BETS

BETS infers directed gene regulatory networks from transcriptional time-series data to identify causal activating and inhibitory relationships among genes using Granger causality principles.

Key Features:

• Elastic net with stability selection: Combines elastic net regression and stability selection across bootstrapped samples to determine causal relationships among genes.
• Granger causality framework: Applies principles of Granger causality to infer directed temporal causal relationships from time-series expression data.
• Sign inference: Distinguishes activating versus inhibitory influences in inferred edges.
• Parallelization: Implements a highly parallelized workflow for analysis of large transcriptional datasets.
• Benchmark performance: Demonstrated competitive speed and precision on the DREAM4 100-gene network inference challenge.
• False discovery rate reporting: Reports inferred networks with controlled FDR, exemplified by results at FDR ≤0.2.

Scientific Applications:

• A549 glucocorticoid time-series: Applied to A549 cells exposed to glucocorticoids over 12 hours, inferring a network of 2768 genes and 31,945 directed edges at FDR ≤0.2.
• Experimental validation: Inferred causal relationships validated using overexpression experiments within the same glucocorticoid system.
• Genetic association validation: Inferred edges were supported by genetic variants associated with the edges in primary lung tissue from GTEx v6.

Methodology:

Uses elastic net regression combined with stability selection across bootstrapped samples and applies Granger causality principles; the implementation is highly parallelized.