (Cho et al., 2014;Mehari et al., 2020;Somporn et al., 2011). With the comparison to bound TPC values, it contributes to the accumulation of free phenolic compounds, which avoids the huge decrease but is relatively higher in the free TPC values. Furthermore, a similar trend and significant differences were shown in the bound TPC value that the total bound phenolic compounds decreased from 19.32 ± 1.29 mg GAE/g to 15.83 ± 1.28 mg GAE/g with the increasing roasting degree. However, the existence of Maillard reaction products, especially melanoidins, has to be considered because of its interaction with Folin-Ciocalteu reagent, which could increase the value of total phenols (TPC) to some extent (Pérez-Hernández et al., 2012). It is necessary to cross-analyze through other antioxidant assays to estimate the properties.
When it comes to TFC and TCT, comparing to the bound phenolic, the results of free phenolic compounds exhibited a reverse trend that the free TFC and TCT values increased from 0.97 ± 0.01 mg QE/g and 1.87 ± 0.23 mg CE/g to 1.16 ± 0.04 mg QE/g and 5.46 ± 0.21 mg CE/g with significant differences when roasting degree increased. Similar results were observed by Hecimovic et al. (2011), Odzakovic et al. (2016 and Król et al. (2020) that the content of total flavonoids and tannins was directly proportional to the roasting degree. More and more bound phenolic compounds were released by the increasing roasting temperature, which improved the free TFC and TCT while reduced that of the bound reasonably. Suitable roasting could degrade condensed tannins into the lower molecular mass of flavonoids, such as anthocyanin, which could improve free TFC value to some extent. However, tannins perform high thermal resistance whose content could be slightly reduced with the temperature lower than 210°C (Ahmad et al., 2018;Van Cuong et al., 2014). Flavan-3-ol is the monomer of condensed tannins which belongs to flavonoids, whereas gallic acid is a component of hydrolyzable tannins belonging to nonflavonoids (Mehari et al., 2020). Thus, various compounds including flavan-3ols complexes, quinolactones, and gallic acids complexes could be formed via the isomerization and polymerization of polyphenolic compounds and interactions with proteins and sugars during thermal processing, which could induce the increase in TPC and TCT (Farah & Donangelo, 2006;Hecimovic et al., 2011;Kim et al., 2011;Król et al., 2020).
TA B L E 1 Determination of phenolic content in coffee beans with three roasting degrees and their antioxidant activity continual intensification of thermal processing so that the proportion of phenolic compounds decreased in the dark-roasted coffee beans (Hecimovic et al., 2011;Król et al., 2020). The influence on the specific phenolic compounds and their changes would be further investigated.

| LC-MS/MS-based characterization of phenolic compounds
Totally, 23 phenolics in three different roasted coffee beans were identified and characterized, which were mainly adopted by 15 phenolic acid and 5 other polyphenols.

Phenolic acids
Five different subclasses of phenolic acids were characterized in the light-, medium-, and dark-roasted coffee beans as shown in Table 2, has also been discovered in roasted coffee beans by Gorecki and Hallmann (2020) and Król et al. (2020). During roasting, chlorogenic acids, as the main component of phenolic fraction in coffee beans, would be hydrolyzed into various aromatic metabolites including salicylic acid (Gorecki & Hallmann, 2020;Król et al., 2020). Therefore, the content of salicylic acid in coffee beans would be increased along with the intensification of roasting within certain temperatures.
Previously, several other derivatives of hydroxybenzoic acids including p-hydroxybenzoic, syringic, and 2,4-dihydroxybenzoic acid were detected in canned coffee drink, honey, and iced tea as well (Shalash et al., 2017;Zeb, 2021).  Note: Ionization mode with ** represents that the compound was detected in both positive and negative modes but only one mode's data were presented. For compounds found in more than one sample, only results for samples with * were shown in the table. Roasted coffee beans samples mentioned in abbreviations are Light roasted "L", Medium roasted "M", and Dark roasted "D." Abbreviation: RT, retention time.

TA B L E 2 (Continued)
Hydroxycinnamic acids. In our study, hydroxycinnamic acids

3-feruloylquinic acid and 3-caffeoylquinic acid, also known as 3-
FQA and 3-CQA, are the two typical isomers of caffeoylquinic acid and feruloylquinic acid, which are the two primary subclass of chlorogenic acids in the coffee beans (Monteiro et al., 2007;Rostagno et al., 2015). As the esterified product of quinic acid and transcinnamic acid derivatives, chlorogenic acids continually degrade along with the roasting period which allows the generation of these phenolic compounds, especially 3-CQA (Rostagno et al., 2015). When it comes to two detected compounds belonging to hydroxyphenylpentanoic and hydroxyphenylacetic acids, there is no detection in coffee beans and drinks in the previous studies.

