Application of Attenuated Total ReflectanceâFourier Transformed

Subscriber access provided by LAURENTIAN UNIV

Article

Application of Attenuated Total Reflectance-Fourier Transformed Infrared (ATR-FTIR) Spectroscopy to Determine Chlorogenic acid Isomer Profile and Antioxidant Capacity of Coffee Beans Ningjian Liang, Xiaonan Lu, Yaxi Hu, and David Kitts J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jafc.5b05682 • Publication Date (Web): 02 Jan 2016 Downloaded from http://pubs.acs.org on January 3, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Agricultural and Food Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 37

Journal of Agricultural and Food Chemistry

1

Application of Attenuated Total Reflectance-Fourier Transformed Infrared

2

(ATR-FTIR) Spectroscopy to Determine Chlorogenic acid Isomer Profile and

3

Antioxidant Capacity of Coffee Beans

4 5

Ningjian Liang, Xiaonan Lu, Yaxi Hu, and David D. Kitts *

6 7

Food, Nutrition, and Health Program, Faculty of Land and Food Systems, The University

8

of British Columbia, 2205 East Mall, Vancouver, British Columbia, V6T 1Z4, Canada.

9 10

*

11

[email protected])

Corresponding

author

(Tel:

+16048225560;

Fax:

+16048225143;

Email:

12 13 14 15 16 17 18 19 20 1

ACS Paragon Plus Environment


21

Page 2 of 37

Abstract

22

The chlorogenic acid isomer profile and antioxidant activity of both green and

23

roasted coffee beans is reported herein using ATR-FTIR spectroscopy combined with

24

chemometric analyses. High performance liquid chromatography (HPLC) quantified

25

different chlorogenic acid isomer content for reference, while ORAC, ABTS, and DPPH,

26

were used to determine the antioxidant activity of the same coffee bean extracts. FTIR

27

spectral data and reference data of 42 coffee bean samples were processed to build

28

optimized PLSR models, and 18 samples were used for external validation of constructed

29

PLSR models. In total, six PLSR models were constructed for six chlorogenic acid

30

isomers to predict content, with three PLSR models constructed to forecast the free

31

radical scavenging activities, obtained using different chemical assays. In conclusion,

32

FTIR spectroscopy, coupled with PLSR serves as a reliable, non-destructive and rapid

33

analytical method to quantify chlorogenic acids and to assess different free radical

34

scavenging capacities in coffee beans.

35 36

Keywords

37

Coffee, FTIR, Chlorogenic acid Isomer, Antioxidant Activity, chemometrics

38

39

2


Page 3 of 37

40


Introduction

41

Chlorogenic acids are important dietary phenolic compounds formed from the

42

esterification of trans-cinnamic and quinic acids.1 Coffee is a popular source of

43

chlorogenic acid in the human diet, accounting for up to 0.5-1.0 gram per day consumed

44

by typical coffee consumers.2 Green coffee beans contain approximately 12% of the dry

45

weight;3 however, with roasting chlorogenic acid losses occur along with the removal of

46

water which corresponds to the generation of high molecular weight browning products

47

(e.g. melanoidins) due to the onset of the Maillard reaction. As a consequence, one cup

48

of coffee can deliver a wide range of chlorogenic acids (70 to 300 mg), depending on the

49

source of the coffee bean and the time-temperature processing that was used to generate

50

different roasts.4 Total chlorogenic acids refer to the mixture of different specific isomers,

51

depending on number and position of the acyl residues. Major chlorogenic acid isomers

52

in coffee beans have been identified and include both mono-caffeoylquinic acids [e.g.

53

3-caffeoylquinic acid (1 in figure 1), 4-caffeoylquinic acid (2 in figure 1), and

54

5-caffeoylquinic

55

3,4-dicaffeoylquinic acid (5 in figure 1), 3,5-dicaffeoylquinic acid (4 in figure 1), and

56

4,5-dicaffeoylquinic acid (6 in figure 1)], the latter group present in relatively lower

57

concentrations.5 In addition, there are many more minor chlorogenic acid isomers

58

including feruloylquinic acids, p-coumaroylquinic acids, caffeoylferuloylquinic acids,

acid

(3

in

figure

1)]

and

di-caffeoylquinic

acids

[e.g.

3



Page 4 of 37

6

59

diferuloylquinic acids, di-p-coumaroylquinic acids, dimethoxycinnamoylquinic acids;

60

however, their presence in the coffee bean at only trace amounts makes routine

61

quantitation difficult and impractical.

62

Chlorogenic acid concentration and isomer composition contribute to the sensory

63

characteristics of coffee beans, namely the final acidity, astringency, and bitterness of the

64

coffee beverage. Chlorogenic acids also have a role in the formation of pigments, aroma,

65

and flavor of coffee beans during the roasting process, as a precursor for the Maillard

66

reaction.7 Lipid oxidation in roasted coffee beans has been attributed to loss of quality

67

attributes, such as aroma, when beans are exposed to conditions that are susceptible to

68

oxidation.8, 9 It is conceivable that antioxidant components present in coffee beans are

69

important in protecting against deterioration caused by lipid oxidation. Studies that

70

reported chlorogenic acid isomers have different capacities to scavenge free radicals in

71

vitro and are differentially absorbed and metabolized in humans further support that a

72

comprehensive analyses of the isomeric composition of chlorogenic acid in coffee is

73

important.10, 11 The value of antioxidants present in coffee beans has also been linked to

74

the fact that the frequency of coffee consumption for regular coffee drinkers has resulted

75

in the recognition that coffee is a major source of dietary antioxidants.12 Associated with

76

this positive aspect is the inverse relationship between coffee consumption and the risk

77

of Type-2 diabetes and cardiovascular disease.13 Thus, the chlorogenic acid composition

4


Page 5 of 37


78

in coffee could be a valued chemical biomarker to estimate both the sensory quality of

79

coffee and moreover, the overall health benefits attributed to coffee consumption.

80

The standard method for separation and quantification of chlorogenic acid isomers

81

includes the use of HPLC with tandem photodiode array detection.5,14 Although

82

advantages of HPLC for routine chlorogenic acid quantitation include high accuracy and

83

precision, this procedure is destructive to the sample and to some degree, will require

84

relatively longer time to complete the analysis compared to FTIR methods.

85

A battery of different chemical free radical scavenging assays has been

86

recommended to demonstrate antioxidant activity of soluble constituents present in

87

either hydrophilic or hydrophobic phases, which respond to underlying mechanisms

88

defining specific radical scavenging activity.13,

89

that coffee constituents can display antioxidant activity by sequestering potential

90

prooxidants,17 more recent characterization of antioxidant activity of coffee has come

91

from chemical free radical assays, such as oxygen radical absorbance capacity (ORAC)

92

assay,

93

2,2-diphenyl-1-picrylhydrazyl (DPPH) assay. Different free radicals are involved in

94

these assays and different mechanisms of reactions will depend on the chemical (e.g.

