Multi-objective Reconstruction Phylogenetic Tree Algorithm Based On Membrane Structure And Consensus

Evolution is a series of irreversible genetic system changes produced by living beings under the influence of living environment and self.Phylogenetic analysis is used to infer or evaluate the evolutionary history of organisms,usually using evolutionary trees to describe evolutionary relationships.However,the current multi-objective evolutionary algorithms used to reconstruct the evolutionary tree are often subject to unbearable time consumption because of the complex calculation of the objective function and the slow convergence of the evolutionary algorithm.The starting point of our research is to solve the difficulty of reconstruct phylogenetic tree and achieve the goals of multi-objective optimization problem.By analyzing the algorithm flow of the multi-objective evolutionary algorithm,we designed a highly efficient multi-objective reconstruction tree algorithm that can solve the slow convergence speed.The specific work can be summarized as follows:The first work of this paper designed a multi-objective evolutionary algorithm(MOEA-MC)combining membrane structure and consensus.In the initial stage of the multi-objective evolutionary algorithm,we divide the initial population into a specified number of membrane structures according to the weight vectors.Then,each membrane structure independently iterates through the evolution algorithm.At the end of each algorithm iteration,retaining the internal consensus branch of the membrane structure to the external solution,and then selecting the excellent solution in the membrane as the parents for the next iteration.Experiments verify that MOEA-MC can achieve excellent distribution and convergence speed.In the second work,we parallelized MOEA-MC.By studying the independence between the subpopulations in each time-consuming step of the multi-objective evolutionary algorithm,we chose to design the steps that do not require global information as parallel steps.The objective function and running time of MOEA-MC after parallelization are verified.Experiments prove that the parallelized MOEA-MC is superior to the traditional serial multi-objective evolutionary algorithm.In the third work,we first collected two data sets,one is the existing COVID-19 gene data,and the other is the sequence similarity obtained by BLAST search using the longest complementary palindrome sequence of COVID-19,and then use parallel MOEA-MC to reconstruct the evolutionary tree of COVID-19.At the molecular level,the possible origin and intermediate host of COVID-19 are speculated,which provides biologists with information that may be of reference value.