Semi-supervised network inference using simulated gene expression dynamics.

Area of Science:

• Systems biology
• Computational biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) inference from high-throughput data is challenging due to noise and limited sample sizes.
• Existing methods often produce networks with missing or incorrect links, hindering biological insights.
• Integrating prior network information, such as pathway databases, can enhance GRN reconstruction accuracy.

Purpose of the Study:

• To develop a semi-supervised algorithm for reconstructing gene regulatory networks.
• To effectively synthesize information from partially known networks and time-course gene expression data.
• To improve the accuracy of network inference by incorporating prior biological knowledge.

Main Methods:

• Developed a semi-supervised network reconstruction algorithm utilizing time-course gene expression data.
• Adapted partial least squares-variable importance in projection (VIP) for time-course analysis.
• Generated reference distributions from simulated expression data using prior networks to estimate edge probabilities.

Main Results:

• The algorithm successfully integrated prior network information (KEGG pathways) with gene expression data.
• Applied to sleep deprivation data and DREAM challenge datasets, the method recovered true network edges.
• Identified errors in existing networks, demonstrating the ability to derive accurate posterior networks reflecting gene expression dynamics.

Conclusions:

• The developed semi-supervised approach enhances gene regulatory network reconstruction.
• The method effectively leverages prior biological knowledge and gene expression dynamics.
• This approach offers a robust tool for discovering novel and correcting erroneous regulatory links in biological networks.