Microbial Gene Ontology informed deep neural network for microbe functionality discovery in human diseases.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• The human microbiome is integral to health and disease, but its functional roles in various pathologies are challenging to determine.
• High dimensionality of metagenomic data and limited interpretability of existing models hinder the exploration of microbe functionality in human diseases.

Purpose of the Study:

• To develop and validate a novel neural network model that integrates Gene Ontology (GO) relationships to elucidate microbe functionality in human diseases.
• To identify and visualize key microbial genes and their associated functions contributing to disease states.

Main Methods:

• A neural network model incorporating a Gene Ontology (GO) relationship network was developed.
• The model was applied to four benchmark datasets: diabetes, liver cirrhosis, inflammatory bowel disease, and colorectal cancer.
• Gene and GO term importance scores were calculated for network-based discovery and visualization.

Main Results:

• The model successfully identified and visualized novel candidate microbiome genes and their functions.
• The approach provided a higher degree of biological interpretability compared to non-GO-informed models.
• The model demonstrated competitive performance in disease prediction across the tested datasets.

Conclusions:

• The proposed model offers a powerful approach to uncover microbe functionality in human diseases.
• The identified microbial genes and functions provide novel insights into the contribution of the microbiome to disease pathogenesis.
• This work advances our understanding of the human microbiome's role in health and disease.