Easy and fast prediction of sow milk immunoglobulins and main milk components using mid-infrared analysis of milk and colostrum


  • Grelet, C. , Leblois, J. , Dufourny, S. , Arévalo Sureda, E. , Schroyen, M. , Wavreille, J. & Dehareng, F. (2025). Easy and fast prediction of sow milk immunoglobulins and main milk components using mid-infrared analysis of milk and colostrum. Poster in: EAAP 2025, Innsbruck, Austria, 27/08/2025.
Type Poster
Year 2025
Title Easy and fast prediction of sow milk immunoglobulins and main milk components using mid-infrared analysis of milk and colostrum
Event name EAAP 2025
Event location Innsbruck, Austria
Event date 27/08/2025
Endnote keywords Transfer, mid-infrared, sow, milk
Abstract Quality of sow colostrum and milk is an essential element regarding piglets immunity and growth. Among others, the content in immunoglobulins in colostrum is directly impacting piglets immunity in the first weeks of life, thereby impacting their intestinal health at weaning. In addition, the global composition of milk, through its fat, protein and lactose content is the main source of energy transfer from the sow to the piglets. Therefore, even if milk sampling is laborious, in research frame there is an interest in measuring colostrum and milk composition, but this usually implies reference chemical methods which are tedious and expensive. The objective of this study is to derive a key method in dairy animals to be used with sow milk: the mid-infrared (MIR) analysis of milk. Milk analysis of milk requires only a few seconds to run, is cost-effective and non-destructive. For this, 37 sows from CRA-W experimental farm were milked from week 0 to 3 after farrowing, and milk was analyzed both in chemical reference methods and in MIR spectrometry. First, an attempt was made to transfer the existing models predicting fat and protein for dairy cows, to be used for predicting sow milk main components. Second, new MIR models were developed to attempt predicting the immunoglobulins A, G and M contents in colostrum, in milk or in a common model using both colostrum and milk. Results show that cow MIR models were perfectly linear when predicting sow milk fat and protein, with an R² of 0.99, but where biased and needed to be adjusted for slope and bias. Then MIR models to predict immunoglobulins provided very heterogeneous results, with R² ranging from 0 to 0.93 depending of the model (colostrum, milk or both) and the molecules (immunoglobulins A, G, M or total). These results are only preliminary findings, as a proof of concept with a limited dataset. However, they show that models from cow milk can be easily transferred to sow milk for accurate estimation of fat and protein. The preliminary findings show that high IG contents (colostrum) may be easily discriminated from low IG contents (milk) with MIR. However, while some R²cv appear good (>0.9), the very high relative errors of models make them not suitable for quantitative screening.
Fichier
Authors Grelet, C., Leblois, J., Dufourny, S., Arévalo Sureda, E., Schroyen, M., Wavreille, J., Dehareng, F.

Equipe