Revealing differences in gene network inference algorithms on the network level by ensemble methods.

Area of Science:

• Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Inferring gene regulatory networks from expression data offers causal insights but faces challenges with observational data.
• Current understanding of network inference method capabilities and limitations remains incomplete.

Purpose of the Study:

• To statistically analyze and compare four network inference algorithms: ARACNE, CLR, MRNET, and RN.
• To assess the inferability of regulatory networks down to individual edges using ensemble methods.
• To provide guidance on interpreting inferred networks and understanding their biases.

Main Methods:

• Statistical analysis of ARACNE, CLR, MRNET, and RN algorithms.
• Application of ensemble methods for detailed edge-level inferability assessment.
• Investigation of local network-based measures to differentiate algorithm performance.

Main Results:

• Identified differences and similarities among the four network inference algorithms.
• Revealed biases inherent in these methods concerning various network components.
• Provided a framework for interpreting inferred regulatory networks from expression data.

Conclusions:

• The study offers crucial insights into the performance and biases of common network inference algorithms.
• Findings guide the selection and interpretation of methods for analyzing gene regulatory networks.
• Predicts regulatory interactions in human B cells, hypothesizing roles for Myc and its targets.