Network aggregation improves gene function prediction of grapevine gene co-expression networks.

Area of Science:

• Plant Genomics
• Bioinformatics
• Functional Genomics

Background:

• Gene co-expression network (GCN) analysis is widely used in plants for gene function prediction.
• Grapevine GCN analysis shows promise but lacks measurable progress.
• Accumulated transcriptome datasets enable meta-analysis approaches.

Purpose of the Study:

• To explore how meta-analysis through aggregation influences the functional connectivity of grapevine gene co-expression networks.
• To evaluate the performance of aggregated networks compared to individual networks.
• To demonstrate the utility of aggregated networks for gene function prediction in grapevine.

Main Methods:

• Utilized 33 grapevine microarray experiments (1359 samples) for network aggregation.
• Employed guilt-by-association neighbor voting and functional enrichment metrics for performance evaluation.
• Assessed two annotation schemes (MapMan BIN, Pfam) and two sparsity thresholds (top 100, 300).

Main Results:

• Network aggregation dramatically improved functional connectivity and outperformed individual networks.
• Network size and sparsity were key performance factors; annotation scheme had minimal impact.
• The aggregate microarray network surpassed state-of-the-art microarray and RNA-seq networks in predictive performance.

Conclusions:

• Aggregating grapevine gene co-expression networks enhances robustness and functional connectivity.
• Network aggregation offers a powerful strategy for improving gene function prediction in grapevine.
• Publicly available aggregate networks (VTC-Agg) facilitate future functional genomics studies.