Breeding energy cane for cellulosic biofuel production involves manipulating various traits. An important trait to optimize is cell wall degradability as defined by enzymatic hydrolysis. We investigated the feasibility of using near-infrared spectroscopy (NIRS) combined with multivariate calibration to predict energy cane cell wall digestibility based upon fiber samples from a range of sugarcane genotypes and related species. These samples produced digestibility values ranging between 6 and 31%. To preserve the practicality of the technique, spectra obtained from crudely prepared samples were used. Various spectral pre-processing methods were tested, with the best NIRS calibration obtained from second derivative, orthogonal signal-corrected spectra. Model performance was evaluated by cross-validation and independent validation. Large differences between the performance results from the two validation approaches indicated that the model was sensitive to the choice of test data. This may be remedied by using a larger calibration training set containing diverse sample types. The best result was obtained through independent validation which produced a R2 value of 0.86, a root mean squared error of prediction (RMSEP) of 1.59, and a ratio of prediction to deviation (RPD) of 2.7. This study has demonstrated that it is feasible and practical to use NIRS to predict energy cane cell wall digestibility.
Barrie Fong Chong and Michael G. O'Shea, "Advancing Energy Cane Cell Wall Digestibility Screening by Near-Infrared Spectroscopy," Appl. Spectrosc. 67, 1160-1164 (2013)