MIDER: network inference with mutual information distance and entropy reduction.

Area of Science:

• Bioinformatics and Systems Biology
• Information Theory
• Network Science

Background:

• Network inference, or reverse engineering, is crucial for understanding complex biological systems.
• Existing information-theoretic methods often have limited generality and lack public implementations.
• Accurate network inference requires methods that can handle time delays and distinguish direct/indirect interactions.

Purpose of the Study:

• To present MIDER, a novel information-theoretic method for inferring general-purpose network structures.
• To provide a robust and versatile tool for network inference applicable to diverse biological and non-biological systems.
• To offer a publicly available implementation of an advanced network inference technique.

Main Methods:

• MIDER utilizes information-theoretic concepts, including mutual information, for network inference.
• The method represents networks based on statistical closeness of nodes and refines link predictions.
• It processes time-series data, accounts for time delays, and allows flexible mutual information definitions.

Main Results:

• MIDER was evaluated on seven benchmark problems across metabolic, gene regulatory, and signaling networks.
• Performance comparisons demonstrated MIDER's competitive and versatile capabilities against state-of-the-art methods.
• The method successfully distinguishes direct from indirect interactions and assigns directionality.

Conclusions:

• MIDER provides a powerful and generalizable approach to network inference using information theory.
• Its freely available implementation and adaptive nature facilitate broad application without requiring prior user knowledge.
• The method advances the field of systems biology by offering a reliable tool for uncovering complex network architectures.