Netter

Netter re-ranks links between genes using confidence scores from various network inference methods together with structural network properties to improve rankings of regulatory interactions for gene network inference.

Key Features:

• Re-ranking Algorithm: Re-ranks gene links using confidence scores from various network inference methods and incorporates structural properties such as graphlets and other graph-invariant characteristics.
• Flexibility and Applicability: Accepts any input ranking of regulatory interactions derived from different network inference methodologies and is applicable to diverse omics data analyses.
• Improvement Across Benchmarks: Improves predictions on multiple benchmarks, including artificially generated datasets and the DREAM4 and DREAM5 challenges, notably enhancing the E.coli community prediction in DREAM5.
• Comparison with Other Algorithms: Shows favorable performance relative to other post-processing algorithms and is not limited to correlation-like predictions.
• Robust Performance: Maintains consistent effectiveness across a wide range of parameter settings.
• User Customization: Supports customization via user-defined graph properties to integrate specific prior knowledge into the re-ranking process.

Scientific Applications:

• Gene regulatory network inference: Serves as a second-step re-ranking method to enhance prediction quality of inferred gene regulatory networks.
• Systems biology: Improves modeling accuracy in systems biology by integrating structural network insights.
• Genomics: Refines regulatory interaction predictions from expression-based network inference in genomics analyses.
• Personalized medicine: Enables more precise models of genetic interactions to support applications in personalized medicine.

Methodology:

Re-ranks regulatory interaction predictions by combining confidence scores from network inference methods with structural network properties, specifically graphlets and other graph-invariant characteristics.