Characterization of the Key Aroma Compounds in Chinese Syrah Wine by Gas Chromatography-Olfactometry-Mass Spectrometry and Aroma Reconstitution Studies

The key aroma compounds and the organoleptic quality of two Chinese Syrah wines from the Yunnan Shangri-La region and Ningxia Helan mountain region were characterized. The most important eighty aroma-active compounds were identified by Gas Chromatography-Olfactometry. In both Syrah samples, ethyl 2-methylpropanoate, ethyl 3-methylbutanoate, 3-methylbutyl acetate, 2- and 3-methyl-1-butanol, ethyl hexanoate, ethyl octanoate, 2-phenethyl acetate, methional, 3-methylbutanoic acid, hexanoic acid, octanoic acid, β-damascenone, guaiacol, 2-phenylethanol, trans-whiskylactone, 4-ethylguaiacol, eugenol, 4-ethylphenol, and sotolon were detected to have the highest odor intensities. In the chemical analysis, 72 compounds were quantitated by Stir Bar Sorptive Extraction combined with Gas Chromatography Mass Spectrometry. Based on the Odor Activity Value (OAV), the aromas were reconstituted by combining aroma compounds in the synthetic wine, and sensory descriptive analysis was used to verify the chemical data. Fatty acid ethyl esters, acetate esters, and β-damascenone were found with higher OAVs in the more fruity-smelling sample of Helan Mountain rather than Shangri-La.


Introduction
Wine aroma characteristics originate from grape berries, wine fermentation, and aging techniques [1,2]. Over one thousand volatile and semi-volatile compounds have been identified in wine contributing to its aroma quality [3]. High contributing aroma-active volatiles in wine include alcohols, esters, fatty acids, ketones, terpenes, phenols, and aldehydes. They exist in varying concentrations, from 10 −12 to 10 −4 g/L [4,5]. Volatiles in wine are complex and heterogeneous, and those concentrations that are higher than their odor threshold are considered as potential contributors to wine aroma [6][7][8][9].
Solid Phase Extraction (SPE) based on LiChrolut-EN is a fast and user friendly technique that has been widely utilized to isolate and concentrate volatile compounds in wine [10]. This technique concentrates and purifies the analytes by adsorbing the compounds onto the resin bed and recovering them using a small volume of suitable solvents [11]. Solid Phase Microextraction (SPME) and Stir Bar Sorptive Extraction (SBSE) techniques based on polydimethylsiloxane (PDMS) as the absorbent for nonpolar and intermediate polar compounds are of high selectivity and only require a small amount which were determined to play significant roles in wine aroma quality. In the GC-O study, the odorants in the extracts were described as fruity, floral, green, spicy, tobacco, woody, potato, nutty, herb, brandy, sweaty, and so on.
Some studies reported links of odorants with sensory attributes in wine which showed positive correlations between fruity (berry or dried fruit) and ethyl ester compounds [21,26], woody aromas, and volatile phenols [27]. Straight and branched chain fatty acids such as propionic acid, 3-methylbutyric acid, hexanoic acid, and octanoic acid in wine (10 −6 g/L level concentration) were considered as unpleasant odorants, but they contribute significantly to the complexity of the wine total flavor [28]. C 13 -norisoprenoid compounds β-damascenone (fruity-flowery smelling) and β-ionone (violet smelling) were reported frequently in premium wine [29,30]. To understand the correlations between volatile compounds and aroma attributes, two samples are comparatively discussed in the GC-O section, quantitative analysis section, and aroma reconstitution test section. Compared with YS, NS showed higher intensities in odorants: 3-methylbutyl acetate, ethyl hexanoate, ethyl octanoate, methional, benzaldehyde, ethyl decanoate, guaiacol, trans-whiskylactone, γ-decalactone, sotolon, and ethyl vanillate. These compounds mainly contribute to the fruity, nutty, and clove characters of wine. In contrast, compounds 1,1-diethoxyethane, 2-methylpropanoic acid, α-terpineol, octanoic acid, and m-cresol in NS were detected to have lower odor intensities than that in YS.

