SMARTS

SMARTS reconstructs condition-specific regulatory response networks by integrating static and time series expression data from multiple individuals to model population heterogeneity and identify transcription factor drivers.

Key Features:

• Data integration: Integrates static and time series expression data from multiple individuals to reconstruct condition-specific response networks.
• Population heterogeneity modeling: Accounts for variation across individuals such as baseline expression differences, age-related and genetic variability, and life-experience–driven heterogeneity.
• Temporal variability handling: Accommodates differences in individual start times, end times, and progression rates within time series datasets.
• Probabilistic graphical models: Employs probabilistic graphical models as the core algorithmic framework.
• Iterative reconstruction and assignment: Iteratively reconstructs regulatory networks while simultaneously assigning individuals to these networks.
• Unsupervised analysis: Operates without predefined labels, autonomously identifying patterns and groupings in the data.
• Patient clustering: Clusters patients into distinct groups based on response profiles derived from time series data.
• Transcription factor identification: Pinpoints transcription factors (TFs) that differentiate response groups.

Scientific Applications:

• Patient stratification: Stratifies patients into response-based groups to reveal population-level heterogeneity in dynamic responses.
• Regulatory driver discovery: Identifies transcription factors associated with differential dynamic responses.
• Human influenza response analysis: Applied to human responses to influenza to improve baseline groupings and identify relevant TFs.
• Mouse brain development studies: Applied to mouse brain development to enhance grouping and propose regulatory hypotheses.

Methodology:

Applies probabilistic graphical models in an unsupervised framework to iteratively reconstruct regulatory networks while simultaneously assigning individuals to networks, integrating static and time series expression data and accommodating variation in baseline expression, start/end times, and progression rates.