95

phenolic vs non-phenolic) composition of coffee. No single method has been agreed

96

upon for routine analysis. It is challenging for the coffee industry to perform a battery of

97

antioxidant assays due to the complexity and time required to conduct multiple assays.

2,2′-azino-bis

15, 16

(3-ethylbenzothiazoline-6

While former studies have shown

sulphonic)

(ABTS)

assay,

and

5



Page 6 of 37

98

The development of rapid methods for the estimation of chlorogenic acid isomer

99

composition, and in association, antioxidant capacity in coffee beans, could have value

100

for quality control monitoring.

101

FTIR spectroscopy coupled with multivariate statistical analysis offers the potential

102

to facilitate real-time measurement of critical quality attributes. FTIR spectroscopy has

103

been used as a less destructive and high-throughout method for coffee quality parameter

104

analysis. Some examples of applying FTIR spectroscopic method in coffee include

105

estimating caffeine content in coffee,4 distinguishing between defective and

106

non-defective coffee beans, and distinguishing between Arabica and Robusta coffee

107

varieties.

108

antioxidant capacity of various foods.20,

109

accessories, such as attenuated total reflectance cells, provides simplified handling and

110

enables samples to be examined directly in their original state. The main challenges of

111

applying ATR-FTIR spectroscopy for the determination of specific properties of

112

biological materials are to identify suitable spectral pre-treatment and calibration

113

strategies.

114

critical for developing spectroscopic models. After pretreatment, multivariate statistical

115

analysis, such as partial least squares regression (PLSR), is commonly used to establish

116

regression models that can confirm spectral data with reference analyte quantification,

117

determined by conventional HPLC methods.22, the objective of our current research was

18, 19

19

FTIR spectroscopy has also been successfully used to determine 21

The development of FTIR spectrometer

Pre-treatment of spectra, such as baseline correction and filtering, are both

6


Page 7 of 37


118

to use ATR-FTIR spectroscopy combined with PLSR analysis as an analytical platform

119

for rapid and accurate determination of chlorogenic acid isomer composition and

120

antioxidant activity in both green and roasted coffee beans.

121 122

Materials and Methods

123

Coffee bean samples. Three batches each of green coffee beans obtained from five

124

locations, namely Sumatra, Dominica, Peru, Ethiopia, and Papua New Guinea (PNG),

125

were individually combined, and three sub-samples were subsequently collected for each

126

roast. Coffee beans were roasted at 210 °C for 12 min, 223 °C for 14 min, or 235 °C for

127

15 min to produce light, medium and dark coffee roasts, respectively. Moisture loss

128

ranged from 15-18% for different coffee roasts. Green and roasted coffee beans were

129

ground to a powder and passed through a 0.5 mm sieve to obtain uniform particle size.

130

Extraction of chlorogenic acids from coffee beans. Aqueous methanol (70%) was

131

added to 1 gram of ground coffee powder to produce a final concentration of 60 mg/mL;

132

this solution was extracted at 20 °C for 6 h in the dark using a shaker set at a speed of 250

133

rpm. The mixture was centrifuged at 2900×g for 5 min and the pellet was extracted with

134

two more volumes of 70% methanol at 20 °C for 6 h; thus each sample was extracted

135

three times. The supernatants from three extractions were pooled and the final volume

136

was adjusted to 50 mL. The methanol extracts were stored at -80 °C for future HPLC

137

analyses and antioxidant activity analyses. 7



Page 8 of 37

138

Reference HPLC method for the determination of chlorogenic acid profile. Coffee

139

bean extracts from each coffee roast were filtered through 0.2-µm nylon syringe filters

140

prior to analysis using HPLC to quantify chlorogenic acid isomer content. An 1100 series

141

HPLC (Agilent, Palo Alto, CA) equipped with a binary pump, auto-sampler with

142

thermostat control set at 30 °C, and a diode array detector was used. HPLC procedures

143

followed a previous method with some modifications.23 Chromatographic separation was

144

carried out using a column (250 mm × 4.6 mm i.d., 5 µm, Agilent RP-18) with gradient

145

mobile phase A (10 mM citric acid) and B (100% methanol) set at a flow rate of 1 ml/min.

146

The elution gradient was: 0-5 min, 85% A; 5-40 min, 85%-60% A; 40-45 min, 60% A,

147

45-50 min, 60%-85% A. The diode array detector was set at 325 nm for chlorogenic acid

148

isomers. The sample injection volume was 10 µL. Standard solutions of pure chlorogenic

149

acid isomers (Chengdu Must Bio-Technology Co., Ltd, Chengdu, China) were prepared

150

in 70% aqueous methanol and analyzed by HPLC to build calibration curves. The limit of

151

quantification was defined as the minimum concentration at the signal-to-noise ratio (S/N)

152

of 10. All of the analyses were run three times.

153

Reference DPPH method for the determination of antioxidant activity. Scavenging

154

capacity of DPPH (Sigma-Aldrich, St. Louis, MO) free radical by coffee bean extracts

155

was determined using a previously modified method.24 DPPH in ethanol at the

156

concentration of 1 mM was mixed with a range of concentrations of coffee bean

157

extract/Trolox and this was kept at 20 °C for 30 min. The absorbance reading at 517 nm 8


Page 9 of 37


158

was recorded using a Multiskan Spectrum spectrophotometer (ThermoLabsystems,

159

Helsinki, Finland). The DPPH radical scavenging activity was calculated as a percentage

160

of inhibition (% inhibition):

161

% Inhibition=[(Abcontrol - Absample)/Abcontrol]×100

162

where Abcontrol is the absorbance of the negative control which does not contain

163

coffee/Trolox; Absample is the absorbance of the reaction mixture containing coffee/Trolox.

164

Trolox was used as the reference in the current study. The % inhibition vs. concentration

165

of coffee brew extract was plotted to obtain a regression equation. The ratio between the

166

slope of coffee sample regression equation and the slope of Trolox regression equation

167

was defined as DPPH quenching capacity of coffee brew samples. All analyses were run

168

three times in triplicate. The DPPH scavenging capacities of coffee bean extracts were

169

expressed as mmol equivalents of Trolox per gram of coffee beans (dry matter).

170

ABTS method for the determination of antioxidant activity. The ABTS assay was

171

followed according to the procedure described previously with modifications.25,26 Briefly,

172

ABTS radical cations were generated by mixing 7 mM ABTS in distilled water with 2.45

173

mM potassium persulfate in distilled water. The mixed solution was kept in the dark at

174

20 °C for 16 h. An aliquot of stock solution was diluted in distilled water to prepare the

175

ABTS working solution with an absorbance of at least 0.70±0.02 at 734 nm.