Quantitative Analysis of Aroma-Active Compounds
To gain insight into the aroma characteristics of the two Syrah wines, 72 important compounds (shown in Table 2), from the GC-O data and from the literature [23,31] were quantitated and their odor activity values (OAVs) were calculated. As expected, volatiles were detected at levels ranging from ng/L to mg/L. The highest concentration in both wines was found for isoamyl alcohol (123 mg/L for NS and 106 mg/L for YS), followed by ethyl acetate, isobutyl alcohol, propanol, diethyl succinate, 2-phenethanol, and acetaldehyde. Since those fermentation byproducts have a relatively high odor threshold (especially diethyl succinate of 120 mg/L and 2-phenethanol of 14 mg/L), they were usually not necessarily considered as high aroma contributors to wine. Among the 72 quantitated compounds, in the Ningxia Syrah wine, 18 compounds reached concentrations higher than their odor threshold, whereas in the Shangri-La Syrah, 19 compounds were found to have an OAV > 1. Those aroma-active compounds (23 compounds in NS and 24 compounds in YS) with OAV between 0.1 and 1 may also contribute to the formation of wine aroma due to the interaction effect in the wine matrix. In the wine from the Ningxia Helan Mountain region, these fusel alcohols were at higher concentrations than that in YS.
According to the odor threshold, the highest OAV in the Syrah from the Ningxia region was determined for β-damascenone (tobacco and burnt sugar smelling, OAV 56.1), followed by the fruity-like ethyl octanoate (OAV 54.9), and acetaldehyde (fruity and musty or pungent smelling, OAV 44.7). The fruity straight and branched esters ethyl isovalerate (OAV 43.8), ethyl hexanoate (OAV 38.1), ethyl isobutyrate (OAV 17.6), and isoamyl acetate (OAV 13.6) also had high OAVs are therefore very significant aroma contributors in NS. Those ester compounds produced by yeasts during fermentation are well known to contribute to and enhance the fruity aroma of wines [32]. β-Damascenone and β-ionone mainly come from the degradation of carotenoids in grapes, and are well known as important characteristic flavor compounds in wine, which were characterized by "fruity-flowery" and "violet" notes [33,34]. They are important C 13 -norisoprenoids due to their low odor threshold and high contribution to wine aroma complexity [29,30]. Fatty acids are the precursors of esters, terpenes, and alcohols in grapes and wine that mainly contribute to the flavor of wine [32]. Butanoic acid, 2and 3-methylbutyric acid, hexanoic acid, and octanoic acid that give a cheesy or sweaty smell and contribute complexity to wine also had concentrations higher than their thresholds. Volatile acids in wine are important components in flavor quality; they impart woody, brandy, almond, etc., pleasant aromas to wine in proper concentration levels, while high concentrations (above 20 mg/L) of acids will give a negative aroma hints of wine [20,21]. Oak-derived phenolic compounds guaiacol (OAV 1.25), eugenol (OAV 1.12) and ethyl dihydrocinnamate (OAV 1.05) were also found in concentrations above their thresholds, and they were reported as important aroma potentials in aged wine. Most of these compounds are formed through hydrolysis of glycosidically conjugated forms during wine aging, or are extracted from oak during barrel aging, contributing floral and cherry notes to wine [3][4][5]35].
In the Syrah wine from the Shangri-La region, compounds with the highest OAVs were mostly similar to those in NS. Unlike Syrah from the Ningxia region, acetaldehyde (OAV 63.8) was the highest in YS. The fruity-smelling ester compounds ethyl isovalerate (OAV 27.3), ethyl octanoate (OAV 11.7), ethyl hexanoate (OAV 10.8), ethyl isobutyrate (OAV 10.9), and isoamyl acetate (OAV 8.1) in YS were detected to have much lower OAVs compared with those in NS. The high concentration of acetaldehyde and lower concentration of esters in YS were supposed to be generated during relatively high alcohol fermentation temperatures. According to the record profile of the winery, the atmospheric temperature during the alcohol fermentation was higher than usual without a cooling system. Molina et al. [36] and Culleré et al. [37] reported that higher wine fermentation temperature interferes with the reactions of esters and increases the formation of acetaldehyde. Besides, the addition of SO 2 before alcohol fermentation may affect the formation of acetaldehyde in wine. The Syrah wine from the Shangri-La region in this study was determined to have a lower concentration of esters and lower intensities of fruity odors. Mayr et al. [23] determined two Australian Syrah wines. The wine from the warmer Barossa Valley was found to have higher concentrations of ethyl propionate and oak-derived compounds, whereas the cooler Margaret River Syrah had above threshold concentrations of 2-and 3-methylbutanoic acid, as well as rotundone. β-Damascenone (OAV 24.4) was found with a lower odor activity value in YS than NS, as well as the fatty acids 2-and 3-methylbutyric acid (OAV 11.8). Table 2.
The oak-derived compounds guaiacol (OAV 1.58) and eugenol (OAV 1.19) were detected to have similar concentrations in two Syrah wines. Compared with those in NS, ethyl acetate (OAV 6.62), trans-whiskylactone (OAV 1.82), vanillin (OAV 0.5-1), and phenolic compounds (all with OAV 0.5-1): 4-methylguaiacol, 4-ethylguaiacol, isoeugenol, 4-ethylphenol, and 3-ethylphenol were found to have higher odor activity values in YS. Fusel alcohols including propanol, isobutanol, 2-and 3-methyl-1-butanol, 1-hexanol, benzyl alcohol, and 2-phenethanol in wine are mainly formed through alcoholic fermentation from sugar and amino acid catabolism. They might show either positive or negative impacts on total aroma depending on the concentration level. It was reported that a level of higher alcohols below 300 mg/L confers a desirable complexity to the wine whereas concentrations over 400 mg/L have a negative effect [38]. In this study, both wines had fusel alcohols below 300 mg/L, which contributed positively to the wine quality. 2-Phenylethanol that gave wine a positive rose aroma was also reported as a significant compound. In NS and YS wine, linalool and rose oxide have OAV between 0.1 and 0.5; those terpenoids are responsible for the odor of citrus (lemon) and floral aromas in wine. Furans and lactones with low odor thresholds in wine were always reported as the typical compounds in the aging wine. In this study, the low OAV of this group of compounds was consistent with the young wine vintage. Methoxypyrazine compounds 3-isopropyl-2-methoxypyrazine (IPMP), 2-sec-butyl-3-methoxypyrazine (SBMP), and 3-isobutyl-2-methoxypyrazine (IBMP) contributing green bell pepper or bean notes to wine had very low odor thresholds (IBMP, 2 ng/L). The concentrations of methoxypyrazines in the sample wines did not reach the threshold, but they might be potential important odorants in wine due to the synergistic effect of the interaction between compounds [7,39].

