Joint inference of exclusivity patterns and recurrent trajectories from tumor mutation trees

Abstract

Cancer progression is an evolutionary process shaped by both deterministic and stochastic forces. Multi-region and single-cell sequencing of tumors enable high-resolution reconstruction of the mutational history of each tumor and highlight the extensive diversity across tumors and patients. Resolving the interactions among mutations and recovering recurrent evolutionary processes may offer greater opportunities for successful therapeutic strategies. To this end, we present a novel probabilistic framework, called TreeMHN, for the joint inference of exclusivity patterns and recurrent trajectories from a cohort of intra-tumor phylogenetic trees. Through simulations, we show that TreeMHN outperforms existing alternatives that can only focus on one aspect of the task. By analyzing datasets of blood, lung, and breast cancers, we find the most likely evolutionary trajectories and mutational patterns, consistent with and enriching our current understanding of tumorigenesis. Moreover, TreeMHN facilitates the prediction of tumor evolution and provides probabilistic measures on the next mutational events given a tumor tree, a prerequisite for evolution-guided treatment strategies.

EvAM-Tools: tools for evolutionary accumulation and cancer progression models

“…In Table 1, I have not listed TreeMHN (Luo et al, 2023), Hintra (Khakabimamaghani et al, 2019), or REVOLVER (Caravagna et al, 2018), as they focus on inferring recurrent evolutionary trajectories from multiple within-patient (or intra-tumor) phylogenetic trees; these data often arise from multi-region, single-cell sequencing (i.e., the data come from multiple patients, where each patient contributes multiple samples, from different tissues or sampling times). ToMExO is focused on a somewhat different task (Neyshabouri and Lagergren, 2022, p. 3): "(...) simultaneously identify critical driver genes, group them as sets of mutually exclusive genes (driver pathways), and arrange them in a tree structure representing the order in which they get mutated" (see also discussion in section 4.5).…”

“…e The current code models the hypercubic transitions using pairwise dependencies, but could be extended to use higher-order dependencies f This limit applies to the MHN implementation. TreeMHN (Luo et al, 2023) can be used with a larger numbers of features, its runtime and memory usage being limited by sample size.…”

“…Data with more complex dependency structures can also be analyzed by some of these methods, or by closely related methods. In particular, data with phylogenetic/genealogical relationships (where some samples share common ancestors that are not shared by other samples) are analyzed by two methods discussed here (HyperTraPS, section 2.3.2, and HyperHMM section 2.3.3) as well as by TreeMHN (Luo et al, 2023) and REVOLVER (Caravagna et al, 2018) (see section 2.4).…”

“…11. Bulk sequencing can create mutational profiles that do not correspond to any existing genotype (Alves et al, 2017;Luo et al, 2023). Under within-cell XOR restrictions we can observe mutational profiles ("tumor genotypes") that violate the XOR and, thus, that are not possible under the DAG; this would lead to inferring incorrect restrictions.…”

“…Most of the cancer progression models assume that each subject or sample provides a single data point and that different subjects are independent: these are cross-sectional data, and this is the standard use-case for these models I will focus on here. But, as I will mention later, some methods for cross-sectional data can also handle phylogenetic (and longitudinal dependencies) in the data (Johnston and Røyrvik, 2020;Greenbury et al, 2020;Peach et al, 2021;Moen and Johnston, 2023); as well, some cancer progression models have been extended to analyze data sets with multiple observations per subject, where each patient contributes multiple samples, from different tissues or sampling times (Luo et al, 2023;Caravagna et al, 2018).…”