TDARACNE

TDARACNE infers gene regulatory networks from time-series expression data by extending the ARACNE algorithm to detect time-delayed dependencies in time-course profiles.

Key Features:

• Time-Delayed Dependencies: Detects dependencies between genes at different time delays using mutual information to capture temporal gene regulation.
• Probabilistic Model: Assumes an underlying stationary Markov Random Field probabilistic model for gene interactions.
• Filtering Mechanism: Applies an auto-calculated threshold to filter out less informative dependencies.
• Network Representation: Represents inferred gene regulatory networks as directed graphs indicating directional interactions.
• Handling Small Networks: Reconstructs small local networks, identifying feed-forward and cyclic interactions from medium-small numbers of measurements.

Scientific Applications:

• Synthetic benchmark networks: Tested on synthetic networks for method validation.
• Saccharomyces cerevisiae cell cycle: Applied to microarray time-course expression profiles from the S. cerevisiae cell cycle.
• Escherichia coli SOS pathways: Applied to microarray expression data from E. coli SOS pathways.
• In vivo reverse-engineering benchmark: Evaluated on a network developed for assessing reverse engineering algorithms in vivo.
• Comparative performance: Compared with ARACNE, Dynamic Bayesian Networks, and systems of ordinary differential equations (ODEs), demonstrating good accuracy, recall, and F-score.

Methodology:

Extends the ARACNE algorithm to time-course data, computes mutual information across time delays, assumes a stationary Markov Random Field, applies an auto-calculated threshold to retained dependencies, and outputs directed graph representations.