A phylogenetic approach to inferring the order in which mutations arise during cancer progression

Although the role of evolutionary process in cancer progression is widely accepted, increasing attention is being given to the evolutionary mechanisms that can lead to differences in clinical outcome. Recent studies suggest that the temporal order in which somatic mutations accumulate during cancer progression is important. Single-cell sequencing (SCS) provides a unique opportunity to examine the effect that the mutation order has on cancer progression and treatment effect. However, the error rates associated with single-cell sequencing are known to be high, which greatly complicates the task. We propose a novel method for inferring the order in which somatic mutations arise within an individual tumor using noisy data from single-cell sequencing. Our method incorporates models at two levels in that the evolutionary process of somatic mutation within the tumor is modeled along with the technical errors that arise from the single-cell sequencing data collection process. Through analyses of simulations across a wide range of realistic scenarios, we show that our method substantially outperforms existing approaches for identifying mutation order. Most importantly, our method provides a unique means to capture and quantify the uncertainty in the inferred mutation order along a given phylogeny. We illustrate our method by analyzing data from colorectal and prostate cancer patients, in which our method strengthens previously reported mutation orders. Our work is an important step towards producing meaningful prediction of mutation order with high accuracy and measuring the uncertainty of predicted mutation order in cancer patients, with the potential to lead to new insights about the evolutionary trajectories of cancer.Author summary: Cancer evolves as a consequence of the accumulation of somatic mutations, and diverse clones are formed during this process, resulting in intratumoral heterogeneity (ITH). Similar cancer subtypes often display different landscapes of genetic alterations, and tumors that harbor the same mutations sometimes respond differently to therapy. Recent studies have provided evidence that mutation order is critical in cancer. With application to noisy single-cell sequencing data, we develop a computational framework to reconstruct the mutation order along a phylogeny with high accuracy and to measure the uncertainty in a unique way. We demonstrate that our method exhibits robust performance, both for predicting mutation order and for assessing uncertainty of predicted mutation order. We include applications to prostate cancer and colorectal cancer patients, where we identify mutation orders that have been reported to be important in cancer progression. Knowledge of the tumor evolutionary history, especially the mutation order, would greatly improve our understanding of a tumor’s ITH and aid in treatment decisions. Our work is a powerful computational tool that can be applied to address research questions in the field of cancer and has important translational applications for improving cancer diagnosis and personalized therapy.