CTD

CTD identifies subsets of nodes in weighted graphs that exhibit significant connectedness to detect and interpret perturbed molecular modules in transcriptomics and metabolomics.

Key Features:

• Connected-subgraph detection: Identifies subsets of nodes within weighted graphs that exhibit significant connectedness, where nodes represent genes or metabolites and edges denote interactions or correlations.
• Information-theoretic approach: Uses an information-theoretic framework to detect highly connected subgraphs.
• Data-compression p-value bounds: Leverages data compression techniques to provide upper bounds on p-values for connectivity scores.
• Permutation-free validation: Avoids computationally intensive permutation testing for statistical validation of connectivity scores.
• Flexible scoring functions: Supports tailored scoring functions, including formulations for the Maximum Clique problem and edge-weighted scoring mechanisms.
• Computational efficiency: Provides a fast algorithm suited to large datasets and analyses when only a small fraction of nodes are considered.
• GMRF integration: Interprets results within disease-specific Gaussian Markov Random Field networks constructed from molecular profiling data.

Scientific Applications:

• Metabolomics — inborn errors of metabolism: Interprets multi-metabolite perturbations associated with inborn errors of metabolism.
• Transcriptomics — breast cancer: Interprets multi-transcript perturbations linked to breast cancer.
• Network analysis in omics: Detects and characterizes perturbed molecular modules in transcriptomics and metabolomics datasets.

Methodology:

Employs an information-theoretic algorithm and data-compression techniques to detect highly connected subgraphs and provide upper bounds on p-values without permutation testing; supports tailored scoring functions (e.g., Maximum Clique and edge-weighted) and operates within disease-specific Gaussian Markov Random Field networks constructed from molecular profiling data.