Reverse network diffusion to remove indirect noise for better inference of gene regulatory networks

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding gene interactions, but inferring them from multi-omics data is challenging.
• Current GRN inference methods often produce false positives due to indirect correlation effects and noise, obscuring true biological relationships.
• Accurate GRN inference is essential for downstream analyses like identifying functional modules and disease-related genes.

Purpose of the Study:

• To address the limitations of current GRN inference methods, we developed a novel network denoising approach.
• The primary goal is to improve the accuracy and reliability of GRNs by effectively removing spurious edges caused by indirect effects.
• To enhance the signal-to-noise ratio in inferred gene networks for more robust biological interpretation.

Main Methods:

• We introduced REverse Network Diffusion On Random walks (RENDOR), a novel method for denoising gene regulatory networks.
• RENDOR models higher-order indirect gene interactions using transitive closure and eliminates false positives via inverse network diffusion.
• The method takes noisy networks as input and outputs refined, more accurate GRNs.

Main Results:

• Comparative assessments on simulated and real GRN data demonstrated that RENDOR significantly enhances network accuracy.
• The denoised networks derived from RENDOR more effectively capture true gene interactions compared to original inferred networks.
• Our findings highlight the importance of removing indirect noise for accurate GRN inference and validate RENDOR's effectiveness.

Conclusions:

• The proposed RENDOR method successfully denoises gene regulatory networks, improving the accuracy of inferred gene interactions.
• By mitigating the false-positive problem caused by indirect correlations, RENDOR facilitates more reliable downstream biological analyses.
• RENDOR offers a valuable tool for researchers working with multi-omics data to obtain higher-quality gene regulatory networks.