176

Concentrations of Trolox (0-0.25 mM), or appropriate concentrations of coffee brew

177

extract were mixed with ABTS working solution and incubated at room temperature for 9



Page 10 of 37

178

10 min. The absorbance was measured at 734 nm. A control contained ABTS working

179

solution. All the readings were subtracted from the reading of the blank, which was the

180

well with only distilled water. The % of inhibition was plotted against the concentration

181

of coffee/Trolox to generate regression equations. The ratio between the slope of coffee

182

sample regression equation and the slope of Trolox regression equation was defined as

183

the coffee sample antioxidant capacity. All of the analyses were run in triplicate. The

184

ABTS scavenging capacities of coffee bean extracts were expressed as mmol equivalents

185

of Trolox per g of coffee beans (dry matter).

186

ORAC method for the determination of antioxidant activity. The antioxidant

187

activities of coffee brew extracts were assessed using the oxygen radical absorbance

188

capacity fluorescein (ORAC) assay.27 Briefly, 100 µL samples in 75 mM phosphate

189

buffer (pH 7.0) or Trolox standard (final concentrations of 0-6.0 mM) and 60 µl

190

fluorescein (Sigma, St. Louis, MO) (final concentration 60 nM) were added to 96-well

191

black-walled plates and incubated at 37 °C for 10 min. This was followed by adding 60

192

µl AAPH (final concentration 12 mM) and the quench of fluorescence was monitored

193

every minute for 60 min at the excitation and emission wavelength of 485 nm and 527

194

nm, respectively. The area under the fluorescence decay curve (AUC) was used as

195

indication of the retention of fluorescein. The regression equation between the

196

concentration of Trolox/coffee and the AUC was calculated and the ratio between the

197

slope of coffee regression equation, to the slope of Trolox regression equation was used 10


Page 11 of 37


198

to express the antioxidant activity of the coffee brew extract. All of the analyses were run

199

in triplicate. The peroxyl radical scavenging capacities (ORAC value) of coffee bean

200

extracts were expressed as mmol Trolox/g coffee brew extract.

201

FTIR spectral collection. A Perkin Elmer Fourier transform infrared spectrometer

202

(Perkin Elmer, Waltham, MA) equipped with a Universal Attenuated Total Reflectance

203

sensor was used for infrared spectral acquisition. Pulverized coffee beans were applied

204

onto the ATR-IR ZnSe crystal and a 145 force gauge was applied on top of the sample.

205

Spectra were collected over a wavenumber range of 4000-400 cm-1 at 2 cm-1 spectral

206

resolution with 16 scans for each spectral collection. Six spectra were collected for each

207

sample. A background spectrum was collected at the beginning of the measurements and

208

every other hour thereafter.

209

Data processing and multivariate analysis. Multivariate analysis was used for

210

quantitative determination based upon FTIR spectra and reference values determined by

211

HPLC and antioxidant chemical assay results using the chemometrics software, Solo

212

(Eigenvector Research Inc., Manson, WA).28 The reference values were mean centered

213

before PLSR model construction. Similarly, the FTIR spectra were baseline corrected and

214

mean-centered before PLSR model construction. The effect of first and second derivative

215

transformations using a 9-point Savitsky-Golay filter on the performance of PLSR was

216

also evaluated according to previous reports.29,30 In order to build robust PLSR

217

calibration models, different varieties of coffee beans with different roasting conditions 11



Page 12 of 37

218

were analyzed to cover a wide range of samples. In total, 60 coffee bean samples were

219

included in the analysis. As a general rule, approximately 70% of samples should be used

220

to build the calibration model, with the remaining 30% of samples used to evaluate the

221

performance of the calibration model. In the current study, 42 samples were randomly

222

selected to form a training set used to build the calibration model and 18 samples were

223

used for external validation testing. During the calibration model construction process,

224

v-fold (also known as venetian blinds) cross-validation was used to execute internal

225

validation. Briefly, the v-fold cross-validation partitions the training set into equal sized v

226

(v≤n, n=42) segments and then a single segment is retained as the internal validation data

227

for testing the model, and the remaining v-1 segments are used to build the model. The

228

v-fold cross-validation process is then repeated v times with each of the v segments used

229

exactly once as the validation data.31 Usually, v is smaller than n and it is often reported

230

that the optimal v-fold value falls within a range of 5-10.31 In our study, we chose the

231

v-fold cross-validation (v=7) rather than a leave-one-out cross-validation, because using

232

the leave-one-out cross-validation could predispose for higher variance and bias when the

233

size of the training set samples is relatively large (>20).32 The PLSR method is commonly

234

used to extract orthogonal factors (latent variables) that are linear combinations of the

235

spectra variables, accounting for as much of the manifest factor variation as possible

236

while modeling the response analytes well.22 The number of latent variables determines

237

the polynomial model. The cross validation selects the optimal number of latent variables 12


Page 13 of 37


238

by two criteria: (a) minimizing the root mean square error of cross validation (RMSEcv)

239

and similarity criterion by means of the RMSECAL to RMSECV ratio (the closer to 1; the

240

better);33 (b) avoiding overfitting by limiting the number of latent variables smaller than 7.

241

RMSEcv is calculated as the formula in Equation (I):34

242

RMSEcv = ∑yi − ŷi

(I)

243

where N is the number of samples in the training set, v is the number of segments, n is the

244

number of samples in a given segment, yiv is the reference value for sample i and segment

245

v, and ŷiv is predicted value for sample i in chunk v. In the current study, our components

246

were: N=42, v=7, and n=6. After constructing models with different spectral

247

pre-processing methods (1st derivative and 2nd derivative), the optimal calibration model

248

was selected on the basis of lowest RMSECV values and the highest correlation coefficient

249

(R2) value.35 Optimal calibration models were established from the data of training sets

250

corresponding to chlorogenic acid isomer content and antioxidant activity for use in later

251

validation of the 18 unknown coffee samples that represented the testing set.

