Discovering the Inter-species Interaction among Microorganisms Based on Iterative Random Forest Algorithm

EasyChair Preprint 1016

Abstract

Microorganisms are widely distributed in various parts of the human body. They interact to form a complex and diverse microbial community ecosystem. Changes in the microbial community may lead to the imbalance of the body system and even cause disease. Therefore, it is important to study the microbial composition and interaction of different populations for disease treatment and health monitoring. In this paper, we propose a microbial inter-species interaction discovery strategy using an iterative random forest (iRF) algorithm. The iRF algorithm combines a random forest (RF) algorithm with a random intersection tree (RIT) algorithm to identify stable inter-species interactions. We applied the algorithm to the microbial dataset of human intestinal cirrhosis. In the microbial data classification and prediction stage, the experiments have shown that the average AUROC of the iRF algorithm reaches 0.952, and the highest AUROC reaches 0.97, which has obvious advantages compared with other recent approaches such as SVM, RF and CNN et al.. In addition, we verified the metabolic interactions of the discovered microbial inter-species interactions, indicating that the inter-species interactions extracted by the iRF algorithm are effective.

Keyphrases: Bioinformatics, Classification, Random Forest, machine learning, microbial interaction

@booklet{EasyChair:1016,
  author    = {Xinzhe Pang and Xingpeng Jiang},
  title     = {Discovering the Inter-species Interaction among Microorganisms Based on Iterative Random Forest Algorithm},
  howpublished = {EasyChair Preprint 1016},
  year      = {EasyChair, 2019}}