Integrated transcriptomic and machine learning analysis of the candidate biomarker potential of ACSL6 in bovine mastitis

Bovine mastitis is a complex inflammatory disease that has a substantial negative influence on milk productivity and costs the dairy industry greatly. Identifying strong molecular biomarkers for early diagnosis is still a key area of research. In this study, integrated analyses of transcriptomic data and machine-learning approaches were used to identify candidate bovine mastitis-associated genes. The GSE15020 and GSE24217 data sets were used as discovery cohorts, while the third dataset GSE50685 served as an independent validation set. Differential gene expression analysis following quality control, normalization and batch effect adjustment using the ComBat method identified 1143 significant genes in the discovery cohort. Candidate biomarkers were subsequently ranked and selected using Random Forest and XGBoost algorithms. All candidate genes were evaluated in an independent dataset, where ACSL6 showed the highest discriminative power (p < 0.001) relative to all other candidates and PTX3 also showed potential as a good classifier. The robustness of the model was evaluated using ROC curve, confidence intervals, and permutation tests. Functional enrichment analysis showed that the DEGs were mainly involved in immune response, inflammatory pathways and defense response. Despite the limited sample size of the validation cohort, findings of present study support ACSL6 as a potential biomarker for bovine mastitis. However, larger independent cohorts and experimental validation are needed to support its diagnostic relevance and biological significance. This analysis collectively demonstrates that the combination of transcriptomic profiling and machine learning is a valuable means to identify candidate biomarkers for bovine mastitis.