Other polyphenols
As for other polyphenols, four subclasses in roasted coffee beans have been characterized, which include two hydroxybenzaldehydes, one hydroxycinnamaldehydes, one curcuminoid, and one tyrosol. All five compounds were not detected in the dark-roasted coffee beans in our study.
Compound 18 was tentatively identified as 4-hydroxybenzaldehyde based on the precursor ion at both positive and negative mode with m/z at 121.0295 and confirmed through the product ion at m/z 77, which indicated the loss of CO 2 from the precursor . 4-hydroxybenzaldehyde is one of the major fragrance and flavor components of natural vanilla which is usually used as a flavoring for coffee and chocolate owing to its aromatic properties (Linares et al., 2019). Mahmud et al. (2020) et al., 2020). Curcumin is the major substance of turmeric and also can be detected in the herbal remedy (Sharma et al., 2005;Soleimani et al., 2018). Traditionally, curcumin is used as a spice and coloring agent in most cuisines or for therapeutic applications, such as anti-inflammatory and antimicrobial activities (Mohajeri et al., 2018;Soleimani et al., 2018). Demethyloleuropein (compound 21) plays a significant role as a protector and natural bioactive component and is commonly discovered in olive fruits (Sivakumar et al., 2007). Previously, the track of both curcumin and demethyloleuropein in roasted coffee beans has not been reported.

Lignans and stilbenes
Sesamin ( Sesamin is the primary lignin that could be discovered from sesame seeds and sesame oil. It is also present in flax, barley, buckwheat, millet, oats, rye, nuts, and legumes . Previously, Sesamin has been isolated from sesame by Majdalawieh et al. (2020).
However, the presence of sesamin has not been reported in roasted coffee beans. Resveratrol could also be discovered and isolated from grapes as well (Roat & Saraf, 2017;Sasot et al., 2017).

