PMCE

PMCE infers expressive probabilistic graphical models of cancer evolution from cross-sectional sequencing mutational profiles to represent alternative evolutionary routes and relate evolutionary paths to clinical outcomes.

Key Features:

• Expressive Framework: Incorporates arbitrary logical formulas that connect mutational events to represent alternative evolutionary routes and convergent evolution scenarios.
• Probabilistic Graphical Models: Constructs probabilistic graphical models to represent mutation accumulation patterns and predict evolutionary trajectories of cancer subpopulations.
• Robustness and Accuracy: Demonstrates superior accuracy and robustness to noise in simulation studies compared with state-of-the-art methods.
• Clinical Relevance: Correlates predicted evolutionary paths with overall survival across 7,866 samples from The Cancer Genome Atlas (TCGA), enabling stratification into risk groups in seven tumor types.

Scientific Applications:

• Modeling evolution dynamics: Characterizes accumulation of driver (epi)genomic alterations and alternative evolutionary routes in heterogeneous tumors.
• Resistance and relapse studies: Identifies evolutionary trajectories that underlie drug resistance and relapse by linking mutational events to selection in subpopulations.
• Prognostic stratification and biomarker discovery: Stratifies patients into risk groups and identifies biomarkers predictive of overall survival using inferred evolutionary models.
• Translational cohort analyses: Applies inferred models to large cross-sectional cohorts such as TCGA to associate evolutionary paths with clinical outcomes.

Methodology:

Infers probabilistic graphical models from cross-sectional sequencing mutational profiles by integrating arbitrary logical formulas that connect mutational events; evaluations include simulation studies and application to TCGA cohort analyses.