252

Results and Discussion

253

FTIR spectral features of pure chlorogenic acid isomer standards. The molecular

254

structures of major chlorogenic acid isomers are shown in Figure 1 and related FTIR

255

spectra in Figure 2A. Despite chlorogenic acid isomers having specifically different FTIR

256

spectra, all chlorogenic acid isomer standards generated strong infrared absorption

257

spectra in the regions of 1300-800 cm-1 and 1700-1500 cm-1, respectively. A distinctive 13



Page 14 of 37

258

band at the wavenumber of 809 cm-1 was assigned to cyclohexane C-O twisting.36 The

259

bands appearing at 1120 cm-1 and 1165 cm-1 are derived from cyclohexane CH, C-OH

260

bending and the phenyl ring bending vibration, respectively.19 The band at 1276 cm-1-was

261

due to phenyl CH rocking vibrations.36 All chlorogenic acid isomers have a prominent

262

infrared absorption band at 1605 cm-1, which is assigned to the phenyl ring stretch.37 The

263

band appearing at 1627 cm-1 can be attributed to C=C ethylenic stretching.36

264

FTIR spectral features of green and roasted coffee beans. Representative FTIR

265

spectra of green and roasted coffee beans are shown in Figure 2B. Several prominent

266

bands were shown in the wavenumber region of 1300-800 cm-1. The bands identified at

267

809, 1032, 1120, and 1165 cm-1 from ground coffee beans are the characteristic bands of

268

pure chlorogenic acid isomers. These chlorogenic acid isomers contributed to FTIR

269

spectral features of coffee beans. We therefore conclude, that the spectral feature in this

270

region has the potential to be used to develop chemometric models for quantitative

271

analysis of chlorogenic acid isomer composition in both the native form and after

272

different thermal treatments. A distinct FTIR absorption peak was also observed in the

273

wavenumber region of 1700-1600 cm-1. A previous study reported that this FTIR region

274

in is highly related to chlorogenic acid concentration in coffee.38 Our FTIR spectra of

275

pure chlorogenic acid isomers also showed that chlorogenic acid isomers have strong

276

absorption bands at 1605, 1627, and 1680 cm-1, indicating that spectral feature in the

277

wavenumber region of 1700-1600 cm-1 have an important role in further characterizing 14


Page 15 of 37


278

chlorogenic acid isomer composition in coffee. A sharp band at 1744 cm-1 is shown in all

279

of the coffee samples and this band is derived from carbonyl vibration in esters, as

280

reported previously.39

281

Quantitative analysis of chlorogenic acid isomers. The limits of quantification for

282

chlorogenic acid isomers by HPLC were 3.125 µM for 1, 2, 3 and 1.56 µM for 4, 5, 6. In

283

green beans, 3 is the most abundant chlorogenic acid isomer form with content ranging

284

from 39.88±1.19 mg/g (Ethiopia) to 45.00±1.59 mg/g (Sumatra). The content of 2 ranged

285

from 5.14±0.17 mg/g (Ethiopia) to 7.33±0.28 mg/g (Sumatra). 3 accounted for 68.9%,

286

70.8%, 68.1%, 71.2% and 72.7% of total caffeoylquinic and dicaffeoylquinic acids in

287

Dominica, Peru, Sumatra, PNG, and Ethiopia green beans, respectively. A significant loss

288

(around 90%) of 3 occurred in beans when roasted at 235 °C for 15 min. The content of 1

289

and 2 increased after light roasting in beans from all the regions except Sumatra. For

290

example, 1 increased from 3.80±0.13 to 4.92±0.09 mg/g, while a similar increase in 2

291

occurred from 6.15±0.41 mg/g to 6.94±0.12 mg/g after light roasting in Dominica beans.

292

In contrast, 3 decreased from 41.6±1.80 mg/g to 13.2±0.21 mg/g. It is highly possible that

293

part of the decrease in 3 reflected the increase in both 1 and 2, which occurred by acyl

294

migration through a formation of an orthoester.40 The dicaffeoylquinic acid isomers

295

contribute less to the total chlorogenic acid content compared to caffeoylquinic acid

296

isomers. As the roasting conditions increased from light to dark, the level of 4, 5, and 6

297

decreased significantly (P < 0.05). Our result indicates that the greater the extent to which 15



Page 16 of 37

298

the coffee beans are roasted, the lower the recovery of both total caffeoylquinic acids and

299

dicaffeoylquinic acids in these beans. Our findings confirm previous work that roasting

300

conditions used for coffee beans will result in degradation of total chlorogenic acid.5

301

PLSR models using the wavenumber range of 1800-700 cm-1 were constructed to

302

study the correlation between HPLC reference values obtained for six chlorogenic acid

303

isomer concentrations in coffee beans with processed (i.e., baseline corrected and mean

304

centered) FTIR spectral features. These calibration models were constructed by using the

305

data set of the randomly selected 42 coffee bean samples. The effects of FTIR spectra

306

pre-processing methods (first and second derivative transformations with a 9-point

307

Savitsky-Golay filter) on PLSR performances were evaluated. The statistical parameters

308

for PLSR models of chlorogenic acid isomer content in coffee beans when constructed on

309

the basis of FTIR raw spectra and spectra with different pre-processing methods are

310

shown in Table 1. A robust PLSR model should possess a high regression coefficient (e.g.

311

R2 > 0.90), preferably a small number of latent variables (0.90). Table 2 lists the

325

performance of these optimal PLSR models for six chlorogenic acid isomers in coffee

326

when tested using the remaining 18 coffee samples. Good agreement between results

327

obtained from predicted FTIR values and HPLC reference values were obtained for all

328

six chlorogenic acid isomers present in coffee.

329

The loading plot for the PLSR calibration model represents an interpretative tool

330

that illustrates the total variance associated with the spectral feature at each wavenumber.

331

Typically, as the loading value increases, the contribution from corresponding

332

wavenumber also increases. PLSR models developed for the content of six chlorogenic

333

acid isomers within a wavenumber range of 1800-700 cm-1 had similar loading plots. A

334

representative PLSR loading plot for chlorogenic acid isomer (e.g. structure 1) is shown

335

in Figure 3. The first two latent variables explained the majority of the variance (>90%)

336

in the PLSR model for the determination of 1 content in coffee beans. The most

337

distinctive loadings of these two latent variables were derived from the wavenumbers 17



Page 18 of 37

338

around 1700-1500 cm-1 and 1300-800 cm-1. These specific wavenumber regions are

339

consistent with those characterizing the pure chlorogenic acid isomers, as shown in

340

Figure 2A. These findings indicate that the PLSR model successfully captured important

341

FTIR spectral features for each chlorogenic acid isomer present in various coffee bean

342

samples. As a result, we used this information to construct PLSR models and determine

343

individual chlorogenic acid isomer content in ground coffee beans.

