A novel near infrared spectroscopy analytical strategy for meat and bone meal species discrimination based on the insight of fraction composition complexity

This study analyzed the meat and bone meal (MBM) matrix complexity from the perspective of fraction composition diversity and a classification strategy was proposed to accurately and rapidly identify the MBM species based on near infrared spectroscopy (NIRS). Partial Least Squares-Discrimination Analysis (PLS-DA) based on full samples, meat meal (MM), MBM and bone meal (BM) performed with decreasing classification errors of 0.115, 0.079, 0.044 and 0.039 which were partly caused by wide sample range; bone fraction content had positive correlation with most of MBM species differences reflected by principal component scores; and PLSDA classification errors among MM, MBM and BM were lower than 0.013. To take fully advantage of the above results, a sequential classification strategy was proposed; near infrared spectra were selected (belong to MM, MBM or BM) and then species discrimination analysis was conducted based on the specific PLS-DA model.

Automated summary

This study analyzed the complexity of meat and bone meal (MBM) and proposed a classification strategy to accurately and rapidly identify MBM species. Using Partial Least Squares-Discrimination Analysis (PLS-DA) based on full samples, lower classification errors were achieved, and bone fraction content showed positive correlation with most MBM species differences.