Investigation on the effectiveness of mid-infrared spectroscopy to predict detailed mineral composition of bulk milk

This Research Communication investigated the potential of mid-infrared spectroscopy to predict detailed mineral composition of bovine milk. A total of 153 bulk milk samples were analysed for contents of Ca, Cl, Cu, Fe, K, Mg, Na, P and Zn. Also, soluble and colloidal fractions of Ca, Mg and P were quantified. For each milk sample the mid-infrared spectrum was captured and stored. Prediction models were developed using partial least squares regression and the accuracy of predic- tion was evaluated using both cross- and external validation. The proportion of variance explained by the prediction models in cross-validation ranged from 34% (Na) to 77% (total P), and it ranged from 13% (soluble Mg) to 54% (Cl−) in external validation. The ratio of the standard deviation of each trait to the standard error of prediction in external validation, which is an indicator of the prac- tical utility of the prediction model, was low and never greater than 2. Results from the current study supported the limited usefulness of mid-infrared spectroscopy to predict minerals present in low con- centration in bulk milk. For major mineral components, results from the present research did not match previous findings demonstrating the need for further studies using larger reference datasets.