344

Determination of Antioxidant Activity. In addition, the antioxidant activities of coffee

345

bean extracts were determined using three different chemical assays (i.e. DPPH, ABTS,

346

and ORAC), all of which measured free radical scavenging capacity. Coffee methanol

347

extracts gave relatively higher antioxidant capacity values for the ORAC assay (range =

348

233.7 to 587.3 µmol Trolox equivalent/g of coffee) compared to the ABTS assay (range =

349

77.1 to 429.3 µmol Trolox equivalent/g of coffee) and DPPH assay (range = 70.3 to 202.9

350

mmol Trolox equivalent/g), respectively. The wide range of activities common for each

351

chemical antioxidant assay reflects the changes in coffee beans produced during thermal

352

treatment of roasting process. Moreover, absolute measurements of antioxidant activity

353

were specific to each assay, which indicates that coffee constituents react differently in

354

relative capacities to scavenge peroxyl, ABTS·+ and stable DPPH radicals,

355

respectively.26,43

356

Calibration models for the determination of antioxidant activity of coffee beans were

357

developed using baseline corrected and mean-centered FTIR spectra, first derivative 18


Page 19 of 37


358

FTIR spectra, and second derivative FTIR spectra in the wavenumber region of 1800 and

359

700 cm-1. The statistical parameters for PLSR models of antioxidant activity in coffee

360

beans constructed on the basis of FTIR spectra and spectra with different pre-processing

361

methods are shown in Table 3. The optimal model corresponding to individual

362

antioxidant activity for each chemical assay is highlighted in Table 3 by considering R2,

363

RMSECAL, RMSECV, and the number of latent variables. We found that the optimal

364

calibration models for the prediction of DPPH, ABTS and ORAC values were obtained

365

by using the first derivative FTIR spectra and the R-CV values were 0.811, 0.842, and

366

0.867 for DPPH, ABTS, and ORAC, assays, respectively. These regression coefficients

367

derived from models that predict antioxidant activities were comparatively lower than the

368

R2-CV derived from models that predicted chlorogenic acid isomer content. It is known

369

that FTIR spectra reveal functional group characteristics of specific chemical components

370

in coffee beans; however, FTIR spectra do not reflect potential interactions, competitive

371

inhibitions, or synergistic effects among coffee antioxidant components. These features in

372

turn are pertinent the free radical scavenging reaction processes and may explain why the

373

PLSR models for predicting chlorogenic acid isomer content in coffee gave higher

374

regression coefficients pertinent to those models that predicted antioxidant activity. We

375

also noticed that the PLSR model for predicting ORAC values had higher R-CV compared

376

to those models that predicted DPPH and ABTS values, respectively. ORAC values

377

reflect the antioxidant activity of coffee components involved in both initiation and 19



Page 20 of 37

378

propagation process,44 whereas ABTS and DPPH values reflect the direct free radical

379

scavenging capacity of coffee components. Therefore, the ORAC values had a higher

380

correlation with the FTIR spectra that reflect the whole chemical composition of coffee

381

beans.

382

These optimal PLSR models were used to predict antioxidant activity for each assay

383

system tested and for different coffee bean samples in the validation set (n=18). Table 4

384

compares the FTIR-PLSR-predicted antioxidant values and the values obtained from

385

chemical reference assays. It is noteworthy that the FTIR-PLSR method showed a better

386

reproducibility among all measurements, compared to that of chemical antioxidant

387

activity assays, as shown by a lower value of coefficient of variability. This finding is to

388

be expected since the radical scavenging PLSR models generated for different chemical

389

assays at best will correspond to a complex mixture of components present in coffee

390

beans that have different capacities to scavenge free radicals. This simple and direct

391

method for the determination of chlorogenic acid composition and antioxidant capacity

392

could be used to enable the coffee industry to efficiently assess the quality of green and

393

processed coffee beans, both in terms of ensuring optimal taste and flavor, while also

394

offering a coffee brew that contains potential antioxidant related health benefits when

395

consumed. In conclusion, the PLSR models constructed in this study were based on FTIR

396

spectra of coffee bean constituents and validated to be suitable practical analytical

397

methods to predict the coffee bean chlorogenic acid isomer profile. A similar potential 20


Page 21 of 37


398

exists for constructing PLSR models to predict antioxidant capacities by selected

399

chemical assays (e.g. ORAC). FTIR coupled with multivariate analysis provides a rapid,

400

inexpensive method for the coffee industry to efficiently and accurately monitor the

401

quality of coffee beans.

402 403

Author Information

404

Corresponding Author

405

*

406

University of British Columbia, 2205 East Mall, Vancouver, British Columbia, V6T 1Z4,

407

Canada. Tel: +1-604–822–5560. Fax: +1-604–822–5143. E-mail: [email protected].

408

Funding

409

This work was funded by the Natural Sciences and Engineering Research Council of

410

Canada (NSERC) Discovery Grant (DDK). NL is a recipient of a 4 year UBC academic

411

research scholarship.

412

Notes

413

The authors declare no competing financial interest.

414

Abbreviations used

415

ATR-FTIR, attenuated total reflectance-Fourier transform infrared; ORAC, oxygen

416

radical absorbance capacity; ABTS, 2,2’-azino-bis(3-ethylbenzothiazoline-6 sulphonic;

417

DPPH, 2,2-diphenyl-1-picrylhydrazyl; PLSR, partial least squares regression; PNG,

Food, Nutrition, and Health Program, Faculty of Land and Food Systems, The

21



Page 22 of 37

418

Papua New Guinea; RMSEcv, root mean square error of cross validation; RMSECAL, root

419

mean square error of calibration.

420 421

References

422

(1) Clifford, M. N., Chlorogenic acids and other cinnamates – nature, occurrence, dietary

423

burden, absorption and metabolism. J. Sci. Food Agric. 2000, 80, 1033-1043.

424

(2) del Castillo, M. D.; Ames, J. M.; Gordon, M. H., Effect of roasting on the antioxidant

425

activity of coffee brews. J. Agric. Food Chem. 2002, 50, 3698-3703.

426

(3) Farah, A.; de Paulis, T.; Trugo, L. C.; Martin, P. R., Effect of roasting on the

427

formation of chlorogenic acid lactones in coffee. J. Agric. Food Chem. 2005, 53,

428

1505-1513.

429

(4) Singh, B. R.; Wechter, M. A.; Hu, Y.; Lafontaine, C., Determination of caffeine

430

content in coffee using Fourier transform infrared spectroscopy in combination with

431

attenuated total reflectance technique: a bioanalytical chemistry experiment for

432

biochemists. Biochem. Educ. 1998, 26, 243-247.

433

(5) Moon, J.-K.; Yoo, H. S.; Shibamoto, T., Role of roasting conditions in the level of

434

chlorogenic acid content in coffee beans: correlation with coffee acidity. J. Agric. Food

435

Chem. 2009, 57, 5365.

436

(6) Perrone, D.; Farah, A.; Donangelo, C. M.; de Paulis, T.; Martin, P. R.,

437

Comprehensive analysis of major and minor chlorogenic acids and lactones in 22


Page 23 of 37


438

economically relevant Brazilian coffee cultivars. Food Chem. 2008, 106, 859-867.

439

(7) Schols, H. A.; Boekel, v. T.; Smit, G.; Bekedam, E. K., Incorporation of chlorogenic

440

acids in coffee brew melanoidins. J. Agric. Food Chem. 2008, 56, 2055-2063.

441

(8) Ortolá, M. D.; Gutiérrez, C. L.; Chiralt, A.; Fito, P., Kinetic study of lipid oxidation

442

in roasted coffee. Food Sci. Technol. Int. (London, U. K.) 1998, 4, 67-73.

