Pairtree

Pairtree reconstructs clone trees from bulk DNA sequencing data to infer the evolutionary relationships among somatic mutation–defined subpopulations in cancer.

Key Features:

• Clone Tree Construction: Pairtree constructs clone trees by analyzing DNA sequencing data from one or more bulk samples, characterizing somatic mutations specific to genetically distinct subpopulations to elucidate their evolutionary ancestry.
• Bayesian Inference and MCMC: Pairtree computes posterior distributions over pairwise evolutionary relationships using Bayesian inference and uses those pairwise posteriors within a Markov Chain Monte Carlo (MCMC) algorithm to infer posterior distributions over clone trees.
• Detection of Infinite Sites Assumption Violations: Pairtree leverages pairwise mutation relationships to detect mutations that violate the infinite sites assumption.
• Scalability and Performance: Pairtree can analyze up to 100 bulk samples per cancer and identify 30 or more cell subpopulations; its performance improves with additional bulk samples and, on simulated data, it outperformed other methods and reproduced or improved expert-derived reconstructions in B-progenitor acute lymphoblastic leukemias with up to 90 samples.

Scientific Applications:

• Reconstructing tumor evolutionary history: Pairtree provides insights into the natural history of cancer by revealing the evolutionary relationships among subclonal populations.
• Understanding disease development: Pairtree aids in identifying critical points in disease progression by resolving subpopulation ancestries.
• Informing treatment strategies and biological interplay: Pairtree informs treatment strategy considerations and enables exploration of interactions between cancer, host biology, and therapeutic interventions.

Methodology:

Pairtree analyzes bulk DNA sequencing data to characterize somatic mutations in genetically distinct subpopulations, computes posterior distributions over pairwise evolutionary relationships via Bayesian inference, uses those pairwise posteriors in a Markov Chain Monte Carlo (MCMC) algorithm to infer posterior distributions over clone trees, and leverages pairwise mutation relationships to identify violations of the infinite sites assumption.