Evaluation of a hybrid near infrared reflectance spectroscopy approach to calculate nutrient digestibility of rabbit feeds

The coefficient of total tract apparent digestibility (CTTAD) of nutrients is an important measurement in rabbits as it is closely associated with feed efficiency. Near Infrared spectroscopy (NIRS) is a non-destructive method that has been used to predict not only the chemical composition of feed and feces but also the digestibility of nutrients. The prediction of digestibility is indirect, based on the assumption that the spectral data of feed and/or feces samples may contain combined chemical and physical information that enhances the quantity of information related to digestion processes. We investigated the possibility of calculating the digestibility in a two-step method by predicting the phytogenic silica (PS) marker in feed and feces with NIRS, thus, correlating the NIRS predictive ability to PS, an internal marker. A total of 27 feed samples and 282 fecal samples from an in vivo digestibility experiment were used. In addition, 43 feed samples were collected from commercial units to expand the dataset. Each data set was randomly split in calibration (n = 50 and 190 for feeds and feces, respectively) and validation sets (n = 20 and 92 for feeds and feces, respectively). First, calibrations were developed for predicting the chemical composition of feed and feces including PS. Fecal spectra were used for predicting the CTTAD of dry matter, organic matter and protein. The calibrations were evaluated internally by cross-validation; leave one out for feed and group k fold for feces and externally with an independent set. Most chemical parameters of feed and feces were predicted accurately (R²_val> 0.8). The prediction of CTTAD of nutrients from fecal spectra had a good R²_val value (> 0.75). Regarding PS marker, in feed it was predicted accurately in the independent set (R²_val= 0.75) and less accurately, yet adequately, in feces (R²_val= 0.7). The next step was to calculate the CTTAD of nutrients and then compare with those obtained by fecal NIRS. The NIRS models demonstrated accurate prediction of the PS marker and the calculated CTTADs were comparable with those of the fecal NIRS did not reveal any significant differences. In conclusion, our approach can adequately calculate the CTTAD of nutrients but more research is required to.