443

(9) Mitjavila, M. T.; Moreno, J. J., The effects of polyphenols on oxidative stress and the

444

arachidonic acid cascade. Implications for the prevention/treatment of high prevalence

445

diseases. Biochem. Pharmacol. 2012, 84, 1113-1122.

446

(10) Xu, J.-G.; Hu, Q.-P.; Liu, Y., Antioxidant and DNA-protective activities of

447

chlorogenic acid isomers. J. Agric. Food Chem. 2012, 60, 11625-11630.

448

(11) Monteiro, M.; Farah, A.; Perrone, D.; Trugo, L. C.; Donangelo, C., Chlorogenic acid

449

compounds from coffee are differentially absorbed and metabolized in humans. J. Nutr.

450

2007, 137, 2196.

451

(12) Natella, F.; Nardini, M.; Giannetti, I.; Dattilo, C.; Scaccini, C., Coffee drinking

452

influences plasma antioxidant capacity in humans. J. Agric. Food Chem. 2002, 50,

453

6211-6216.

454

(13) Ranheim, T.; Halvorsen, B., Coffee consumption and human health - beneficial or

455

detrimental? - Mechanisms for effects of coffee consumption on different risk factors for

456

cardiovascular disease and type 2 diabetes mellitus. Mol. Nutr. Food Res. 2005, 49,

457

274-284. 23



Page 24 of 37

458

(14) Moeenfard, M.; Rocha, L.; Alves, A., Quantification of caffeoylquinic acids in

459

coffee brews by HPLC-DAD. J. Anal. Methods Chem. 2014, 2014, 1-10.

460

(15) Frankel, E. N.; Meyer, A. S., The problems of using one‐dimensional methods to

461

evaluate multifunctional food and biological antioxidants. J. Sci. Food Agric. 2000, 80,

462

1925-1941.

463

(16) Ewa, B. G. N., Antioxidant properties of a and Robusta coffee extracts prepared

464

under different conditions. Dtsch. Lebensm.-Rundsch. 2008, 69-77.

465

(17) Arosha, N. W. D. D. K., Modulation of metal-induced cytotoxicity by Maillard

466

reaction products isolated from coffee brew. J. Toxicol. Environ. Health, Part A 1998, 55,

467

151-168.

468

(18) Craig, A. P.; Franca, A. S.; Oliveira, L. S., Evaluation of the potential of FTIR and

469

chemometrics for separation between defective and non-defective coffees. Food Chem.

470

2012, 132, 1368-1374.

471

(19) El-Abassy, R. M.; Donfack, P.; Materny, A., Discrimination between Arabica and

472

Robusta green coffee using visible micro Raman spectroscopy and chemometric analysis.

473

Food Chem. 2011, 126, 1443-1448.

474

(20) Lu, X.; Wang, J.; Al-Qadiri, H. M.; Ross, C. F.; Powers, J. R.; Tang, J.; Rasco, B. A.,

475

Determination of total phenolic content and antioxidant capacity of onion (Allium cepa)

476

and shallot (Allium oschaninii) using infrared spectroscopy. Food Chem. 2011, 129,

477

637-644. 24


Page 25 of 37


478

(21) Versari, A.; Parpinello, G. P.; Scazzina, F.; Rio, D. D., Prediction of total antioxidant

479

capacity of red wine by Fourier transform infrared spectroscopy. Food Control 2010, 21,

480

786-789.

481

(22) Miller, C. E., Chemometrics for on-line spectroscopy applications—theory and

482

practice. J. Chemom. 2000, 14, 513-528.

483

(23) Fujioka, K.; Shibamoto, T., Chlorogenic acid and caffeine contents in various

484

commercial brewed coffees. Food Chem. 2008, 106, 217-221.

485

(24) Jing, H.; Kitts, D. D., Antioxidant activity of sugar–lysine Maillard reaction products

486

in cell free and cell culture systems. Arch. Biochem. Biophys. 2004, 429, 154-163.

487

(25) Re, R.; Pellegrini, N.; Proteggente, A.; Pannala, A.; Yang, M.; Rice-Evans, C.,

488

Antioxidant activity applying an improved ABTS radical cation decolorization assay.

489

Free Radicals Biol. Med. 1999, 26, 1231-1237.

490

(26) Delgado-Andrade, C.; Rufian-Henares, J. A.; Morales, F. J., Assessing the

491

antioxidant activity of melanoidins from coffee brews by different antioxidant methods. J.

492

Agric. Food Chem. 2005, 53, 7832-7836.

493

(27) Kitts, D. D.; Hu, C., Biological and chemical assessment of antioxidant activity of

494

sugar-lysine model Maillard reaction products. Ann. N. Y. Acad. Sci. 2005, 1043,

495

501-512.

496

(28) Yobi, A.; Wone, B. W. M.; Xu, W.; Alexander, D. C.; Guo, L.; Ryals, J. A.; Oliver, M.

497

J.; Cushman, J. C., Comparative metabolic profiling between desiccation-sensitive and 25



Page 26 of 37

498

desiccation-tolerant species of Selaginella reveals insights into the resurrection trait.

499

Plant J. 2012, 72, 983-999.

500

(29) Brereton, R. G., Chemometrics: data analysis for the laboratory and chemical plant.

501

Wiley: Hoboken, NJ; Chichester, West Sussex, England, 2003.

502

(30) Rohman, A.; Man, Y. B. C.; Ismail, A.; Hashim, P., Application of FTIR

503

spectroscopy for the determination of virgin coconut oil in binary mixtures with olive oil

504

and palm oil. J. Am. Oil Chem. Soc. 2010, 87, 601-606.

505

(31) Baumann, K.; von Korff, M.; Albert, H., A systematic evaluation of the benefits and

506

hazards of variable selection in latent variable regression. Part II. Practical applications. J.

507

Chemom. 2002, 16, 351-360.

508

(32) Arlot, S.; Celisse, A., A survey of cross-validation procedures for model selection.

509

Statistics Surveys 2010, 4, 40-79.

510

(33) Faber, N. M.; Rajkó, R., How to avoid over-fitting in multivariate calibration—The

511

conventional validation approach and an alternative. Anal. Chim. Acta 2007, 595, 98-106.

512

(34) Soriano, A.; Pérez-Juan, P. M.; Vicario, A.; González, J. M.; Pérez-Coello, M. S.,

513

Determination of anthocyanins in red wine using a newly developed method based on

514

Fourier transform infrared spectroscopy. Food Chem. 2007, 104, 1295-1303.

515

(35) Tripathi, S.; Mishra, H. N., A rapid FT-NIR method for estimation of aflatoxin B1 in

516

red chili powder. Food Control 2009, 20, 840-846.

