Evolutionary Multitasking Optimization Based Multi-Objective Network Control Principles for Identifying Personalized Drug Targets with Cancer

Multi-objective optimization-based structural network control principles (MONCP) achieved good performance in identifying personalized drug targets.However, the current approaches neglect the fact that for the same personalized gene interaction networks(PGIN), different MONCPs have some similar properties. Therefore, this paper proposes a evolutionary multitasking optimization framework (called DATCMO) which treats different control principles as multiple optimization tasks and allows the exchange of information between tasks on one PGIN for an individual patient. The technical advantage of this method is to establish a respective auxiliary population for each task, and perform knowledge transfer during the evolutionary process for different MONCPs. The experimental results on 112 breast cancer patients data from The Cancer Genome Atlas indicate that, compared to current advance methods, the proposed algorithm proves to be more competitive.