Effectiveness of visible-Near infrared spectroscopy coupled with simulated annealing partial least squares analysis to predict immunoglobulins G, A, and M concentration in bovine colostrum

Visible - near infrared spectroscopy coupled with variable selection using simulated annealing PLS regression was tested to predict immunoglobulin fractions (g/L) of bovine colostrum, namely IgG, IgA and IgM. Immunoglobulins were quantified in 678 samples using the gold standard radial immunodiffusion. Samples were divided in calibration (50%) and validation (50%) datasets. Maximum number of selected variables were limited to 200 and root mean squared error in cross validation (RMSECV) was used as loss function. Performance of the final model developed using the calibration dataset was assessed on the validation dataset. Overall, simulated annealing PLS improved validation RMSECV compared to ordinary PLS regression by 3% to 17%. The present study demonstrated the effectiveness of the calibration model for accurate quantification of IgG, the most abundant immunoglobulin of bovine colostrum (RMSECV = 13.28 g/L; R2 = 0.83). These outcomes could be useful to assess colostrum quality intended for animal and human usage.

Automated summary

The study demonstrated the effectiveness of visible-near infrared spectroscopy coupled with simulated annealing PLS regression in predicting immunoglobulin fractions of bovine colostrum. The model developed using the calibration dataset showed improved accuracy in validation, particularly in quantifying IgG.