Research On Phylogenetic Tree Reconstruction Method Based On Deep Neural Network

Phylogenetic tree is a tree-like diagram used to represent the evolutionary relationship between species or genes.In the past few decades,the theory of phylogenetic analysis has developed rapidly,making phylogenetic tree an indispensable research tool in various branches of biology,and has been accepted and learned by biologists in various fields.Biologists use phylogenetic trees to express the relationship between species or genes,which can help to reveal the origin and transmission of viral infectious diseases,the migration patterns of species in a certain period and geographical range,and the information of microbial communities in macrogenomic data.But the reconstruction of phylogenetic tree has proved difficult in theory.How to reconstruct the effective phylogenetic tree quickly is still a hot topic in the field of computational biology.In this paper,we propose a phylogenetic tree reconstruction method based on deep neural network under the framework of minimum evolution principle.Transformer model was used to construct the attention network model in our method.The reconstruction method mainly consists of three parts.First,the raw DNA sequence is embedded into vector space by using k-mer model.Then,the attention model is used to reconstruct the optimal circular permutation.Finally,the circular permutation is transformed into the optimal phylogenetic tree.The main innovations of our method are as follows.First,we transform the reconstruction of phylogenetic tree into the reconstruction of circular permutation,thus avoiding the calculation of distance matrix and the estimation of branch length.Secondly,we take the lead in applying neural network to the reconstruction of phylogenetic tree,and learn the heuristic reconstruction strategy of phylogenetic tree with the learning ability of neural network.The experimental results show that our reconstruction method has surpassed one traditional method in small-scale phylogenetic tree reconstruction problem,and is close to the theoretical optimal algorithm of minimum evolution principle.And,our reconstruction method is three times faster than other traditional methods.