A while ago we posted this preprint on bioRxiv describing an agent-based model of multiple myeloma. I think it is a nice paper and you should go and read it. In it we describe an agent-based model and show results derived from running many simulations. The following movie shows two typical simulations, one without myeloma cells that shows a variety of normal stromal cells in the bone ecosystem. There is quite a lot going on and this is, again, with no cancer. On the right in the same simulation this time with myeloma cancer cells:

What you see on the right is what bone biologists working on cancer call the vicious cycle. Also, about 10 years ago, we published a paper that was also an agent-based model, also of the bone ecosystem, but in this case studying bone metastatic prostate cancer. Here is a sample simulation with no prostate cancer cells:

What they do have in common is that we calibrated both models so that the bone can remain in dynamic homeostasis for large periods of time. This process is not easy but it's somewhat helped by the fact that normal cells behave in a more consistent manner than cancer cells (which not only varies significantly from patient to patient but also within one patient due to intra tumor heterogeneity). One drawback is that, unfortunately even though both models were developed from scratch by two different people using two different sets of data, two different development frameworks, for two different research questions, the outputs resemble each other closely enough that, superficially, they might look like the same model. In fact, the latter one was even more carefully calibrated and can maintain bone homeostasis for several bone remodeling cycles even if we allow various bone modeling units operating simultaneously in the same section of the bone. The superficial similarities have made some reviewers think that we were effectively using the same model again, diminishing the novelty of the research.

So why would we spend time on this anyway? Well, cancers do not grow in a vacuum and, whether you are interested in evolutionary dynamics of cancer or not, we often cannot assume that tumor cells interact (compete/cooperate/etc) only with other cancer cells. In fact, cancer cells can be seen as an invasive species in an existing ecosystem and whether the tumor will be successful in establishing itself and what tumor cell phenotypes will emerge victorious will depend on whether the tumor contains cells that can take advantage of this ecosystem and co-opt normal cells and their interactions for their own benefit. In fact, recent work by Chris Tape at UCL add further evidence of the importance of the interactions between normal and tumor cells:

So if your model is about the tumor microenvironment or cancer evolution I think it makes sense to ensure that it can show normal cells doing normal things and interacting with each other before we add tumor cells to the mix.