Predictions of Sampling Site Based on Microbial Compositions Using a Decision Tree-based Method

Abstract

ƒThis is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/ license/by-nc/3.0/). The nose and throat are sites commonly used to obtain swab specimens to diagnose upper respiratory tract infections, and some studies have shown differences between the diagnostic accuracies of nose and throat swabs for upper respiratory infections. However, current sampling methods for the diagnosis of upper respiratory tract infections do not differentiate between nose and throat samples. The present study was undertaken to devise a means of determining whether samples were obtained from the nose or throat. Microbiome abundance data of 576 upper respiratory swab samples were obtained from the human microbiome project website. Predictive models were generated to determine sampling sites based on microbiomes using the random forest and regression tree with recursive partitioning methods. The final prediction model showed a near-perfect prediction for sampling sites using only the abundances of Staphylococcaceae and Streptococcaceae. The devised model can be used to predict sampling sites for upper respiratory specimens.