Examining systematically how spatial structure influences tumour evolution

Using computer model simulations that were compared to actual clinical data for various cancer types, researchers from the group of Niko Beerenwinkel and colleagues gained new insights into why human cancers develop in such diverse ways. Characterising the way, manner or pattern of evolution in tumours is important for clinical forecasting and optimising cancer treatment, the authors say in their new study in Nature Ecology & Evolution.

27.12.2021 by Robert Noble

Number of comments

A major challenge in cancer research is to infer properties of a tumour based on limited genetic information. To understand this problem, consider a sports analogy. Suppose you’re told only that in a head-to-head game, Team A scored twice as often as Team B. Can you figure out how much better Team A is than Team B, so you can predict the outcomes of future contests?

One way to answer this question is to use a computer model, in which each team is assigned a probability of scoring on each attempt. After trying many different settings, you can conclude that the most likely scoring probabilities are those for which the simulation outcomes resemble the actual game result. Although you can never be sure what the true probabilities are, you can at least find their most likely ranges.

A major goal of modern cancer research is to characterise the evolutionary process within tumours.Robert Noble, lead author

But knowing the ratio of the final scores is not enough. In high-scoring basketball, it’s unlikely that one team will score twice as many points as their opponents unless they are vastly superior. In soccer, by contrast, it’s not unusual for the better team to lose 2-1 by a stroke of bad luck. To make accurate inferences, you need to know the rules of the game.

Much as sports teams compete to score points, so groups of closely related cells – known as clones – compete within tumours for the space and resources they need to survive and multiply. Oncologists use genetic sequencing to determine the relative sizes of these clones when a patient comes to the clinic. If one clone is larger than another then it might be because its cells have so-called “driver” mutations that cause them to proliferate faster.

But the effect of mutations on tumour development depends on how cells interact with one another, which is governed by the tumour’s spatial structure. Much as coronavirus spreads more slowly when people stay home and avoid mixing, so driver mutations spread more slowly within tumours if cells are confined to small patches, with only rare movement between patches. The rules matter in this game, too.

The new study is the first to systematically examine how spatial structure influences tumour evolution. To do this, the D-BSSE researchers in Niko Beerenwinkel’s group developed a computational model with the flexibility to simulate alternative spatial structures and types of cell dispersal. They then ran thousands of simulations with different structures and parameter values and compared the results to recent, state-of-the-art DNA sequencing data from actual human tumours.

The team found that the diverse spatial structures of human tumours can cause them to evolve in vastly different ways. The computer model predictions are consistent with clinical data for cancer types with matching structures.

“Our findings have important implications for interpreting cancer genetic data,” says Robert Noble, lead author of the team’s paper published in Nature Ecology & Evolution. “A major goal of modern cancer research is to characterise the evolutionary process within tumours. We have shown that to get an accurate picture of what’s going on, you need to account for each tumour’s particular spatial structure.”

Robert, who recently left the Beerenwinkel group to start his own group at City, University of London, concludes that “By mechanistically connecting tumour architecture to the mode of tumour evolution, our work provides the blueprint for a new generation of patient-specific models for forecasting tumour progression and for optimising therapy.”