Supervised learning of gene-regulatory networks based on graph distance profiles of transcriptomics data

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Understanding gene-regulatory network (GRN) interactions is crucial for deciphering cellular phenotypes.
• Reconstructing GRNs from gene expression data remains a significant challenge in systems biology despite technological advancements.

Purpose of the Study:

• To develop a novel supervised learning approach for accurate GRN reconstruction.
• To evaluate the performance of the proposed method against existing state-of-the-art techniques.

Main Methods:

• Developed GRADIS, a supervised learning method utilizing support vector machines.
• Reconstructed GRNs based on distance profiles from graph representations of transcriptomics data.
• Validated the approach using data from Escherichia coli, Saccharomyces cerevisiae, and synthetic networks (DREAM4, network inference challenges).

Main Results:

• GRADIS demonstrated superior performance compared to state-of-the-art supervised and unsupervised methods for GRN reconstruction.
• The approach showed high accuracy in predicting target genes for individual transcription factors and the overall network structure.
• Experimental validation using known GRNs from E. coli and S. cerevisiae confirmed the predictions and provided insights into method performance.

Conclusions:

• The GRADIS approach offers a robust and effective method for gene-regulatory network inference.
• It provides a foundation for utilizing diverse network-based data representations.
• The method is extensible for characterizing other cellular networks, such as protein-protein and protein-metabolite interactions.