The ecology of cancer from an evolutionary game theory perspective.

The accumulation of somatic mutations, to which the cellular genome is permanently exposed, often leads to cancer. Analysis of any tumour shows that, besides the malignant cells, one finds other 'supporting' cells such as fibroblasts, immune cells of various types and even blood vessels. Together, these cells generate the microenvironment that enables the malignant cell population to grow and ultimately lead to disease. Therefore, understanding the dynamics of tumour growth and response to therapy is incomplete unless the interactions between the malignant cells and normal cells are investigated in the environment in which they take place. The complex interactions between cells in such an ecosystem result from the exchange of information in the form of cytokines- and adhesion-dependent interactions. Such processes impose costs and benefits to the participating cells that may be conveniently recast in the form of a game pay-off matrix. As a result, tumour progression and dynamics can be described in terms of evolutionary game theory (EGT), which provides a convenient framework in which to capture the frequency-dependent nature of ecosystem dynamics. Here, we provide a tutorial review of the central aspects of EGT, establishing a relation with the problem of cancer. Along the way, we also digress on fitness and of ways to compute it. Subsequently, we show how EGT can be applied to the study of the various manifestations and dynamics of multiple myeloma bone disease and its preceding condition known as monoclonal gammopathy of undetermined significance. We translate the complex biochemical signals into costs and benefits of different cell types, thus defining a game pay-off matrix. Then we use the well-known properties of the EGT equations to reduce the number of core parameters that characterize disease evolution. Finally, we provide an interpretation of these core parameters in terms of what their function is in the ecosystem we are describing and generate predictions on the type and timing of interventions that can alter the natural history of these two conditions.