| Volatile compounds in different roasted coffee beans
The composition of volatile compounds in two types of coffee beans with different roasting degrees analyzed by the HS-SPME-GC-MS method was identified as shown in Table 3. The content of primary volatile compounds in coffee beans all was improved along with the intensive roasting degree, particularly acetic acids, furans, and furanic compounds, and some heterocyclic nitrogen compounds, which is consisted with the previous research (Caporaso et al., 2018;Hertz-Schunemann et al., 2013;Somporn et al., 2011). Acetic acid was the most abundant organic acid in roasted coffee beans after roasting, which is probably because of the fragmentation of saccharides, especially sucrose (Diviš et al., 2019). During roasting, the hydrolysis of sucrose with the evaporation of residual water could produce fructose which could generate 2,3-endiol via Lobryde-Bruynvan-Eckenstein rearrangement. Thermal dehydration of these sugars would form 1-deoxyglucosone as an acid precursor which could induce the formation of acetic acid eventually (Ginz et al., 2000;Yeretzian et al., 2014).
Similar to acetic acid, the content of furans and furanic compounds in coffee beans was significantly improved by roasting degree. Carbohydrates and amino acids are two typical precursors with a relatively high concentration in green coffee beans (Chaichi et al., 2015). During roasting, furans, such as 2-furanmentaol (furfuryl alcohol), are yielded from the reaction between sucrose, ribose, or deoxyosones and amino acids (cysteine or methionine), which could be partially responsible for the caramel aroma of roasted coffee beans (Caporaso et al., 2018;Hertz-Schunemann et al., 2013;Sanz et al., 2002;Somporn et al., 2011). Furanic compounds including furfural and 5-methylfurfural could also contribute to coffee aroma and come from two pathways (Chaichi et al., 2015). One is derived from the dehydration, cyclization, and polymerization of Amadori rearrangement products after the Maillard reaction, especially deoxyribose (Caporaso et al., 2018). Furanic compounds could also be obtained from the thermal oxidation of furfuryl alcohol, polyunsaturated fatty acids, and ascorbic acid (Anese, 2015;Caporaso et al., 2018;Chaichi et al., 2015). Therefore, the content of furfural and 5-methylfurfral could be promoted dramatically by the improvement of furfuryl alcohol formation, when the roasting degree is enhanced from light to dark, which fits the results of this research.
A clustering was found in the group of pyrazines according to Table 3 because 2,5-dimethylpyrazine and 2,6-dimethylpyrazine generate from the same Maillard reaction with different locations of the functional groups (Baggenstoss et al., 2008;Caporaso et al., 2018;Lee et al., 2016). The content of pyrazines was observed generally stable among three roasting degrees except for some fluctuations. Normally, the content of pyrazines reaches a peak when the temperature is around 250°C. Some other researchers assumed that pyrazines would be incorporated into melanoidins while the temperature is above 250°C, which would lead to a reduction in the pyrazines content (Schenker et al., 2002).
Pyrroles and pyridines, typical roasting products identified in this research, showed an increased tendency related to increasing roasting degree. The principle of these two group compounds formation in roasted coffee beans is similar as that of pyrazines. The Strecker reaction between aldoses (aldehydes) and alkylamines (aminoketones) would occur subsequently when other amino acids take part in, followed by heterocyclization and generating a series of aroma active volatile compounds including pyrroles, pyridines, and pyrazines (Caporaso et al., 2018;Hertz-Schunemann et al., 2013).
Pyridine could also be derived from the degradation of trigonelline (Baggenstoss et al., 2008;Hertz-Schunemann et al., 2013). Therefore, the content of pyridine would be reduced by overroasting.
The tendency of phenol content increased with an intensive roasting degree as well in this research. This is probably because of the formation of phenol in roasted coffee beans which is through the TA B L E 3 The content of volatile compounds identified in different roasted coffee beans by HS-SPME-GC-MS degradation of caffeoylquinic acid and ferulic acid originated from the decomposition of chlorogenic acids when roasting is exothermic (Baggenstoss et al., 2008;Caporaso et al., 2018). Generally, suitable roasting with high intensity could improve the content of volatile compounds in coffee beans which is beneficial to the development of coffee beans flavor.  Table 1. Overall, all six assays result of coffee beans with increasing roasting degree showed basically no significant differences statistically except some fluctuations. It agrees with previous research that antioxidant activity would not linearly increase with increasing roasting temperature (Odzakovic et al., 2016;Somporn et al., 2011).

|
DPPH and ABTS primarily have been used for determining antioxidant activity, especially free radical scavenging capacity via hydrogen atom transferring (Du et al., 2021;Górnaś et al., 2015;Nebesny & Budryn, 2003;Sirivibulkovit et al., 2018). For free phenolic compounds, coffee beans with increasing roasting degrees exhibited similar DPPH values around 145 mg TE/g with no significant difference. Interestingly, the bound DPPH value of light-roasted coffee beans was the lowest with 69.98 ± 2.26 mg TE/g, whereas dark roasted showed the highest with 77.39 ± 0.89 mg TE/g. Combined with the changes in the estimation of their phenolic content, the free radical scavenging capacity probably depends more on the flavonoids owing to their structure, such as the 3',4'-dihydroxy system of the B-ring in quercetin (Choi et al., 2002). Regarding previous studies, some novel substances with outstanding antioxidant activities could be generated by Maillard reaction during roasting, such as melanoidins (del Castillo et al., 2002;Odzakovic et al., 2016). This would be the main reason for the maintenance and improvement of antioxidant activities. The values of free DPPH were more than two times that of the bound which is probably because of the releasement of bound phenolic compounds during roasting.  ) and hydroxyl radical which would be scavenged by antioxidants (Chou et al., 2021). During FICA assays, the antioxidants interfere with the formation of ferrous and ferrozine complex by chelating ferrous ions and result in the drop off of the color complex. Hence, the reduction in the color intensity could be in equivalence to its metal chelating activity (Patel, 2013;Santos et al., 2017). In this study, the bound FICA values of all three roasted coffee beans were similar, which were basically maintained along with intensive roasting. However, from light-to dark-roasted degree, the free FICA values slightly increased from 0.37 ± 0.09 to 0.51 ± 0.03 mg EE/g but exhibiting no significant difference statistically.
Generally, the antioxidant activity of coffee beans would decrease along with intensive thermal processing owing to the degradation of polyphenolic compounds (Hecimovic et al., 2011). However, suitable thermal processing could change the structure of existed antioxidants or catalyze the formation of novel antioxidant compounds so that maintain or enhance the antioxidant capacity (Cho et al., 2014;Somporn et al., 2011). For instance, melanoidins, reductive ketones, and other heterocyclic compounds, which are displayed as effective antioxidants, could be formed through Maillard reaction during roasting (Cho et al., 2014;Delgado-Andrade & Morales, 2005).
Moreover, phenylindans and other polyphenol derivatives with high antioxidant capacity could be generated (Hecimovic et al., 2011).
Although lower molecular mass polyphenols were generated via the degradation of high molecular mass polyphenols, such as phenolic acid, improve the overall antioxidant capacity. The metal chelation activity of these high molecular mass polyphenols would be considerably reduced, which is reflected by FICA assays (Cho et al., 2014;Kim et al., 2011).

