Longitudinal cancer evolution from single cells

Abstract

The rise of longitudinal single-cell sequencing experiments on patient-derived cell cultures, xenografts and organoids is opening new opportunities to track cancer evolution in single tumors and to investigate intra-tumor heterogeneity. This is particularly relevant when assessing the efficacy of therapies over time on the clonal composition of a tumor and in the identification of resistant subclones. We here introduce LACE (Longitudinal Analysis of Cancer Evolution), the first algorithmic framework that processes single-cell somatic mutation profiles from cancer samples collected at different time points and in distinct experimental settings, to produce longitudinal models of cancer evolution. Our approach solves a Boolean matrix factorization problem with phylogenetic constraints, by maximizing a weighted likelihood function computed on multiple time points, and we show with simulations that it outperforms state-of-the-art methods for both bulk and single-cell sequencing data. Remarkably, as the results are robust with respect to high levels of data-specific errors, LACE can be employed to process single-cell mutational profiles as generated by calling variants from the increasingly available scRNA-seq data, thus obviating the need of relying on rarer and more expensive genome sequencing experiments. This also allows to investigate the relation between genomic clonal evolution and phenotype at the single-cell level. To illustrate the capabilities of LACE, we show its application to a longitudinal scRNA-seq dataset of patient-derived xenografts of BRAF V600E/K mutant melanomas, in which we characterize the impact of concurrent BRAF/MEK-inhibition on clonal evolution, also by showing that distinct genetic clones reveal different sensitivity to the therapy.

A neural crest stem cell-like state drives nongenetic resistance to targeted therapy in melanoma

“…Consistently, the findings described herein identify the NCSCs as a melanoma subpopulation that is highly permissive to nongenetic reprogramming and a key driver of nongenetic resistance. Interestingly, we also provide evidence that nongenetic resistance can develop in absence of selection of a single, genetically-distinct, clonal population, a conclusion that is further supported by the recent re-analysis of our single-cell data with LACE (Longitudinal Analysis of Cancer Evolution), an algorithm that processes single-cell somatic mutation profiles from scRNAseq data 52 . Together, these analyses raise the possibility that nongenetic resistance may not be clonal but develop, instead, through collective reprogramming.…”

VERSO: A comprehensive framework for the inference of robust phylogenies and the quantification of intra-host genomic diversity of viral samples

“…The first step of VERSO employs a probabilistic maximum-likelihood framework for the reconstruction of robust phylogenetic trees from binarized mutational profiles of clonal variants (or, alternatively, from consensus sequences). This step relies on an evolved version of the algorithmic framework introduced by Ramazzotti et al 105 for the inference of cancer evolution models from single-cell sequencing data, and can be executed independently from step #2, in case raw sequencing data are not available.…”

VERSO: a comprehensive framework for the inference of robust phylogenies and the quantification of intra-host genomic diversity of viral samples

“…The first step of VERSO employs a probabilistic maximum-likelihood framework for the reconstruction of robust phylogenetic trees from binarized mutational profiles of clonal variants (or, alternatively, from consensus sequences). This step relies on an evolved version of the algorithmic framework introduced in 105 for the inference of cancer evolution models from single-cell sequencing data, and can be executed independently from STEP #2, in case raw sequencing data are not available.…”