Sensory Evaluation of Syrah Wine Samples and Aroma Reconstitution Test
Sensory descriptive analysis was performed to compare the aroma differences of the Syrah wine samples NS and YS, and to relate aroma compounds with sensory attributes. As shown in Figure 1, the two Syrah wines had similar intensities of floral, black pepper, green or bell pepper, sour, caramel, smoky, woody, spicy, and rancid or cheesy attributes. NS had a higher score of fruity (3.5) and berry (3.2) terms (p < 0.05) than that of YS (2.7 and 2.5, respectively), which could be explained by the concentration differences of the esters with high OAVs and their synergetic interactions. For other attributes, the two wines showed similar results: the panelists could not detect the difference. Figure 1 illustrates that fruity, berry, floral, sour, caramel, smoky, and woody terms showed lower intensities in reconstitution wine than that in the wine samples. the threshold, but they might be potential important odorants in wine due to the synergistic effect of the interaction between compounds [7,39].

Sensory Evaluation of Syrah Wine Samples and Aroma Reconstitution Test
Sensory descriptive analysis was performed to compare the aroma differences of the Syrah wine samples NS and YS, and to relate aroma compounds with sensory attributes. As shown in Figure 1, the two Syrah wines had similar intensities of floral, black pepper, green or bell pepper, sour, caramel, smoky, woody, spicy, and rancid or cheesy attributes. NS had a higher score of fruity (3.5) and berry (3.2) terms (p < 0.05) than that of YS (2.7 and 2.5, respectively), which could be explained by the concentration differences of the esters with high OAVs and their synergetic interactions. For other attributes, the two wines showed similar results: the panelists could not detect the difference. Figure  1 illustrates that fruity, berry, floral, sour, caramel, smoky, and woody terms showed lower intensities in reconstitution wine than that in the wine samples.

