Raw cocoa has an astringent, unpleasant taste and flavour, and has to be fermented, dried and roasted in order to obtain the characteristic cocoa flavour and taste. During the fermentation microbial activity outside the cocoa beans induces biochemical and physical changes inside the beans. The process is complex involving activity of several different groups of microorganisms which bring about numerous biochemical and physical changes inside the beans. Due to the complexity of these processes no thorough investigations of the interactions between the microbial activities on the outside of the beans and the chemical processes inside the beans have been carried out previously.

Recently it has been shown that Denaturing Gradient Gel Electrophoresis (DGGE) offers an efficient tool for monitoring the microbiological changes taking place during the fermentation of cocoa. Near Infrared (NIR) spectroscopy has previously been used to determine various components in cocoa beans, offering a rapid alternative compared to traditional analytical methods for obtaining knowledge about changes in the chemical composition of the cocoa beans during fermentation.

During a number of cocoa fermentations bean samples were taken with 24 h intervals to be dried and analysed by NIR. Cocoa pulp samples taken simultaneously during the same fermentations have previously been characterised using DGGE [Nielsen, D.S., Teniola, O.D., Ban-Koffi, L., Owusu, M., Andersson, T., Holzapfel, W.H. (2007). The microbiology of Ghanaian cocoa fermentations analysed using culture dependent and culture-independent methods. International Journal of Food Microbiology 114, 168-186.]. Here we report the first study where microbiological changes during the fermentation determined using DGGE are correlated to changes inside the beans determined by NIR using multivariate data analysis.

Following data pre-processing (baseline correction followed by Co-shift correction or Correlation Optimised Warping) the DGGE spectra were analysed using Principal Component Analysis (PCA). A clear grouping according to fermentation time was seen demonstrating the microbial succession taking place during the fermentation. Subsequently the DGGE spectra were correlated to the NIR spectra using Partial Least Squares regression models (PLS2). Correlations of 0.87 (bacterial derived DGGE spectra) and 0.81 (yeast derived DGGE spectra) were obtained indicating the relationship between the microbial activities in the pulp and the (bio)chemical changes inside the beans. By comparing the X-block loadings of the PLS2 models and the DGGE spectra it was possible to directly link several microbial species with changes in the NIR spectra and consequently also with changes inside the beans.