A Novel Approach to Differential Expression Analysis of Co-Occurrence Networks for Small-Sampled Microbiome Data

Area of Science:

• Microbiome Research
• Graph Theory
• Machine Learning

Background:

• Pharmacogenomics and cellular-level phenotypic effects are crucial for drug development and precision medicine.
• Understanding drug or disease impacts on biological networks is an accelerating research area.
• Small sample microbiome datasets present unique challenges like sparsity and compositional complexity.

Purpose of the Study:

• To propose a novel graph-theoretic approach for inferring co-occurrence networks from small sample microbiome data.
• To develop a method capable of extracting statistically significant and biologically meaningful networks from limited data.
• To demonstrate the method's applicability and extensibility to multi-omics integration.

Main Methods:

• Inference of a co-occurrence network from 16S microbiome data using graph theory.
• Enrichment and statistical filtering steps to refine inferred networks.
• Application to chicken Eimeria tenella challenge data, analyzing intestinal ecosystems across disease stages.

Main Results:

• The method successfully extracts informative networks from small datasets.
• Analysis revealed evolving node feature distributions correlating with disease progression.
• Identified clusters of mutualistic and parasitic species, and a 'persistent microbiome' sub-network.

Conclusions:

• The graph-theoretic approach effectively characterizes complex interactions in small microbiome datasets.
• The method provides biologically intuitive insights into ecosystem dynamics during disease progression.
• The approach is broadly applicable and adaptable for multi-omics data integration.