Model-based tumor subclonal reconstruction

Abstract

The vast majority of cancer next-generation sequencing data consist of bulk samples composed of mixtures of cancer and normal cells. To study tumor evolution, subclonal reconstruction approaches based on machine learning are used to separate subpopulation of cancer cells and reconstruct their ancestral relationships. However, current approaches are entirely data-driven and agnostic to evolutionary theory. We demonstrate that systematic errors occur in subclonal reconstruction if tumor evolution is not accounted for, and that those errors increase when multiple samples are taken from the same tumor. To address this issue, we present a novel approach for model-based subclonal reconstruction that combines data-driven machine learning with evolutionary theory. Using public, synthetic and newly generated data, we show the method is more robust and accurate than current techniques in both single-sample and multi-region sequencing data. With careful data curation and interpretation, we show how the method allows minimizing the confounding factors that affect non-evolutionary methods, leading to a more accurate recovery of the evolutionary history of human tumors.

Testing for phylogenetic signal in single-cell RNA-seq data

“…Within-organism cancer evolution is increasingly being studied using population genetics approaches, including phylogenetics (Navin et al 2011; Yuan et al 2015; Alves et al 2017; Schwartz et al 2017; Caravagna et al 2018; Singer et al 2018; Alves et al 2019; Caravagna et al 2019; Detering et al 2019; Malikic et al 2019; Werner et al 2019; Kuipers et al 2020), to understand molecular dynamics of cancer cell populatons. These approaches have shown promise to be developed into therapeutic applications in the personalized medicine framework (Gerlinger et al 2012; Abbosh et al 2017; Rao et al 2020).…”

Mapping the breast cancer metastatic cascade onto ctDNA using genetic and epigenetic clonal tracking

“…These results are consistent with a more recent analysis of primary-metastatic pairs in colorectal cancer 20 . In this regard, careful identification of genetically distinct subpopulations of cancer cells is crucial, and liberal subclonal classifications can instead produce overcomplicated patterns 21 where the inferred evolutionary history of the tumour is driven by measurement noise. Moreover, the identification of the correct metastatic cascade is limited often by single samples taken from each metastatic site, instead of multiregion sampling from each metastatic deposit.…”

Assessing reliability of intra-tumor heterogeneity estimates from single sample whole exome sequencing data

“…However, each sample is a priori heterogeneous itself, requiring a first multi-sample deconvolution. This first step can be challenging, as it is thought that multi-sample reconstruction is subject to a larger statistical bias compared to single sample reconstruction [69], and the accuracy of this first step will be critical in the final results. A final possibility is to rely on simulated data, which have the major drawback to not be necessarily representative of the true biological data, as recently highlighted for ITH in [69], that point to an aspect of the input data so far overlooked by the community.…”