517

(36) Eravuchira, P. J.; El-Abassy, R. M.; Deshpande, S.; Matei, M. F.; Mishra, S.; Tandon, 26


Page 27 of 37


518

P.; Kuhnert, N.; Materny, A., Raman spectroscopic characterization of different

519

regioisomers of monoacyl and diacyl chlorogenic acid. Vib. Spectrosc. 2012, 61, 10-16.

520

(37) Biswas, N.; Kapoor, S.; Mahal, H. S.; Mukherjee, T., Adsorption of CGA on

521

colloidal silver particles: DFT and SERS study. Chem. Phys. Lett. 2007, 444, 338-345.

522

(38) Ribeiro, J. S.; Salva, T. J.; Ferreira, M. M. C., Chemometric studies for quality

523

control of processed Brazilian coffees using drift. J. Food Qual. 2010, 33, 212-227.

524

(39) Kemsley, E. K.; Ruault, S.; Wilson, R. H., Discrimination between Coffea arabica

525

and Coffea canephora variant robusta beans using infrared spectroscopy. Food Chem.

526

1995, 54, 321-326.

527

(40) Deshpande, S.; Jaiswal, R.; Matei, M. F.; Kuhnert, N., Investigation of acyl

528

migration in mono- and dicaffeoylquinic acids under aqueous basic, aqueous acidic, and

529

dry roasting conditions. J. Agric. Food Chem. 2014, 62, 9160-9170.

530

(41) Lu, X.; Ross, C. F.; Powers, J. R.; Rasco, B. A., Determination of quercetins in onion

531

(Allium cepa) using infrared spectroscopy. J. Agric. Food Chem. 2011, 59, 6376-6382.

532

(42) Abdi, H., Partial least squares regression and projection on latent structure regression

533

(PLS Regression). Wiley Interdisciplinary Reviews: Comput. Stat 2010, 2, 97-106.

534

(43) Liang, N. J.; Kitts, D. D., Antioxidant property of coffee components: assessment of

535

methods that define mechanisms of action. Molecules 2014, 19, 19180-19208.

536

(44) Cao, G.; Alessio, H. M.; Cutler, R. G., Oxygen-radical absorbance capacity assay for

537

antioxidants. Free Radicals Biol. Med. 1993, 14, 303-311. 27



Page 28 of 37

FIGURE CAPTIONS Figure 1. Chemical structure of chlorogenic acid isomers (1, 2, 3, 4, 5, and 6 represent 3-caffeoylquinic acid, 4-caffeoylquinic acid, 5-caffeoylquinic acid, 3,5-dicaffeoylquinic acid, 3,4-dicaffeoylquinic acid, and 4,5-dicaffeoylquinic acid, respectively).

Figure 2. Raw FT-IR spectra of: A. chlorogenic acid isomers [(a) 4,5-dicaffeoylquinic acid; (b) 3,4-dicaffeoylquinic acid; (c) 3,5-dicaffeoylquinic acid; (d) 4-caffeoylquinic acid; (e) 3-caffeoylquinic acid; (f) 5-caffeoylquinic acid]; and B. representative coffee bean samples [(i) green coffee beans; (ii) dark roasted coffee beans; (iii) medium roasted coffee beans; (iv) light roasted coffee beans].

Figure 3. PLSR loading plot of the first two latent variables (blue line: First latent variable; orange line: second latent variable) or the cross-validated models to determine 5-caffeoylquinic acid content in coffee beans.

28


Page 29 of 37


Table 1: PLSR Models for Determination of Chlorogenic Acid Isomer Content in Coffee Beans RMSE CGA N of Range Preprocessing N R2-CALa RMSECALb R2-CVc d method of (mg/g) Isomer Sample (mg/g) CV LV (mg/g) e g 42 1.54-5.72 Non filtering 7 0.886 0.430 0.835 0.517 1 st f 42 1.54-5.72 1 derivative 6 0.946 0.296 0.921 0.359 nd f 42 1.54-5.72 2 derivative 6 0.920 0.361 0.863 0.471 42 2.15-8.47 Non filtering 6 0.907 0.614 0.879 0.701 2 st 42 2.15-8.47 1 derivative 5 0.957 0.417 0.944 0.477 nd 42 2.15-8.47 2 derivative 6 0.947 0.461 0.913 0.593 42 3.47-42.62 Non filtering 5 0.989 1.617 0.986 1.796 3 st 42 3.47-42.62 1 derivative 6 0.994 1.200 0.990 1.549 nd 42 3.47-42.62 2 derivative 4 0.986 1.812 0.982 2.095 42 0.09-2.13 Non filtering 7 0.977 0.128 0.962 0.163 4 st 42 0.09-2.13 1 derivative 8 0.989 0.084 0.979 0.119 nd 42 0.09-2.13 2 derivative 7 0.984 0.106 0.962 0.162 42 0.01-3.93 Non filtering 3 0.968 0.265 0.962 0.290 5 st 42 0.01-3.93 1 derivative 5 0.986 0.177 0.979 0.215 nd 42 0.01-3.93 2 derivative 4 0.985 0.184 0.978 0.218 42 0.05-3.74 Non filtering 5 0.973 0.211 0.966 0.240 6 st 42 0.05-3.74 1 derivative 6 0.989 0.137 0.984 0.167 nd 42 0.05-3.74 2 derivative 6 0.987 0.149 0.973 0.211 a 2 R -CAL, coefficient of correlation for calibration model b RMSECAL, root mean square error of calibration c 2 R -CV, coefficient of correlation for cross-validated model d RMSECV, root mean square error of cross-validation e Baseline corrected, but non filtering f Baseline corrected, and with filtering g 1, 2, 3, 4, 5, and 6 represent 3-caffeoylquinic acid, 4-caffeoylquinic acid, 5-caffeoylquinic acid, 3,5-dicaffeoylquinic acid, 3,4-dicaffeoylquinic acid, and 4,5-dicaffeoylquinic acid, respectively. Their structures are shown in Figure 1.

