networkBMA

networkBMA infers gene regulatory networks from time-series gene expression data using Bayesian Model Averaging (BMA) approaches to identify regulatory relationships in genomic datasets, including mammalian systems.

Key Features:

• Bayesian inference (ScanBMA): Implements ScanBMA within a Bayesian Model Averaging framework and integrates external data sources to refine inferred gene-to-gene relationships.
• Efficient model-space search: Employs a strategy for efficient navigation of the model space tailored to large-scale genomic datasets.
• Data transformation techniques: Applies specific transformations to mitigate spurious associations and enhance robustness of network predictions.
• g-prior integration: Uses a g-prior to improve identification of potential gene regulators.
• Scalability and computational efficiency: Addresses scalability challenges to process extensive time-series datasets with high computational efficiency.
• Benchmark performance: Demonstrates generation of compact networks with a higher proportion of true positives and superior Receiver-Operating Characteristic (ROC) and Precision-Recall performance in benchmark comparisons.

Scientific Applications:

• Time-series GRN inference: Reconstruction of gene regulatory networks from time-course gene expression data.
• Mammalian genome-wide analysis: Inference of regulatory interactions in mammalian genome-wide datasets.
• Systems biology and functional genomics: Discovery of complex regulatory interactions for systems-level and functional genomic studies.
• Regulator identification: Prioritization of candidate gene regulators using model-based priors.
• Benchmarking and validation: Comparative evaluation using yeast time-series data and DREAM simulated datasets for method assessment.

Methodology:

Uses Bayesian Model Averaging (BMA) via ScanBMA with integration of external information, an efficient model-space search strategy, data transformation steps to reduce spurious associations, and a g-prior, with computational strategies to address scalability for time-series expression data.