Explicit evolutionary models are required in maximum-likelihood andBayesian inference, which two methods are regularly used in phylogenetic studies of DNA sequence data nowadays. Appropriate selection of nucleotide substitution models for multiple sequences after their alignment has become standard procedure in phylogenetic analysis. For the most widely used selection program i.e., ModelTest, there are actually more than one selection criteria, such as hierarchical likelihood-ratio test (hLRT), Akaike information criterion (AIC), and Bayesian information criterion (BIC). Therefore, researchers are usually confronted with one puzzle that which criterion is better for their analysis.

When doing extensive DNA phylogenic analyses and DNA taxonomy on various taxa, a group led by Dr. Chao-Dong ZHU also came across tackling this question. To better understand the performance of different model-selection criteria, they used 33,600 simulated data sets to analyze the accuracy, precision, dissimilarity, and biases of the hLRT, AIC, BIC, and DT. They demonstrated that the BIC and DT are the most appropriate model-selection criteria because of their high accuracy and precision. The results also indicate that in some situations different models are selected by different criteria for the same dataset. Such dissimilarity was the highest between the hLRT and AIC, and lowest between the BIC and DT. The hLRT performed poorly when the true model included a proportion of invariable sites, while the BIC and DT generally exhibited similar performance to each other. On the whole, the results indicate that the BIC and DT should be preferred for model selection. Together with model-adequacy tests, accurate model selection will serve to improve the reliability of phylogenetic inference and related analyses.

This project has been mainly done by researchers from the group led by Dr Chao-Dong ZHU in the Institute of Zoology, Chinese Academy of Sciences. It was supported mainly by the Innovation Program in the Chinese Academy of Sciences (KSCX2-YW-NF-02), grants from the National Science Foundation, China (30870268, J0930004), and partially by Public Welfare Project from the Ministry of Agriculture, China (Grant No. 200803006), and Beijing Municipal Natural Science Foundation Key Projects (No. KZ201010028028). The research paper has been published in BMC Evolutionary Biology (IF, 4.29). Since it was online, it has been highly accessed for nearly 1,200 times till now. With the grant from China Scholarship Council (CSC), A-Rong LUO, the first author, one PhD candidate student of the Graduate University, CAS, has been in the Australia Commonwealth Scientific and Research Organization (CSIRO) to collaborate with Australian researchers and do further researches.

Arong Luo, Huijie Qiao, Yanzhou Zhang, Weifeng Shi, Simon YW Ho, Weijun Xu, Aibing Zhang, Chaodong Zhu. 2010. Performance of criteria for selecting evolutionary models in phylogenetics: a comprehensive study based on simulated datasets. BMC Evolutionary Biology. 10: 242.

Internet access: http://www.biomedcentral.com/1471-2148/10/242/abstract