29



Page 30 of 37

Table 2: Chlorogenic Acid Isomers Content in Coffee Beans Determined by the HPLC and Predicted by FT-IR in Randomly Validation Set Samples Analyte Samples HPLCa SDb CV (%)c IRd SDb CV (%)c Ethiopia-Gf 3.45 0.00 0.06 5.23 0.30 5.73 1e Dominica-D 1.84 0.04 1.99 2.35 0.13 5.53 Peru-L 5.46 0.15 2.78 5.14 0.06 1.17 Peru-M 3.33 0.32 9.73 3.63 0.05 1.37 Peru-D 1.62 0.09 5.47 1.84 0.12 6.52 Sumatra-L 4.67 0.12 2.66 4.50 0.04 0.89 Ethiopia-G 6.65 0.03 0.47 6.13 0.54 8.81 2 Dominica-D 2.64 0.11 3.99 3.33 0.17 2.51 Peru-L 7.51 0.26 3.47 7.09 0.13 1.84 Peru-M 4.67 0.21 4.47 5.12 0.05 0.98 Peru-D 2.36 0.09 3.75 2.84 0.19 6.63 Sumatra-L 6.54 0.08 1.16 6.03 0.06 0.94 Ethiopia-G 42.23 0.06 0.14 41.51 2.85 6.86 3 Dominica-D 4.00 0.25 6.13 6.70 0.57 12.83 Peru-L 13.15 0.66 4.98 12.95 0.49 3.78 Peru-M 7.89 0.17 2.20 8.94 0.34 3.80 Peru-D 3.59 0.09 2.46 3.74 0.24 6.42 Sumatra-L 11.56 0.21 1.80 11.04 0.41 3.71 Ethiopia-G 1.56 0.02 1.29 1.99 0.15 7.54 4 Dominica-D 0.14 0.01 8.06 0.25 0.02 8.00 Peru-L 0.84 0.05 5.39 0.80 0.08 10.0 Peru-M 0.35 0.03 7.23 0.51 0.04 7.84 Peru-D 0.07 0.01 14.29 0.17 0.04 23.5 Sumatra-L 0.85 0.02 2.75 0.65 0.07 10.77 Ethiopia-G 2.98 0.04 1.25 3.14 0.35 11.15 5 Dominica-D 0.03 0.01 21.65 0.17 0.04 23.53 Peru-L 0.47 0.02 4.30 0.49 0.04 8.16 Peru-M 0.12 0.00 1.59 0.27 0.04 14.81 Peru-D 0.04 0.01 25.00 0.01 0.01 100.00 Sumatra-L 0.41 0.07 15.74 0.29 0.02 6.90 Ethiopia-G 3.07 0.05 1.54 3.40 0.35 10.29 6 Dominica-D 0.08 0.03 35.94 0.30 0.04 13.33 Peru-L 0.89 0.04 4.49 0.91 0.08 8.79 Peru-M 0.36 0.01 2.78 0.56 0.02 3.57 Peru-D 0.04 0.02 50.00 0.14 0.08 57.14 Sumatra-L 0.80 0.06 7.40 0.68 0.06 8.82 30


Page 31 of 37


a

HPLC, chlorogenic acid isomer content measured by HPLC (mg/g); SD, standard deviation; c CV (%) , coefficient of variation; d IR, chlorogenic acid isomer content predicted by FTIR (mg/g); e 1, 2, 3, 4, 5, and 6 represent 3-caffeoylquinic acid, 4-caffeoylquinic acid, 5-caffeoylquinic acid, 3,5-dicaffeoylquinic acid, 3,4-dicaffeoylquinic acid, and 4,5-dicaffeoylquinic acid, respectively. Their structures are shown in Figure 1. f “G” represents “green”; “L” represents “light roasted”, “M” represents “medium roasted”, and “D” represents “dark roasted”; b

31



Table 3: PLSR Models for Determination of Antioxidant Activity of Coffee Beans Assays N of Range Preprocessing N of LV R2-CALa method samples (µM Trolox/g) DPPH 42 81.49-202.92 Non filteringe 5 0.844 st f 42 81.49-202.92 1 derivative 6 0.886 nd 42 81.49-202.92 2 derivative 5 0.882 ABTS 42 111.41-429.32 Non filtering 5 0.836 st 42 111.41-429.32 1 derivative 6 0.900 nd 42 111.41-429.32 2 derivative 5 0.898 ORAC 42 270.72-587.34 Non filtering 6 0.864 st 42 270.72-587.34 1 derivative 6 0.915 nd 42 270.72-587.34 2 derivative 5 0.911

RMSECALb (µM Trolox/g) 13.001 11.146 11.312 38.379 29.998 30.294 35.370 28.038 28.960

Page 32 of 37

R2-CVc 0.804 0.811 0.795 0.796 0.842 0.839 0.819 0.867 0.863

RMSECVd (µM Trolox/g) 14.651 14.392 14.978 42.984 37.955 38.181 41.062 35.158 35.620

a

R-CAL, coefficient of correlation for calibration model RMSECAL, root mean square error of calibration c R-CV, coefficient of correlation for cross-validated model d RMSECV, root mean square error of cross-validation e Baseline corrected, but non filtering f Baseline corrected, and with filtering b

32


Page 33 of 37


Table 4: Antioxidant Activity of Coffee Beans Methanol Extract Determined by Reference Assays (DPPH, Predicted by FT-IR in Validation Set Samples. Assays Samples Reference value SDa CV (%)b FT-IR predicted (µM Trolox/g) value (µM Trolox/g) DPPH Ethiopia-G 161.35 16.41 10.17 189.89 Dominica-D 79.56 6.32 7.95 112.04 Peru-L 137.62 12.15 8.83 138.97 Peru-M 107.17 7.21 6.73 112.64 Peru-D 76.94 6.62 8.60 89.89 Sumatra-L 139.59 12.51 8.96 146.09 ABTS Ethiopia-G 327.59 19.76 6.03 393.01 Dominica-D 85.36 30.11 35.28 69.92 Peru-L 281.47 19.63 6.97 247.66 Peru-M 181.46 26.84 14.79 199.09 Peru-D 127.42 13.17 10.34 113.70 Sumatra-L 269.35 33.69 12.51 277.93 ORAC Ethiopia-G 494.62 20.39 4.12 555.53 Dominica-D 249.83 17.20 6.88 315.41 Peru-L 422.47 17.55 4.15 416.02 Peru-M 333.85 20.20 6.05 367.70 Peru-D 252.12 18.40 7.30 282.49 Sumatra-L 416.98 18.35 4.40 432.52

ABTS, and ORAC) and SDa

CV (%)b

9.78 4.52 4.96 6.44 4.60 0.72 20.07 13.14 7.25 8.06 10.65 2.16 24.92 11.56 6.26 6.33 7.43 0.53

5.15 4.03 3.57 5.72 5.12 0.49 2.63 18.79 3.37 7.63 8.87 2.71 4.49 3.67 1.50 1.72 2.63 0.12

a

SD (standard deviation), b CV (%) (coefficient of variation). “G” represents “Green coffee”, “L” represents “Light roasted coffee”, “M” represents “Medium roasted coffee”, “D” represents “Dark roasted coffee”.

33



Page 34 of 37

Figure 1

34


Page 35 of 37


A

B

Figure 2

35



Page 36 of 37

Figure 3

36


Page 37 of 37


Table of Content Graphics

For Table of Contents Only

37


Application of Attenuated Total ReflectanceâFourier Transformed

Recommend Documents

Application of Attenuated Total ReflectanceâFourier Transformed