ToMExO

ToMExO models cancer progression using a probabilistic tree-structured framework to elucidate relationships among cancer driver genes and their mutation patterns.

Key Features:

• Tree-Structured Model: ToMExO employs a hierarchical tree in which each node represents an individual cancer driver gene or a set of genes, capturing mutual exclusivity and temporal ordering of mutational events.
• Dynamic Programming for Likelihood Calculation: It uses a dynamic programming approach to efficiently compute likelihoods for noisy cross-sectional mutation datasets.
• Markov Chain Monte Carlo (MCMC) Inference Algorithm: It performs inference via an MCMC algorithm with engineered moves that accelerate likelihood calculations, enabling analysis of datasets with hundreds of genes and thousands of tumors.

Scientific Applications:

• Mutation Accumulation Analysis: Modeling how driver gene mutations accumulate to identify subtype-specific progression processes.
• Pattern Identification and Validation: Identifying significant patterns and relationships among genes involved in cancer development and validating them using method-independent metrics to assess causality and significance.
• Cohort Analyses of Specific Cancers: Applied to large-scale datasets from colorectal cancer, glioblastoma, and pancreatic cancer to uncover causal relationships and gene interactions.

Methodology:

Analyzes cross-sectional tumor data using a probabilistic tree-structured model, computes likelihoods via dynamic programming, performs inference with a Markov Chain Monte Carlo algorithm that includes engineered moves for rapid likelihood evaluation, and has been applied to synthetic datasets and benchmarked against state-of-the-art methods on moderate-sized biological datasets.