| Correlation between phenolic compounds and antioxidant potential
Pairwise Pearson's correlation test was performed to evaluate whether the content of phenolic compounds in coffee beans contributed to their related antioxidant activities. The correlation test results were shown in Table 4. TPC was significantly positively correlated with most antioxidant potential estimation assays that the absolute values of the r values for the pairwise correlations were higher than 0.7. With the consideration of a small sample size, the absolute value of the correlation coefficient closer to 1 represents the stronger tendency (Sedgwick, 2012). Therefore, it indicated that phenolic compounds within the coffee beans extract may be the primary constituents responsible for the antioxidant capacity of coffee beans (Wang et al., 2009).
TFC assay performed a significant positive correlation with DPPH, FRAP, and RPA, and negative with FICA. It agreed with previous research conducted by Amin et al. (2013). They stated that the reducing characteristics are commonly related to the presence of reductones which could donate a hydrogen atom and then break the free radical chain to conduct antioxidant action.
Interestingly, TCT exhibited a moderate, negative correlation with other antioxidant assays except for TFC, · OH-RSA, and FICA assays. It could be inferred that roasting could degrade condensed tannins into the lower molecular mass of flavonoids and slightly improve the TFC values, which is consistent with the previous conjecture.
Moreover, it seemed like low degradation of condensed tannins was not enough to maintain the antioxidant capacity with increased roasting temperature.
All assays for the estimation of antioxidant potential showed a significant positive correlation with each other, whereas performed significant negative correlation with · OH-RSA and FICA. The antioxidant in both assays plays a role as the chelating agent. Hence, the hydroxyl radicals scavenging capacity of coffee beans was probably determined more on the novel antioxidant compounds. The metal chelating activity of the high molecular mass of phenolic compounds could be crippled after thermal degradation while still could relieve oxidation via other pathways (Cho et al., 2014).

| CON CLUS ION
According to the current study, it was found that commercial light-roasted coffee beans displayed a relatively higher content of total phenolic compounds (TPC) and antioxidant potential (DPPH, ABTS, FRAP, and FICA). The dark-roasted exhibited higher content of total flavonoids and condensed tannins (TFC and TCT), as well as the better capacity of scavenging hydroxyl radical and reducing power ( · OH-RSA and RPA). Nevertheless, the commercial medium-roasted coffee beans were overall better in all estimation of phenolics content and antioxidant potential. From the advanced LC-ESI-QTOF-MS/MS analytical technique applied for the identification and characterization of the phenolic compounds in roasted coffee beans, a total of 23 phenolic compounds were tentatively identified in our study. Most phenolic compounds were detected in the medium-roasted coffee beans. As for the GC-MS, a total of 20 volatile compounds were identified and quantified in all roasted coffee beans. Generally, the dark-roasted coffee beans performed the highest value of all detected volatile compounds, followed by the medium roasted closely. In conclusion, the content of phenolic compounds in coffee beans would decline along with the intensification of roasting. The antioxidant activities of coffee beans could be at least maintained or improved to some extent even after intensive roasting owing to the generation of novel substances with outstanding antioxidant activity. The medium-roasted coffee beans contain the most various phenolic compounds and relatively outstanding aroma properties.

ACK N OWLED G EM ENTS
We would like to thank Nicholas Williamson, Shuai Nie, and Swati Varshney from the Mass Spectrometry and Proteomics Facility, Bio-21 Molecular Science and Biotechnology Institute, the University of Melbourne, Australia. We are also thankful to the University of Melbourne, Australia, and to Amrit BK, Vigasini Subbiah, and master's students for their moral support.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available in article supplementary material.