Abstract
The investigation of commonalities among different cancers is one of the important problems for deciphering cancers and will be helpful for personalized therapy in cancer treatment. Zhang et al. have presented the ComMDP method to solve this problem in 2017. However, when the number of samples among different cancers varies largely, accumulating the absolute weight value of every cancer, performed by the ComMDP method, may lead to missing some driver pathways. In this paper, an improved mathematical model is proposed by replacing the absolute weight values with the relative ratios of them, and introducing variance to minimize the dispersion of each ratio. By introducing a kind of short chromosome code and a greedy based recombination operator, a pathenogenetic algorithm PGA-MDP is put forward for solving this model. Experimental results indicate that the PGA-MDP algorithm is indeed able to detect some biologically meaningful gene sets which are missed by the ComMDP one. Hence it may become a useful complementary tool for identifying cancer pathways.