Diet Prediction and Feature Importance of Gut Microbiome using Machine Learning

Abstract

The microbiome is the genetic material of all the microbes, bacteria, fungi, protozoa and viruses that live on and inside the human body. The most exciting venture of present-day medicine is determining the microscopic dwellers of every organ and restoring them in balanced proportion when it gets disrupted. It is challenging to conclude whether disease or disrupted microbiota occurred first based on observation of health problems. By collecting the microbiome data and analysing it, precise medicines can be given to the patients. Precision Medicine depends on microbiome composition inside a patient. There are various medicines used to cure the specific disease. Medicines reacts differently on every human being, one of the reason is the difference in the microbiome composition of the person. If one is able to analyse the composition of microbiome, the medicine apt can even be predicted to cure that disease. Diet has abundant influence on once gut microbiome population. This work analyses human gut microbiome data to find the microbiomes which has more influence on human health from different diet pattern scrutinized on humanized gnotobiotic mice. Machine Learning (ML) techniques like Logistic Regression (LR), Random Forest (RF) and Artificial Neural Network (ANN) have been applied on an accessible data set from MINE (Maximal Information-based Non-parametric exploration). By using Machine Learning, the type of diet can be predicted from microbiome composition with great accuracies as well as diet sensitive microbiomes are found which are very less compared to a huge number of microbiomes in data set.