107.

Ribeiro J. S., Teófilo R. F., Augusto F., Ferreira M. M. C., "SIMULTANEOUS MULTIPLE RESPONSE OPTIMIZATION OF MICROEXTRACTION CONDITIONS USING PRINCIPAL COMPONENT ANALYSIS AND RESPONSE SURFACE METHODOLOGY TO COFFEE VOLATILE EXTRACTION". Campinas, SP, Brazil, 14-19/09/2008: 22nd International Conference of Coffee Science, Programme & Abstracts (2008) 132. Poster PC768.


PC768
SIMULTANEOUS MULTIPLE RESPONSE OPTIMIZATION OF
MICROEXTRACTION CONDITIONS USING PRINCIPAL COMPONENT
ANALYSIS AND RESPONSE SURFACE METHODOLOGY TO COFFEE
VOLATILE EXTRACTION
____________________________________________________________________________________________________

RIBEIRO, Juliano S.*, TEÓFILO, Reinaldo, F.*,  AUGUSTO, F.*,  FERREIRA, Márcia M. C*

*Universidade Estadual de Campinas, SP, Brazil

Principal component analysis (PCA) [1] and response surface methodology (RSM) [2] were applied to simultaneous multiple
responses optimization  (MRO)  of  the  headspace-solid-phase  microextraction  (HS-SPME)  conditions  to  extract  volatile
compounds from roasted arabica coffee.   In  a greater number of situations,   some  or  all  chromatographic peaks  present
relatively high correlation.   This fact  is  a  great advantage  in  MRO analysis  because  the  correlated  responses  provide
redundant information.   The  initial responses were 57 peak areas obtained  from  gas chromatographic system  with  flame
ionization detector (GC-FID).

The  basis  of  MRO  consists  in  compact several  correlated peak areas  in  one component through  PCA  and  uses  this
component as response in the central composite design  (CCD).  Through a correlogram map  (Figure 1A),  it was observed
direct correlations among peak areas in two subsets (in red). Negative correlations were observed among peak areas of the
two subsets (in blue),   which means that the responses  of  one subset brings different chemical information than the  other.
Hence,   the multiple response analyses using  PCA  were performed separately  for each subset,   in order to obtain higher
explained variance in PC1.   The first components of  the two subsets explained  64.51  and  81.98 %  of  the data variance,
respectively. ANOVA indicated that both regression models are significant  (p < 0.05)  and lack-of-fit are not significant  (p >
0.05). The response surfaces using the PC1 scores of the subsets are indicated in Figure 1B.
 
 


 
 

Figure 1 - Correlation map of peak areas (A) and response surfaces for the two subset scores (B)
 
 

The new approach introduced in this work using  PCA  and  RSM  is a versatile and interpretable procedure to optimize the
extraction of desired volatile compounds in coffee samples.

[1] Ferreira, M. M. C., Antunes, A. M., Melgo, M. S., Volpe, P. L., Quím. Nova, 22, 724 (1999);

[2] Teófilo, R. F., Ferreira, M. M. C., Quim. Nova, 29, 338 (2006);
 
 
 
 

132