DDN 3.0

DDN 3.0 determines significant rewiring of biological network structures with differential dependency networks, enabling joint learning of common and rewired network structures under different conditions.

Key Features:

• Unbiased Model Estimation: Implements a weighted error measure applicable to imbalanced sample groups for unbiased model estimation.
• Accelerated Learning: Employs multiple acceleration strategies to improve computational efficiency of network learning.
• Data-Driven Hyperparameter Determination: Determines proper hyperparameters through data-driven methods rather than manual tuning.
• Joint Network Learning: Enables simultaneous learning of common and differentially rewired network structures across conditions, avoiding separate network estimation.

Scientific Applications:

• Disease Mechanism Discovery: Identifies significantly rewired molecular players within conserved regulatory circuitry potentially responsible for phenotypic transitions in complex diseases.
• Therapeutic Response Analysis: Detects significant rewiring of biological network structures to understand molecular etiology of disease progression and therapeutic response.
• Differential Network Analysis: Performs differential network analysis with improved accuracy when grouped samples are limited or less consistent.

Methodology:

DDN 3.0 uses differential dependency networks with three algorithms: unbiased model estimation with a weighted error measure, multiple acceleration strategies for learning efficiency, and data-driven hyperparameter determination.