Search

Introduction: To decrease antibiotic resistance, their use as growth promoters in the agricultural sector has been largely abandoned. This may lead to decreased health due to infectious disease or microbiome changes leading to gut inflammation. Objectives: We aimed to generate a m/z signature classifying chicken health in blood, and obtain biological insights from the resulting m/z signature. Methods: We used direct infusion mass-spectrometry to determine a machine-learned metabolomics signature that classifies chicken health from a blood sample. We then challenged the resulting models by investigating the classification capability of the signature on novel data obtained at poultry houses in previously unseen countries using a Leave-One-Country-Out (LOCO) cross-validation strategy. Additionally, we optimised the number of mass/charge (m/z) values required to maximise the classification capability of Random Forest models, by developing a novel ranking system based on combined univariate t-test and fold-change analyses and building models based on this ranking through forward and reverse feature selection. Results: The multi-country and LOCO models could classify chicken health. Both resulting 25-m/z and 3784-m/z signatures reliably classified chicken health in multiple countries. Through mummichog enrichment analysis on the large m/z signature, we found changes in amino acid metabolism, including branched chain amino acids and polyamines. Conclusion: We reliably classified chicken health from blood, independent of genetic-, farm-, feed- and country-specific confounding factors. The 25-m/z signature can be used to aid development of a per-metabolite panel. The extended 3784-m/z version can be used to gain a deeper understanding of the metabolic causes and consequences of low chicken health. Together, they may facilitate future treatment, prevention and intervention. [ABSTRACT FROM AUTHOR]

Direct infusion untargeted mass spectrometry-based metabolomics allows for rapid insight into a sample's metabolic activity. However, analysis is often complicated by the large array of detected m/z values and the difficulty to prioritize important m/z and simultaneously annotate their putative identities. To address this challenge, we developed MetaboShiny, a novel R/RShiny-based metabolomics package featuring data analysis, database- and formula-prediction-based annotation and visualization. To demonstrate this, we reproduce and further explore a MetaboLights metabolomics bioinformatics study on lung cancer patient urine samples. MetaboShiny enables rapid and rigorous analysis and interpretation of direct infusion untargeted mass spectrometry-based metabolomics data. [ABSTRACT FROM AUTHOR]