Model recognition and prediction abilities were defined as the percentage of members of the calibration and evaluation sets that were correctly classified, respectively. The statistical package XLSTAT Sensory 2010 (Addinsoft, New York) was employed for all the chemometric calculations. Average spectra obtained for roasted coffee, coffee husks and corn samples are shown in Fig. 1. A comparative evaluation of the average data indicates somewhat similar spectra, with most SP600125 clinical trial of the significant bands concentrated in

the following ranges: 3000–2800 and 1800–700 cm−1. In general, absorbance values were higher for coffee and lower for corn. Two sharp bands at 2923 and 2854 cm−1 can be clearly seen in the spectrum corresponding to roasted coffee. Such bands have been previously reported present in spectra of roasted Arabica and Robusta coffee samples (Craig et al., 2012b; Kemsley et al., 1995) and also of crude coffee samples (Craig et al., 2011, 2012a). Paradkar and Irudayaraj (2002) also reported two sharp peaks

at 2882 and 2829 cm−1 in samples of caffeinated beverages such as coffee, tea and soft drinks. The band at 2829 cm−1 was attributed to stretching of C–H Obeticholic Acid order bonds of methyl (–CH3) group in the caffeine molecule, being successfully used to develop predictive models for quantitative analysis of caffeine (Paradkar & Irudayaraj, 2002). The same bands can be identified in the spectra obtained for roasted coffee husks and roasted corn at 2923 and 2854 cm−1 and at 2925 and 2848 cm−1, respectively. Both bands present lower absorbance values in the spectra obtained for coffee husks and corn compared to coffee. Furthermore, the second band is less evident in coffee husks and corn in comparison to coffee. Coffee husks have been reported to present similar levels of caffeine (∼1 g/100 g dry basis) in comparison to coffee beans,

whereas corn does not contain any caffeine. Other FTIR studies on corn and corn flour have also reported two bands at 2927–2925 and 2855 cm−1, being respectively attributed to asymmetric and symmetric C–H stretching in lipids (Cremer & Kaletunç, 2003; Greene, Gordon, Jackson, & Bennett, 1992). Although the samples in those studies were not submitted to roasting, the lipids content is not expected to vary during Rho roasting of corn, as it is known to occur with coffee, and the peak assignment to C–H stretching in lipids might still be valid. Furthermore, the reported amounts of lipids (Gouvea, Torres, Franca, Oliveira, & Oliveira, 2009; Moreau, 2002; Oliveira, Franca, Mendonça, & Barros-Junior, 2006) present in coffee husks (1.5–3 g/100 g) are quite low in comparison to coffee beans (12–16 g/100 g) and corn kernels (3–5 g/100 g). Therefore, such bands may be affected by both caffeine and lipids levels in the case of coffee, and are most likely primarily associated to caffeine in the case of coffee husks and only to lipids in the case of roasted corn.