The Improvement Of DNA Evolution Models Containing Indel’s Information

Markov model of DNA evolution is one of the mathematical basics of molecular evolution and phylogenetic reconstruction based on the data of homologous DNA multiple sequence alignment. In recent 3 decades, evolutionary distance formula between homologous DNA sequences and nucleotide mutation probability matrix derivating by such models as JC69, K80, F81, HKY85, F84, TN93, GTR (REV) and UNREST, have been widely used in various distance methods and the maximum likelihood method to study DNA molecular evolution and to infer phylogenetic trees. However, these models involve only the information of nucleotide substitution events and don’t take the information of insertion or deletion of nucleotides mutation events in DNA evolutionary process into account.McGuire et al. (2001), regarding alignment gap as the fifth nucleotide state in markov process, introduced to the maximum likelihood method for the first time, they expand JC69, F81 and F84 models into JC69+gap, F81+gap and F84+gap models which containing 5 kinds of nucleotides substituting each other. In these models, alignment gap is represented a single site indel event. To some extent,these models have eliminated the defects of previous models, however, they simply regard an indel event as a base substitution event, didn’t distinguish them in nature.On the basis of McGuire’s method, this paper expanded JC69, F81, HKY85 and F84 models into JC69+gap’,F81+gap’,HKY85+gap’ and F84+gap’ model by introducing a new parameter to distinguish nucleotide substitution and indel mutation events. In order to compare the existing models and the improved models, real homologous DNA sequences as well as simulated sequences data made by computer program are both used as test data in phylogenetic reconstruction by the maximum likelihood method. Results show that:at first, in the aspect of likelihood ratio test, except for F84+gap performs slightly better than F84+gap’, all other improved models are superior to similar existing models; secondly, in terms of the accuracy of estimated branch length of reconstructed phylogenetic tree, all improved models are superior to existing models.