Wines
Two popular commercial Syrah wines of vintage 2014 were kindly donated by wineries. NS was from the Ningxia region in northwest China. The basic composition of the wine was as follows: alcohol 14.1% v/v, pH 3.48, titratable acidity (TA, as tartaric acid) 6.78 g/L, volatile acidity (VA, as acetic acid) 0.78 g/L, total SO2 64 mg/L, glucose + fructose (G + F) 1.2 g/L. YS was from a high altitude region, Yunnan Shangri-La in southwest China. The basic compositions: alcohol percentage of 13.6% by volume, pH 3.56, TA 6.51 g/L, VA 0.51g/L, total SO2 91 mg/L, G + F 2.0 g/L. Both wine samples were fermented in stainless steel fermenters at 20-32 °C, and the malolactic fermentation was performed ten days after the alcohol fermentation. Both wines were aged in the same type of France oak for half a year and bottled. Once received, the wine samples were stored in a 4 °C controlled temperature room.

Reagents and Chemical Standards
Chemical standards of the compounds in this study were purchased from Sigma-Aldrich (

Wines
Two popular commercial Syrah wines of vintage 2014 were kindly donated by wineries. NS was from the Ningxia region in northwest China. The basic composition of the wine was as follows: alcohol 14.1% v/v, pH 3.48, titratable acidity (TA, as tartaric acid) 6.78 g/L, volatile acidity (VA, as acetic acid) 0.78 g/L, total SO 2 64 mg/L, glucose + fructose (G + F) 1.2 g/L. YS was from a high altitude region, Yunnan Shangri-La in southwest China. The basic compositions: alcohol percentage of 13.6% by volume, pH 3.56, TA 6.51 g/L, VA 0.51g/L, total SO 2 91 mg/L, G + F 2.0 g/L. Both wine samples were fermented in stainless steel fermenters at 20-32 • C, and the malolactic fermentation was performed ten days after the alcohol fermentation. Both wines were aged in the same type of France oak for half a year and bottled. Once received, the wine samples were stored in a 4 • C controlled temperature room.

Reagents and Chemical Standards
Chemical standards of the compounds in this study were purchased from Sigma-Aldrich (St. Louis, MO, USA), TCI America (Portland, OR, USA), EKC Inc. (Rosemont, IL, USA), Alfa Aesar (Ward Hill, MA, USA), and EMD Chemical Inc. (Gibbstown, NJ, USA), and their purities were >90% in all cases. The details are shown in a supplementary table. Milli-Q quality water was obtained from a Milli-Q purification system (Millipore, Boston, MA, USA). Methanol (HPLC grade) was from EM Science (Gibbstown, NJ, USA). Acetonitrile (HPLC grade) was bought from Omnisolv (McLean, VA, USA). Dichloromethane (HPLC grade) from Burdick & Jackson (Muskegon, MI, USA) was freshly distilled before use. Tartaric acid was from Mallinckrodt Inc. (Paris, France). Anhydrous sodium sulfate and sodium chloride (99.9%, ACS certified) were supplied by Mallinckrodt Baker. The LiChrolut-EN cartridge with a two-gram absorbent bed was packed in the laboratory.
Standard stock solutions were prepared in methanol individually except that fatty acids were in acetonitrile. Internal standard solution (IS) of 3,4-dimethylphenol and 4-octanol, was prepared at a concentration of 50 ppm in methanol, separately. All the solutions were stored in dark bottles at −20 • C.

LiChrolut-EN-SPE and Solvent-Assisted Flavor Evaporation (SAFE)
The LiChrolut-EN cartridge was made by packing one-gram of resin into a 15 mL, 2 cm diameter reservoir (Thermo Scientific). The cartridge was conditioned with 10 mL of dichloromethane, air-dried, and then eluted with 10 mL of methanol, and finally washed with 10 mL of Milli-Q water. Two hundred and fifty milliliters of wine sample was percolated through the LiChrolut-EN under vacuum at 3 mL per min. The column was then washed with 10 mL of water and eluted with 20 mL of dichloromethane. The eluent was passed through the SAFE device (BAENG; Glasbläserei Bahr, Manching, Germany) to remove the nonvolatile constituents at 50 • C under a vacuum of 2.80 × 10 −3 torr. The distillate was concentrated in a Kuderna-Danish concentrator in 40 • C water bath with a Vigreaux column to approximately 5 mL. After drying over anhydrous sodium sulfate and transferring, the extract was further concentrated to 0.5 mL with a gentle stream of nitrogen.

