Stratification of Human Gut Microiome and Building a SVM-Based Classifier

Abstract

Gut microbiome plays an important role on our health and disease development. In the recent decade, many research papers reported the correlation between alternations of microbiome pattern and the occurrence/severity of diseases. However, the human microbiome is very complex and divergent between individuals. Little is known about the whole spectrum of healthy human microbiome. Using data from human microbiome project (HMP) database (n= 325), we 1) classify sample by hierarchical clustering; 2) Identification the core taxes of each class and the differential microbe cross classes; 3) examine and compare the within-sample microbial diversity (alpha-diversity) and between-person diversity (beta-diversity); 4) built a SVM-based classifier for stool microbiome classification. The results showed that 1) healthy stool microbiome can be classified into 4 classes; 2) Firmicutes and Bacteroidete are the two dominant phyla, and Bacteroides and Prevotella are the most predominant genera. Alistipes, Oscillibacter and Ruminococcus were the major taxa in certain cases; 3) Classes were differed in their microbial composition and complexity; 4) SVM-based gut microbiome classifier yield a very good classification accuracy, sensitivity and specificity.