Machine Learning Study: Application for Pet Biomarker Discovery from Omics Data

The consumption of food can influence different omics levels, such as genomics, proteomics, transcriptomics, and metabolomics. It causes changes in the number of plasmatic proteins and metabolite concentrations. Additionally, it is well known that balanced dietary intakes play a critical role in maintaining gut microbial balance, which can prevent chronic and inflammatory diseases or enhance recovery from illness more effectively. Therefore, investigating various omics data could reveal new insights into the mechanisms underlying the metabolism of nutrients and health. Because of the large volume and complexity of omics data, machine learning (ML) methods have become popular alternatives to traditional statistical analyses and for predicting the relationship between nutrient intake and biological pathways. In this talk, we applied machine learning models to a publicly available metabolomics dataset to predict feline urolithiasis, a kidney stone disease in cats. Our models can detect key biomarkers from the dataset and predict kidney stone occurrence with a small sample size, enabling nutrition intervention for both pet and human health.