80.

Pedro M. K. A.,Ferreira M. M. C., "Simultaneuosly Calibrating Solids, Sugars and Acidity of Tomato Products Using PLS2 and NIR Spectroscopy". Águas de Lindóia, SP, Brazil, 10-15/09/2006: 10th International Conference on Chemometrics in Analytical Chemistry (CAC-2006, CAC-X), Book of Abstracts (2006) OP35. Oral 035.


10th International Conference on Chemometrics in Analytical Chemistry OP35

Simultaneously Calibration Solids, Sugars and Acidity of Tomato
Products Using PLS2 and NIR Spectroscopy

André M. K. Pedro*1 and Márcia M. C. Ferreira2   andre.k.pedro@unilever.com

1Unilever Brasil Ltda, Av. Invernada 401, Valinhos (SP), 13271-450, Brazil,  2Universidade Estadual de
Campinas (UNICAMP), Chemistry Institute, Physical Chemistry Department, Campinas (SP), 13083-970
Caixa Postal 6154, Brazil

Keywords: Fourier filter, multivariate calibration, tomato quality
_____________________________________________________________________________________
 

    The purpose of  this work is to develop  a  robust calibration model  for determining,  simultaneously  and
non-destructively,   relevant  quality parameters  in  processed tomato products,   namely total  and  soluble
solids, total acidity and sugars.   These are key quality parameters once  they  largely  determine  sensorial
profile and consumer acceptance of tomato products.
    Forty-two samples  of  tomato concentrate products  with total solids content ranging from  6.9  to  35.9%
were collected from  Latin America,  the  US  and  Europe.  Three spectra of each sample were acquired  in
the   4000  to  10000 cm-1  region  using   a  Büchi  NIRLab  N-200  spectrometer  with  a  MSC-100  diffuse
reflectance  cell.   Total  solids   were  determined  by  oven-drying  under  vacuum  (70º,  ~150 mmHg  abs.
pressure) until constant weight. Soluble solids were determined using a Metrohm 702 automatic titrator and
NaOH 0.1 mol L-1 Merck standard solution. Sugars were quantified by HPLC using  an  Shimatzu  refractive
index detector,  a Shodex NH2P-50 4E  column  and  Shodex NH2P-50G  column-guard.  Mobile phase was
acetonitrile:water (75:25),  isocratic, 1 mLmin-1.
    Simultaneous  Partial  Least  Squares  (PLS2)1,2   was  used  to  build  the  calibration  models.   Original
spectra  were  pre-processed  by  applying  a  50-points mean smoother,  followed  by  multiplicative  signal
correction (MSC)1-3.  Alternatively, a Fourier filter with Gaussian convolution function was applied,  followed
by  MSC  and  first-   or   second-derivatives  according   to   Savitsky  and  Golay4.   Seven  samples  were
separated  for  external validation  and  the others  were  used in the calibrations.   The optimum number of
factors was determined by  leave-one-out cross validation1,2.   Comparisons between different models were
performed through RMSEP and  rval1,2.  PLS1 models were built for comparison as well1,2.
    Results  have  shown  that  PLS2  gave  models  with  better  prediction  abilities  and/or  with   a  smaller
number of factors than PLS1.   The  best  pre-processing  strategy was mean smoothing followed by  MSC;
despite  the  Fourier filter followed  by MSC giving quite similar results,  it is a more complex approach.  The
RMSEP  for  each  property  was:  total solids,  0.6294;  soluble solids,  0.6755;  total acidity,  0.2189;  total
sugar,  1.9648;  glucose,  0.5365 and fructose,  0.8802.  The final model required only 4 factors,  much less
than PLS1 models previously reported5.
    As conclusion,  PLS2 gave a good model  for predicting the levels of important properties  of concentrate
tomato products.   Despite  its  usage  not  being  widely  reported  in the literature,   this application clearly
shows  the advantages  of  PLS2  over  PLS1  for  some applications,  as  a better  and  simpler model was
obtained.

Acknowledgement.   The authors would like to thank Colin Haine,  Unilever's Head of the Tomato Global
Technology Centre, and his team, for providing the samples for this study.
__________________________________________________________________________________________________________________________

References

1Martens H.; Naes T. Multivariate Calibration (2nd edn), vol. 1. Wiley: Chichester, UK, 1989.
2Beebe K. R.; Pell R. J.; Sheasholtz M.B. Chemometrics: a Practical Guide, Wiley: New Yourk, US, 1998.
3Isakson T.; Naes T. Appl. Spectrosc. 1988, 42, 1273-1286.
4Savstsky A.; Golay M. Anal. Chem. 1964, 36, 1627.
5Ferreira M.M.C., Pedro A.M.K. Anal. Chem. 2005, 77, 2505-2511.