Spatially constrained tumour growth affects the patterns of clonal selection and neutral drift in cancer genomic data

Abstract

Quantification of the effect of spatial tumour sampling on the patterns of mutations detected in next-generation sequencing data is largely lacking. Here we use a spatial stochastic cellular automaton model of tumour growth that accounts for somatic mutations, selection, drift and spatial constrains, to simulate multi-region sequencing data derived from spatial sampling of a neoplasm. We show that the spatial structure of a solid cancer has a major impact on the detection of clonal selection and genetic drift from bulk sequencing data and single-cell sequencing data. Our results indicate that spatial constrains can introduce significant sampling biases when performing multi-region bulk sampling and that such bias becomes a major confounding factor for the measurement of the evolutionary dynamics of human tumours. We present a statistical inference framework that takes into account the spatial effects of a growing tumour and allows inferring the evolutionary dynamics from patient genomic data. Our analysis shows that measuring cancer evolution using next-generation sequencing while accounting for the numerous confounding factors requires a mechanistic model-based approach that captures the sources of noise in the data. SummarySequencing the DNA of cancer cells from human tumours has become one of the main tools to study cancer biology. However, sequencing data are complex and often difficult to interpret. In particular, the way in which the tissue is sampled and the data are collected, impact the interpretation of the results significantly. We argue that understanding cancer genomic data requires mathematical models and computer simulations that tell us what we expect the data to look like, with the aim of understanding the impact of confounding factors and biases in the data generation step. In this study, we develop a spatial simulation of tumour growth that also simulates the data generation process, and demonstrate that biases in the sampling step and current technological limitations severely impact the interpretation of the results. We then provide a statistical framework that can be used to overcome these biases and more robustly measure aspects of the biology of tumours from the data.

Subclonal reconstruction of tumors by using machine learning and population genetics

“…We used simulated data from a stochastic spatial branching process model of tumor growth 15 to assess the confounding factors of spatial sampling discussed above, and to provide a rationale to handle them and interpret the data correctly.…”

“…A second simulator model described in ref 15 was used to generate several synthetic multi-region sequencing datasets to test the behavior of MOBSTER in a multivariate setting and to assess confounders (Figures 4 and 5).…”

“…We also expand MOBSTER to analyze multivariate data from multiple samples of the same tumor. We show that unavoidable sampling bias and lineage admixture caused by the spatial structure of the tumor 15 produces additional confounding factors that need to be considered when interpreting the output of the subclonal reconstruction. Rational curation based on understanding how a tumor expands, how its growth is affected by stochastic forces such as drift 16 , and a careful spatial sampling strategy, can be combined with MOBSTER to accurately reconstruct the tumor phylogenetic history.…”

The fidelity of cancer cells in PDX models: Characteristics, mechanism and clinical significance

“…Due to ITH and limited tissue quantities, resected tumors, biopsy specimens and CTCs can hardly represent the whole landscape of parental tumors . Different specimens of the same parental tumor develop over different trajectories in multiple xenograft tumors due to variable clonal composition (Fig.…”

Model-based tumor subclonal reconstruction

“…= {40,60,80,100,200}) and otherwise identical parameters were created. A second simulator model described in ref 15 was used to generate several synthetic multi-region sequencing datasets to test the behavior of MOBSTER in a multivariate setting and to assess confounders (Figures 4 and 5).…”