An Iterated Local Search Algorithm for the Clonal Deconvolution Problem

Cancer is a disease characterized by the continuous acquisition of random mutations by cells, which are subsequently subjected to selection forces that favour the survival of some cells over others. The result of this evolutionary process called clonal evolution is a genetically heterogeneous mass known as a tumor, and identifying its composition is crucial not only for gaining further understanding of the disease, but also for designing effective therapies tailored to the particularities of the whole tumor. Thus, the clonal deconvolution problem tries to identify the different cell subpopulations that form the tumor and the phylogenetic tree that describes the evolutionary process that led to it from a series of biopsies that are an admixture of those subpopulations. This problem has been tackled from different perspectives, but as far as we know, metaheuristics have not been explored. In this article, we propose an Iterated Local Search (ILS) approach as a first metaheuristic approximation to solve this problem. Preliminary results on simulated data show that our method outperforms two well-established heuristic algorithms when running time is constrained. Moreover, the algorithm has the advantage that it is a flexible approach in which assumptions on the tumor development mode are not directly implemented, and it can therefore be easily adapted to accommodate new discoveries made on the evolution mechanisms in cancer.