PSeq-Gen: an application for the Monte Carlo simulation of protein sequence evolution along phylogenetic trees.

Abstract

PSeq-Gen will simulate the evolution of protein sequences along evolutionary trees following the procedures previously reported for the DNA sequence simulator Seq-Gen (Sequence-Generator, Rambaut and Grassly, 1997). Statistics calculated from these sequences can be used to give expectations under specific null hypotheses of protein evolution. This Monte Carlo simulation approach to testing hypotheses is often termed 'parametric bootstrapping' (see, for example, Efron, 1985; Huelsenbeck et al., 19%; Huelsenbeck and Rannala, 1997), and has many powerful applications, such as testing the molecular clock (Goldman, 1993), detecting recombination (Grassly and Holmes, 1997), and evaluating competing phylogenetic hypotheses (Hillis et al., 19%). Three common models of amino acid substitution are implemented (PAM: Dayhoff et al., 1978; JTT: Jones et al., 1992; mtREV: Adachi and Hasegawa, 1995). These models use instantaneous rate matrices derived from observed patterns of accepted point mutations for aligned nuclear (PAM, JTT) and mitochondrial (mtREV) genes. The amino acid frequencies found for these alignments are the default in each model, but it is also possible to specify their frequencies independently in the implementation here [following Adachi (1995), but see also Kishino et al. (1990)]. In addition, sitespecific rate heterogeneity following a gamma distribution is allowed (as described in Rambaut and Grassly, 1997; Yang, 1993). As with Seq-Gen, any number of trees may be read in and any number of data sets can be simulated for each tree, allowing large sets of replicate simulations to be created easily. Maximum-likelihood algorithms exist which will reconstruct phylogenies from protein sequences using the PAM, JTT and mtREV substitution models (e.g. Adachi and Hasegawa, 1995; Yang, 1996), and hence bootstrapping to obtain confidence limits about phylogenetic parameters is easily achieved.