Consistency of biological networks inferred from microarray and sequencing data

Area of Science:

• Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Sparse Gaussian graphical models are widely used for inferring biological networks, including gene regulatory networks.
• Investigating model consistency across diverse data platforms (microarray, next-generation sequencing) is crucial for reliable biological network inference.

Purpose of the Study:

• To assess the consistency of sparse Gaussian graphical models across different data platforms.
• To identify factors affecting the reproducibility of inferred biological networks.
• To propose strategies for improving cross-platform network inference.

Main Methods:

• Utilized a rich dataset with samples profiled across multiple platforms (microarray, next-generation sequencing).
• Analyzed the impact of individual node variances on network connectivity.
• Compared network reproducibility based on individual edges versus functional group enrichment.

Main Results:

• Individual node variances significantly influence network connectivity and can lead to platform-specific inconsistencies.
• Failure to scale data prior to network analysis results in non-reproducible and potentially misleading networks.
• Network reproducibility is substantially higher when summarized by pathway enrichment compared to individual edge analysis.

Conclusions:

• Proper pre-processing of transcriptional data is essential for consistent network inference.
• Summarizing biological networks beyond individual edges, such as through pathway enrichment, enhances cross-platform consistency.
• Caution is advised in the interpretation of gene regulatory networks inferred from biological data due to potential inconsistencies.