Evolutionary unpredictability in cancer model systems

Abstract

Abstract Despite the advent of advanced molecular prognostic tools, it is still difficult to predict the course of disease for cancer patients at the individual level. This lack of predictability is also reflected in many experimental cancer model systems, begging the question of whether certain biological aspects of cancer (eg. growth, evolution etc.) can ever be anticipated or if there remains an inherent unpredictability to cancer, similar to other complex biological systems. We demonstrate by a combination of agent-based mathematical modelling, analysis of patient-derived xenograft model systems from multiple cancer types, and in-vitro culture that certain conditions increase stochasticity of the clonal landscape of cancer growth. Our findings indicate that under those conditions, the cancer genome may behave as a complex dynamic system, making its long-term evolution inherently unpredictable.