GC-MS-Olfactometry Analysis
The GC-O analyses were performed on an Agilent 6890 GC (Agilent Technologies Inc., Santa Clara, CA, USA) equipped with an Agilent 5973 mass selective detector and a Gerstel Olfactory Detection Port (ODP). One microliter concentrated extract was injected in split mode (split ratio 1:10) and separated by a ZB-Wax column (30 m × 0.25 mm i.d., 0.50 µm film thickness, Phenomenex, Torrance, CA, USA). Helium was used as the carrier gas at a constant flow rate of 2.5 mL/min. At the exit of the capillary column, the effluents were split 1:1 (by volume) into a sniffing port and a MS detector. The GC injector and ODP temperature were both set at 250 • C. The oven temperature was programed at 40 • C for a 4 min holding and ramped up to 100 • C at a rate of 4 • C/min, then 3 • C/min to 230 • C with a 10 min holding. The MS transfer line and ion source temperature were 250 • C and 230 • C, respectively. Electron ionization mass spectrometric data from m/z 35-350 were collected using a scan rate of 5.27/s, with an ionization voltage of 70 eV. Three experienced panelists (two females and one male, with over 30 h of training) were selected for the GC-O analysis. The retention time, odor descriptor, and its intensity (5-point scale from 1 to 5 represent very weak, weak, moderate, strong, very strong) were recorded. Each sniffing session lasted 1 h and the panelists smelled each extract sample twice. The average intensity of the descriptors was calculated. A standard mixture of n-alkane C 5 -C 30 was prepared and analyzed under the same GC conditions above. Retention Indices (RI) were calculated in accordance with a modified Kovats method based on individual retention times from the lab using pure reference compounds. Mass spectra of unknown compounds were compared with those present in the Wiley 275.L database (Agilent Technologies Inc.).

Static Headspace-GC-FID Analysis
Acetaldehyde, ethyl acetate, propanol, isobutyl alcohol, isoamyl acetate, and isoamyl alcohol were quantitated using the static headspace-GC-FID method described previously [46] due to their high concentrations in the sample. A Varian CP 3800 gas chromatograph equipped with a flame ionization detector (Varian, Inc., Palo Alto, CA, USA) was used. One milliliter of wine was added into a 20 mL auto sampler vial and 20 µL of internal standard (IS, 5 mg/L methyl propionate in methanol) was spiked. Samples were equilibrated at 70 • C for 15 minutes with agitation at 500 rpm. One thousand microliters of the headspace sample was injected using a heated (70 • C) gastight syringe (2.5 mL) in split mode 10:1. Separation was performed by a DB-FFAP capillary column (30 m × 0.32 mm i.d., 0.5 µm film thickness, Agilent Technologies). Helium was used as the carrier gas at a constant rate of 1.5 mL per minute. The oven temperature was set at 35 • C for 4 min holding, raised to 150 • C at a rate of 10 • C/min, and held at the final temperature for 5 min. The injector and detector temperature were both set at 250 • C, respectively. A standard calibration curve was prepared by spiking known amounts of standards into one mL of synthetic wine (12% ethanol (v/v), 3.5 g/L tartaric acid, pH 3.5) and 20 µL of IS (5 mg/L methyl propionate) was added. Data were collected by the Varian Star workstation. Standard curves and sample concentrations were calculated using interactive graphics.

