[Paper Review] Learning mutational graphs of individual tumor evolution from multi-sample sequencing data

Daniele Ramazzotti, Alex Graudenzi+3 more|arXiv (Cornell University)|Sep 4, 2017

Cancer Genomics and Diagnostics4 citations

TL;DR

TRaIT is a unified statistical framework that infers the order of genomic alterations in tumor evolution from both multi-region and single-cell sequencing data, explicitly modeling violations of the infinite sites assumption such as convergent evolution. It outperforms state-of-the-art phylogenetic methods in accuracy, robustness to noise, and performance on small samples, enabling more accurate quantification of intra-tumor heterogeneity and generation of testable hypotheses.

ABSTRACT

Phylogenetic techniques quantify intra-tumor heterogeneity by deconvolving either clonal or mutational trees from multi-sample sequencing data of individual tumors. Most of these methods rely on the well-known infinite sites assumption, and are limited to process either multi-region or single-cell sequencing data. Here, we improve over those methods with TRaIT, a unified statistical framework for the inference of the accumula- tion order of multiple types of genomic alterations driving tumor development. TRaIT supports both multi-region and single-cell sequencing data, and output mutational graphs accounting for violations of the infinite sites assumption due to convergent evolution, and other complex phenomena that cannot be detected with phylogenetic tools. Our method displays better accuracy, performance and robustness to noise and small sample size than state-of-the-art phylogenetic methods. We show with single-cell data from breast cancer and multi-region data from colorectal cancer that TRaIT can quantify the extent of intra-tumor heterogeneity and generate new testable experimental hypotheses.

Motivation & Objective

• To address the limitations of existing phylogenetic methods that rely on the infinite sites assumption and cannot handle complex evolutionary phenomena like convergent evolution.
• To develop a unified framework capable of analyzing both multi-region and single-cell sequencing data for tumor evolution inference.
• To improve accuracy, robustness to noise, and performance on small sample sizes compared to state-of-the-art phylogenetic methods.
• To generate mutational graphs that account for violations of the infinite sites assumption and other complex evolutionary dynamics.
• To enable the quantification of intra-tumor heterogeneity and the generation of testable experimental hypotheses from real sequencing data.

Proposed method

• TRaIT employs a unified statistical framework to infer the accumulation order of multiple types of genomic alterations in tumor evolution.
• It models the evolutionary process using probabilistic graphical models that account for convergent evolution and other phenomena violating the infinite sites assumption.
• The method integrates both multi-region and single-cell sequencing data within a single inference pipeline, enabling consistent analysis across data types.
• TRaIT uses a likelihood-based inference approach to estimate the most probable mutational graph structure from observed mutation profiles.
• It incorporates uncertainty quantification and robustness to noise through statistical modeling of sequencing errors and sampling variability.
• The framework supports the inference of complex mutational graphs that go beyond simple clonal or mutational trees, capturing non-linear evolutionary trajectories.

Experimental results

Research questions

• RQ1 Can a unified framework accurately infer tumor evolution from both multi-region and single-cell sequencing data while accounting for evolutionary complexities?
• RQ2 How does TRaIT perform in comparison to state-of-the-art phylogenetic methods in terms of accuracy, robustness to noise, and small sample sizes?
• RQ3 To what extent can TRaIT detect and model violations of the infinite sites assumption, such as convergent evolution, in real tumor sequencing data?
• RQ4 Can TRaIT generate biologically meaningful mutational graphs that reveal new insights into intra-tumor heterogeneity and suggest testable hypotheses?
• RQ5 How well does TRaIT reconstruct tumor evolutionary histories in real-world datasets, such as breast cancer single-cell and colorectal cancer multi-region sequencing data?

Key findings

• TRaIT demonstrates superior accuracy in inferring mutational graphs compared to state-of-the-art phylogenetic methods, particularly in scenarios involving complex evolutionary dynamics.
• The method shows enhanced robustness to noise and improved performance on small sample sizes, making it suitable for challenging clinical sequencing datasets.
• TRaIT successfully models violations of the infinite sites assumption, such as convergent evolution, which are undetectable by standard phylogenetic tools.
• In single-cell breast cancer data, TRaIT quantified intra-tumor heterogeneity more effectively than existing methods, revealing complex subclonal architectures.
• In multi-region colorectal cancer data, TRaIT generated mutational graphs that provided new, testable experimental hypotheses about tumor evolution.
• The framework produces biologically plausible mutational graphs that capture non-linear evolutionary trajectories beyond the scope of traditional tree-based models.