Boffo E. F., Tavares L. A., Ferreira A. G., Ferreira M. M. C., Tobias A. C. T, "Determination of the authenticity of commercial coffees by application of chemometrics to 1H NMR and FT-IR spectra". Nottingham, UK, 16-19/07/2006: 8th International Conference on The Applications of Magnetic Resonance in Food Science, Online Book of Abstracts (2006). Poster.
Determination
of the authenticity of commercial coffees by
application
of chemometrics to 1H NMR and FT-IR spectra
Leila
A. Tavares1,
Elisangela F. Boffo1, Antonio G. Ferreira1
& Márcia M C. Ferreira2
(1)Universidade
Federal de São Carlos, São Carlos, SP, BR.
(2)Universidade
Estadual de Campinas, Campinas, SP, BR.
giba@dq.ufscar.br
Coffee is one of
the most widely consumed beverages in the world and
Brazil is the
first producer.
Its quality control is difficult because depends
of the weather, harvest
conditions, species
used, soil nutrients, etc and it can be also adulterated with cheaper
substitutes like barley,
chicroy, cereals, malt, maltodextrins, caramelised
sugar, etc
[1,2].
Many techniques have been
investigated to tackle the problem of coffee
adulteration.
Spectroscopic methods
such as NMR and FT-IR than can monitor
a wide range of
chemicals in a
single spectrum have been a great success. However, the
richness of
this information makes the
spectra too complex to be analysed one by one and require
chemometric analysis to
extract the useful information [3].
In order to
verify which methodology is better to determined
barley addition into the
coffees, multivariate analysis
methods were applied to 1H
NMR and FT-IR spectra and
these results
were compared. In addition, k-Nearst
Neighbour (KNN) and Soft
Independent Modelling
of Class Analogy (SIMCA) were
used and compared to
determine the
sample class, and Partial
Least Squares (PLS) and
Principal
Component Regression
(PCR) were used to determine
the content of barley
addition into the commercial
coffees using the Pirouette®
v. 2.02 program.
Considering the classification
methods, KNN presented better efficiency than SIMCA,
for both techniques,
and attributed correctly 100% for the samples from an
external
validation group.
After this, the models were applied to predict
the class for coffee
commercial samples.
The coffee content
for the pattern and commercial samples were
determined using
PLS and PCR which showed
similar predictions. When both methods were applied, the
NMR data results showed
better with relative errors 5.3 and 5.6% for
PLS and PCR,
respectively.
For the FT-IR spectra data we found 16.8
and 17.5% for the same
samples used in the NMR
References:
[1]. Prodolliet, J. et al. Z. Lebensn
Unters Forsch A, 1998, 207, 1.
[2]. Charlton, A. J. et al. J.
Agric. Food Chem. 2002, 50, 3098.
[3]. Suchánek, M. et al. Fresenius
J. Anal. Chem. 1996, 354, 327.
The 8th International Conference on the Applications of Magnetic Resonance in Food Science, 2006, Nottingham