SBSE-GC-MS Analysis
For those compounds having low concentration or low affinity, the quantitation was conducted by the SBSE-GC-MS technique as described previously [47,48]. The comparison of two Twisters (PDMS and EG) is shown in the supplementary materials. A preconditioned PDMS coated stir bar (Twister) (10 mm × 0.5 mm, Gerstel Inc., Linthicum, MD, USA.) or an EG coated stir bar (0.5 mm film thickness, 10 mm length, Gerstel Inc.) was used to extract the aroma compounds. Ten milliliters of wine was pipetted into a 40 mL glass vial and diluted with 10 mL of saturated NaCl solution, and 20 µL of IS (50 ppm 4-octanol for PDMS stir bar set, 50 ppm 3,4-dimethlyphenol for EG stir bar set) solution was added. A PDMS or EG stir bar was then placed into the vial and stirred for 3 h at 1000 rpm at room temperature. After extraction, the stir bar was picked up from the vial, rinsed with Milli-Q water, dried with a Kimwipe, and transferred into a thermal desorption unit (TDU) for GC-MS analysis. Each sample was analyzed in triplicate. Analysis of the absorbed volatile compounds were performed on an Agilent 7890 GC coupled with a 5975 mass selective detector, and a Gerstel MPS-2 multipurpose TDU auto sampler with a CIS-4 cooling injection system (Gerstel Inc.). The analytes were thermally desorbed at the TDU in splitless mode. The CIS-4 was cooled to −80 • C with liquid nitrogen during the sample desorption, and then heated at 10 • C/s to 250 • C and held for 10 min for the PDMS bar, or to 220 • C for the EG bar. Solvent vent mode was used during the injection with a split vent flow of 50 mL/min. A ZB-WAX capillary column (30 m × 0.25 mm i.d., 0.5 µm film thickness, Phenomenex, Torrance, CA) was used. The oven temperature program was set at 40 • C for 4 min, raised to 230 • C at 4 • C/min, and held for 15 min. A constant helium flow of 2 mL/min was used. The MS transfer line and ion source temperatures were 280 • C and 230 • C, respectively. A standard calibration curve was prepared by spiking known amounts of standards into 10 mL of synthetic wine and 10 mL of saturated NaCl solution with 20 µL of IS. Each sample was analyzed in triplicate and the results were calculated through Chemstation software (v.10.1) (Agilent Technologies).

Sensory Evaluation of Wines and Aroma Reconstitution Test
The sensory evaluation of the two wine samples was conducted based on the method by Tao et al. [49]. The panel was trained over 50 days using a "Le Nez du Vin" aroma kit and 11 judges (4 males and 7 females) were selected. The Quantitative Descriptive Analysis (QDA) was performed to pick up the most important terms to describe the aroma characters of the wine samples in this study. The top (high MF values) 11 descriptors: fruity, berry, floral, black pepper, green or bell pepper, sour, caramel, smoky, woody, spicy, and rancid or cheesy were selected to describe the aroma of Syrah wine in this study. Sample wines in a balanced and completed block design were presented (in triplicate) to the panelists. They were required to use the 5 to 6 most significant terms as listed in Li et al. [37] to describe the wine aroma. Panelists were also asked to score the intensity of each term using a 5-point scale: (0) not detected; (1) weak, hardly recognizable note; (2) clear, but weak; (3) clear but not an intense note; and (4) intense note. The data processed were a mixture of intensity and frequency of detection ("modified frequency", MF), which was calculated with the formula proposed by Tao et al. [49,50]: MF = F(%)I(%) F (%) is the detection frequency of an aromatic attribute expressed as a percentage; I (%) is the average intensity expressed as a percentage of the maximum intensity.
The aroma reconstitution test was conducted as described by Chen et al. [18] to resemble the wine according to the quantitative analysis and OAVs. Compounds with high OAVs (≥0.5) were reconstituted with the concentration in the matrix of synthetic wine (14% ethanol (v/v), 3.5 g/L tartaric acid, pH 3.5), labeled as RN and RY for NS and YS, respectively.

Statistical Analysis
The volatile compounds identified and quantified were listed in the table which was formed by Microsoft Office Excel 2013, and so was the mean value, standard deviation, and the OAVs. The concentration differences of volatiles between samples were determined using one-way analysis of variance (ANOVA) established by the Student's t test at a significance level of ≤0.05, carried out using SPSS 20.0 (IBM, Armonk, NY, USA).
By comparing the odor intensities in GC-O, OAVs in the quantitative analysis, and scores of descriptors in the sensory evaluation, a more thorough understanding of the correlation of compounds and aroma was formed. Important aroma contributors: ethyl 2-methylbutyrate, ethyl hexanoate, ethyl octanoate, ethyl decanoate, 3-methylbutyl acetate, guaiacol, eugenol, and decanoic acid were detected at higher odor intensities in the sample Syrah extract from Helan Mountain as compared to Shangri-La. The quantitative analysis and OAV study also showed that most of the ethyl esters and acetate esters were found with higher OAVs in the Syrah wine of Helan Mountain rather than the wine from Shangri-La. It also should be noticeable that the Syrah wine from Ningxia in this study showed stronger fruity characteristics in the sensory evaluation study, which is consistent with the GC-O and quantitative analysis.