Files
Abstract
Phylogenetic trees are fundamental tools for understanding the evolution of species. A number of computational approaches have been developed to estimate phylogenetic trees from molecular sequence data. Although maximum likelihood methods can accurately reconstruct phylogenetic trees, they face computational challenges when searching for optimal solutions in a huge tree space. The computation time for maximum likelihood approaches increases exponentially as the size of molecular sequences increases. In this paper, I propose a method called Maximum K-subtree pseudo-likelihood estimate (MPLE) in which the K-subtree pseudo-likelihood function is used to replace the full likelihood function. The identifiablility and consistency of MPLE method guarantees its accuracy. To learn the lose of efficiency caused by mis-specified assumption by MPLE, fisher information of branch is compared between full likelihood function and pseudo-likelihood function. At last, I construct the MPLE algorithm and compare it with the most popular maximum-likelihood methods RAxML. Extensive experiments on simulated datasets are conducted to evaluate our approach’s effectiveness in terms of time complexity and accuracy with respect to mean square error. The results show that our approach MPLE would obtain accurate estimate and cost less time, compared with representative methods.