Tavares L. A., Barison A., Ferreira A. G., Ferreira M. M. C., Corrêa A., Mattoso L. H., "APPLICATION OF CHEMOMETRIC METHODS IN THE COFFEE BLENDS EVALUATION". São Carlos, SP, Brazil, 01-04/12/2004: V Brazilian Meeting on Chemistry of Food and Beverages (V BMCFB), Book of Abstracts (2004). Poster. Session: Quality control.

APPLICATION OF CHEMOMETRIC METHODS IN THE COFFEE BLENDS
EVALUATION
 

L. A. Tavares*, A. Barison, A. G. Ferreira, M. M. C. Ferreira, A. Corrêa and L. H.
Mattoso

Universidade Federal de São Carlos, Departamento de Química, Laboratório de
Ressonância Magnética Nuclear.
leilaley@yahoo.com
 

        Consumers  approve  a  product   when  it    is  produced  with    a  certain  quality
standard.The manufacturer must  therefore make  sure that the  quality of this product is
controlled. In coffee,  an  important part  of this control is  the coffee specie used. There
are appoximately   eighty   coffee   species   however   only   two    are   of  commercial
importance¹,  arabica  and  robusta.  Arabica  coffee  presents  more  pronounced  and
refined flavor,  and  robusta  generally   is   mixed  with  arabica  to  make  the  product
cheaper.
        Our work proposes an evaluation of coffee blends  with several proportions of  this
species. The information about the extracted components in a cup  of coffee is  obtained
through ¹H NMR spectra, then it is analysed by chemometric methods.
        The  ¹H NMR  spectra were  recorded  on  a   Bruker  DRX400  spectrometer,  in
triplicate. In  0.6  mL  of coffee solution we added three drops  of  D₂O  for  lock  signal
and  the  temperature was maintained  constant,  303 K.  For  chemometric  analysis  the
Piroutte^®  software package was used.  An  exploratory  study  from  spectroscopic  data
was  done  using  the  Principal  Component  Analysis  (PCA)  and  Hierarchical  Cluster
Analysis  (HCA)  methods.  Classification  and  calibration models  also were  built  from
K-th Nearest Neighbor (KNN) and  Partial Least Squares  (PLS)² methods,  respectively.
After  building  the  models,   they   were  validated   with  external  samples,   doing  the
prediction of classes (more arabica content/more robusta content)  or  of  coffee species
amount in an unknown sample.
        This  work  shows  that   ¹H NMR  spectroscopy   together  with  the chemometric
methods can be used in an "easy way" to indentify and quantify the arabica and robusta
contents in coffee blends.

1. Belton, P. S.; Colquhon, I. J.; Kemsley, E. K.;  Delgadillo, I.;  Roma, P.; Dennis,  M.
J.; Sharman,   M.;   Holmes,   E.;   Nicholson,  J.   K.;   Spraul,   M.     Application   of
chemometrics to the  ¹H NMR  spectra  of  apple  juices:  discrimination  between  apple
varieties. Food Chem. 1998, 6, 207-213.
2. Brereton, R. G.  Chemometrics - Data Analysis for Laboratory and Chemical Plant.
Wiley, 2002, p. 183.