Evaluating changes in attractor sets under small network perturbations to infer reliable microbial interaction...

Area of Science:

• Computational Ecology
• Systems Biology
• Bioinformatics

Background:

• Microbiome data analysis is crucial for understanding microbial interactions.
• Current inference methods often fail to reproduce stable microbial community states (attractors).
• Boolean networks offer an efficient framework for network inference but require improvement.

Purpose of the Study:

• To develop a network inference algorithm that accurately reproduces attractors in Boolean dynamics.
• To enhance existing inference methods by leveraging attractor dynamics under network perturbations.
• To investigate differences in microbial interaction networks across health conditions.

Main Methods:

• Studied attractor changes in Boolean threshold dynamics on signed undirected graphs.
• Developed the EDAME algorithm to refine networks inferred by methods like ESABO.
• Applied the method to microbial abundance data from human stool samples (IBD, colorectal cancer, healthy).

Main Results:

• Demonstrated how small network changes affect attractors, enabling algorithmic enhancement.
• Showcased EDAME's ability to generate networks closely matching original attractors.
• Revealed significant diversity in microbial interaction networks of individuals with Inflammatory Bowel Disease (IBD).

Conclusions:

• The EDAME algorithm improves microbial network inference by accurately capturing system attractors.
• Network inference methods can be enhanced by analyzing attractor stability under network perturbations.
• Significant differences exist in microbial interaction networks between IBD patients and healthy individuals.