Application talk
Abstract
Cancer is an evolutionary process—a fact not yet fully accounted for in current treatment decisions. Severe consequences can result from regarding tumours as static, homogeneous entities, as the evolution of cancer can bring resistance to treatment, and ultimately, metastasis and death. Resolving the evolutionary trajectory of a patient’s cancer is now a reality in the academic setting, and bringing this advancement to the clinic is the next step toward improved patient treatment. A central challenge here is the clinical interpretability of cancer evolution data. To this end, we have developed two visualization tools: TimeSweep for temporal and CloneMapper for spatial cancer evolution analysis. While these tools were developed in an academic setting, we believe they have potential for future clinical applications. Advancements in next-generation sequencing and statistical methods have enabled the quantification of a tumour’s evolutionary trajectory. To spatially or temporally quantify this evolution, a tumour is broken down into its constituent subpopulations, or clones, and the prevalence of each clone is tracked over time or space. The precise mutational content, hierarchy and prevalence of clones is determined using sequencing data as input to sophisticated computational and statistical methods. The output of such methods consists of a complex set of interrelated tables in need of integrative and intuitive visualization. We have developed a visualization tool for temporal evolution (Fig. 1a), called TimeSweep, that automatically generates an intuitive interactive visualization of a tumour’s evolutionary trajectory over time. The view integrates time-series observations of a tumour’s mutational content and clonal composition, all in the context of its clonal phylogeny. The plot axes show time horizontally and clonal prevalence vertically, while within the plot, each clone is nested according to its place in the clonal phylogeny. From this view, one can easily track the success of each lineage over time, allowing for a better understanding of the tumour’s clonal dynamics. For a more granular view of individual clonal trajectories, the hierarchical organization may be turned off, displaying each clone changing in prevalence on its own track. The most fundamental use case of TimeSweep is to expose the evolutionary dynamics occurring within a patient’s tumour as a result of selection and drift. At the time of diagnosis, the most abundant clone may subsequently shrink due to intense evolutionary competition from one or more minor clones. In this case there would be little value in treating the most abundant clone alone—the expanding minor clone should be targeted as well. Moreover, applying a selective pressure such as treatment will strongly alter the baseline evolutionary dynamics of a given tumour. Predictive models of cancer evolution in response to treatment are currently under development at cancer centres worldwide, where a patient’s tumour is grown in mice and treated with various drugs. Using TimeSweep, a clinician may track a cancer’s evolutionary response to each drug over time. This is especially important when no drug is completely effective in a mouse model. In this case, the clinician can visualize the evolutionary underpinnings of resistance and devise an effective strategy for combination therapy. For instance, if the TimeSweep displays that application of Drug A was effective against most clones, but led to an expansion of a clone harbouring Mutation X, the clinician may first treat the patient with Drug A, then follow up with a drug that targets Mutation X. A more large-scale application for TimeSweep is in the context of clinical trials. In these trials, treatment effectiveness may be evaluated based on outcomes including tumour diameter decrease and time to progression. The genetic mechanisms behind these outcomes are often ignored, despite having great potential for predicting the responsiveness of a tumour to the drug on trial. Visualizing the evolutionary trajectories of patient cancers during clinical trials is the first step toward stratifying future patients as potential responders or non-responders. For instance, if many unresponsive cancers contained a subpopulation of cells with Mutation Y which expanded post-treatment, perhaps future patients with Mutation Y could be spared the application of this drug and given a more suitable treatment. Thus far we have considered within-tumour evolution as the only source for treatment evasion. However, in the case of metastasis, each anatomic site will experience an evolutionary trajectory that begins with the seeded cell, resulting in high between-tumour evolution. For patients with metastatic cancer, between-tumour evolution can be a significant contributor to treatment ineffectiveness. Using our visualization tool for spatial cancer evolution, CloneMapper, it is possible to simultaneously and intuitively explore both within- and between-tumour evolution for a given patient. CloneMapper (Fig. 1b), built for the analysis of spatial data, displays the clonal composition of tumours at a variety of anatomic locations in the patient. The representation is radially divided to display each tumour sample separately, yet all samples are centrally connected to their anatomic sources. Two views of each sample form a comprehensive understanding of its composition. The first view, a cellular aggregate, visually displays the prevalence of each clone. The second view shows the patient’s clonal phylogeny, highlighting the clones composing the tumour sample. Additionally, the bottom of the view houses a mutation table, providing information about the genetic changes associated with dominant and emerging clones. The primary application for CloneMapper in a clinical context is for treating metastatic cancers. Recent studies have shown striking between-tumour heterogeneity in patients with metastatic cancer, suggesting that a single treatment may not be universally effective in a patient, and highlighting the importance of treating tumours individually. Employing CloneMapper, the clinician may visualize the distribution of mutations across anatomic sites, enabling predictions of therapeutic effectiveness across tumours. Furthermore, mutations unique to potentially unresponsive sites may be leveraged as targets for a secondary therapy. In the future, information about each patient’s cancer evolution, when available in the clinic, could make a significant difference in the treatment decision-making process. The applications above demonstrate how the TimeSweep and CloneMapper visualization tools can present complex cancer evolution data to the clinician in a comprehensive and intuitive manner that can ultimately improve therapeutic success rates.
Authors
Maia Smith, BC Cancer Agency, Canada
Cydney Nielsen, BC Cancer Agency, Canada
Fong Chun Chan, BC Cancer Agency, Canada
Andrew Roth, BC Cancer Agency, Canada
Andrew McPherson, BC Cancer Agency, Canada
Daniel Machev, BC Cancer Agency, Canada
Sohrab Shah, BC Cancer Agency, Canada