Intelligent Search for Personalized Cancer Therapy Synthesis: An Experimental Comparison

CRC is one of the deadliest forms of tumor for adult humans, which often requires expensive and highly toxic treatments. In this work, we exploit a recent System Biology Markup Language (SBML) quantitative model of the tumor growth and its immune response to two immunotherapic drugs and define a non-linear, constrained optimization problem for the synthesis of personalized therapies. We compute a virtual population of physiologically admissible patients and define a parametric therapy model that enables the usage of standard parameter optimization algorithms. The whole approach is implemented with a novel methodology that requires a single tool, namely COPASI, for the therapy template modelling, the definition of the optimization problem and its solution. We present results of an experimental evaluation of our approach by comparing five search algorithms among those implemented in COPASI, in order to find out the most promising kind of algorithms for this problem. The results show that population-based search algorithms perform well and are able to compute personalized therapies within time bounds compatible with clinical practice.