GeneNetWeaver: in silico benchmark generation and performance profiling of network inference methods

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Accurate evaluation of gene regulatory network inference methods is challenging due to inadequate benchmarks and analysis tools.
• Developing robust benchmarks is crucial for advancing gene expression data analysis.

Purpose of the Study:

• To introduce GeneNetWeaver (GNW), a comprehensive open-source software for generating in silico benchmarks and profiling network inference methods.
• To provide a tool for detailed analysis of network inference predictions and identification of systematic errors.

Main Methods:

• Development of GeneNetWeaver (GNW) for creating dynamical models of gene regulatory networks as benchmarks.
• Implementation of network motif analysis within GNW to identify prediction errors.
• Utilizing standard metrics like precision-recall and ROC curves for performance evaluation.

Main Results:

• GNW facilitates the assessment of six different network inference methods, highlighting their strengths and weaknesses.
• The software was used to generate challenges for the Dialogue for Reverse Engineering Assessments and Methods (DREAM) competitions (DREAM3, DREAM4, DREAM5).

Conclusions:

• GeneNetWeaver (GNW) provides a valuable resource for the community to systematically evaluate and improve gene regulatory network inference algorithms.
• The open-source nature of GNW promotes reproducible research and collaborative development in systems biology.