Can Machine Learning Make Fecal Testing Part of CVD Screening?

Machine learning analysis of stool samples may provide a helpful first pass for the mass screening for any type of cardiovascular disease (CVD) in patients, researchers claimed. Various machine learning algorithms were fed gut microbiota data and, with training, were subsequently able to distinguish between people with and without CVD, with ROC curves as high as 0.70, reported a group led by Sachin Aryal, an MS student in bioinformatics at the University of Toledo, Ohio. "While this demonstrates the promising potential of applying microbiome-based ML [machine learning] for predicting CVD, in the future, it will be of interest to further calibrate and improve predictive capability of ML modeling by including more samples from different sources or stratifying specific types of CVD incorporated with combinatorial features such as health records, in addition to gut microbiome data," the authors said. Their study was presented as a poster at the virtual Hypertension meeting, sponsored by the American Heart Association, and was simultaneously published online in the November 2020 issue of Hypertension. Investigators claimed theirs as the first study to apply existing knowledge of dysbiosis of gut microbiota in CVD patients to a machine learning approach to CVD screening.