Milk lipidome alterations in first-lactation dairy cows with lameness: A biomarker identification approach using untargeted lipidomics and machine learning
Format of work:
Journal Article
Event presented at / Journal Name:
Journal of Dairy Science
Speaker / Contact Author's Name:
Ana S. Cardoso
Speaker / Contact Author's E-mail Address:
ana.ferreira1@nottingham.ac.uk
-
Research aim:
The study set out to determine whether milk lipidomic profiles could accurately predict early lameness in first‑lactation dairy cows, and to identify specific lipid biomarkers linked to disease onset. It also aimed to explore disrupted metabolic pathways to better understand the biological processes underlying lameness. The overarching goal was to support the development of early‑detection tools that enhance cow welfare and farm productivity.
-
Background:
Lameness, a painful condition that impairs gait, is a major welfare and productivity issue in dairy cows and increases the risk of premature culling. Traditional detection relies on visual mobility scoring, which identifies lameness only after substantial damage has occurred. Early detection is essential to improve outcomes. Although biomarker‑based diagnostics are common in human medicine, they remain underused in livestock. Milk lipidomics offers a noninvasive approach and has shown promise for detecting health disorders. This study builds on earlier work exploring metabolic changes to identify lipid biomarkers for early lameness detection.
-
Approach:
The study applied untargeted milk lipidomics and machine‑learning methods to identify early biomarkers of lameness in first‑lactation dairy cows. Milk samples from 32 cows (16 lame, 16 controls) were collected after calving and at lameness onset. Lipid profiles were analysed using liquid chromatography–high‑resolution mass spectrometry. Multiple statistical approaches, including univariate and multivariate analyses, five machine‑learning models, stability selection, and Cox–Battey regression, were used to identify discriminative mass ions. Functional pathway analysis was then performed to explore metabolic pathways associated with lameness.
-
Key finding:
Using advanced milk testing and machine‑learning models, scientists were able to predict lameness with up to 100% accuracy at onset and over 80% accuracy using samples collected after calving. The study uncovered early disruptions in key metabolic pathways, including retinol metabolism, and highlighted specific milk lipids linked to inflammation. These results suggest that routine milk screening could offer a simple, noninvasive tool to improve early detection, welfare, and farm productivity.
-
Industry or policy relevance:
Lameness remains one of the most costly and persistent welfare challenges in dairy farming; yet, current detection methods identify cases only after significant damage has already occurred. This study demonstrates that milk lipidomics could offer a practical, noninvasive tool for early lameness detection, enabling farmers to intervene sooner and reduce long‑term welfare and productivity losses. Early‑warning biomarkers have the potential to support precision‑livestock technologies, strengthen welfare assurance schemes, and inform evidence‑based policy aimed at reducing lameness prevalence. Adoption of such tools could improve herd health, farm sustainability, and public confidence in dairy production.
-
Route for practical application:
The identification of milk‑based lipid biomarkers creates a pathway toward developing rapid, on‑farm screening tools for early lameness detection. These biomarkers could be integrated into existing milk‑recording systems, automated milking stations, or precision‑livestock platforms to provide real‑time alerts before visible gait changes occur. With further validation in larger and more diverse herds, the biomarkers could support commercial diagnostic kits or algorithm‑based monitoring software. Embedding this approach into routine herd‑health programmes would enable earlier treatment, reduce chronic lameness, and improve welfare, productivity, and farm sustainability.
-
Confidence in findings and next steps towards realising impact:
The study’s findings are strengthened by the use of multiple complementary analytical approaches, including untargeted LC‑HRMS, five machine‑learning models, stability selection, multivariate analysis, and the Cox–Battey method. The repeated identification of a small set of discriminative mass ions across methods increases confidence in their potential relevance. Elastic net regression showed the highest predictive performance, with 83% accuracy after calving and 100% accuracy at lameness onset. Functional pathway analysis supported these results by revealing dysregulation of retinol metabolism before clinical signs appeared. However, the authors note key limitations: the study involved a single herd of first‑lactation cows, and putative lipid identifications require confirmation. Next steps include validating biomarkers in larger, diverse herds and confirming lipid identities using targeted approaches.
Funders:
Funded by BBSRC iCASE BB/T0083690/1, AHDB, and the University of Nottingham’s School of Veterinary Medicine and Science.
