Early prediction of paratuberculosis health status in cattle by integrating NMR analysis and blood indices through the LASSO method

Abstract

Abstract INTRODUCTION: The early diagnosis of Mycobacterium avium subsp. paratuberculosis (MAP) is one of the current challenges of farmers and veterinarians. OBJECTIVES: This work aimed to investigated the changes in metabolic levels associated with natural MAP infection in infected and infectious dairy cattle METHODS: The study included sera from 23 infectious/seropositive, 10 infected but non-infectious/seronegative, and 26 negative animals. The samples were selected from a collection of samples gathered during a prospective study. The samples were analyzed by quantitative nuclear magnetic resonance (NMR) spectroscopy and routine blood chemistry. The blood indices and the 1H NMR data were concatenated by low-level data fusion, resulting in a unique global fingerprint. Afterwards, the merged dataset was statistically analyzed by the least absolute shrinkage and selection operator (LASSO), which is a shrinkage and selection method for supervised learning. Finally, pathways analysis was performed to get more insights on the possible dysregulated metabolic pathways.RESULTS: The LASSO model achieved, in cross-validation, an overall accuracy of 91.5% with high values of sensitivity and specificity in classifying correctly the negative, infected, and infectious animals. The pathway analysis revealed MAP-infected cattle have increased tyrosine metabolism and enhanced phenylalanine, tyrosine and tryptophan biosynthesis. The enhanced synthesis and degradation of ketone bodies was observed both in infected and infectious cattle.CONCLUSION: In conclusion, fusing data from multiple sources has proved to be useful in exploring the altered metabolic pathways in MAP infection and potentially predicting negative animals within paratuberculosis-infected herds.