Evolutionary Model Selection with a Genetic Algorithm: A Case Study Using Stem RNA

Abstract

The choice of a probabilistic model to describe sequence evolution can and should be justified. Underfitting the data through the use of overly simplistic models may miss out on interesting phenomena and lead to incorrect inferences. Overfitting the data with models that are too complex may ascribe biological meaning to statistical artifacts and result in falsely significant findings. We describe a likelihood-based approach for evolutionary model selection. The procedure employs a genetic algorithm (GA) to quickly explore a combinatorially large set of all possible time-reversible Markov models with a fixed number of substitution rates. When applied to stem RNA data subject to well-understood evolutionary forces, the models found by the GA 1) capture the expected overall rate patterns a priori; 2) fit the data better than the best available models based on a priori assumptions, suggesting subtle substitution patterns not previously recognized; 3) cannot be rejected in favor of the general reversible model, implying that the evolution of stem RNA sequences can be explained well with only a few substitution rate parameters; and 4) perform well on simulated data, both in terms of goodness of fit and the ability to estimate evolutionary rates. We also investigate the utility of several distance measures for comparing and contrasting inferred evolutionary models. Using widely available small computer clusters, our approach allows, for the first time, to evaluate the performance of existing RNA evolutionary models by comparing them with a large pool of candidate models and to validate common modeling assumptions. In addition, the new method provides the foundation for rigorous selection and comparison of substitution models for other types of sequence data.

HIV-Specific Probabilistic Models of Protein Evolution

“…To formally characterize the similarities in the substitution process across the eight matrices, we computed a neighbor-joining tree on the Markov processes defined by each matrix using the total variation metric (TVM) [31]. Briefly, given a specific evolutionary time scale, TVM computes the distance between the expected distributions of characters generated under the two evolutionary models.…”

“…The c-AIC score of a model is defined as c-AIC = 2(− L + ps /( s − p −1)), where L is the log-likelihood score of the model, p is the number of estimated model parameters, and s is the number of independent samples. There are a number of possible ways to estimate the number of independent samples in the alignment [32] and we chose to use the number of alignment columns as an estimate of s. c-AIC performed well in selecting appropriate evolutionary models on biological and simulated alignments of paired RNA sequences [33]. In 2/47 cases, the HIV-W m model with the frequencies from the training set was the best and in 1/47 cases, the HIV-B m model was the best.…”

Paraphyly and budding speciation in the hairy snail (Pulmonata, Hygromiidae)

“…Average p-distances (pairwise exclusion of gaps) were calculated using MEGA version 4 (Tamura et al 2007), which was also used to calculate neighbour-joining (NJ) trees (Saitou & Nei 1987). For the NJ tree of the nc sequences, we used mid-point rooting.…”

Hepatitis C virus in healthy blood donors in Sri Lanka

“…The sequences were aligned in the BioEdit sequence alignment editor version 7.0.9.0[10] by using the Clustal W Multiple alignment. [11] Phylogenetic trees for HCV which were based on Core/E1 and NS5B sequences and genetic distances were calculated with MEGA software version 4[12] using the Maximum Likelihood model. The sequences of Core/E1 and NS5B of HCV strains in Sri Lanka were deposited in NCBI GenBank under the accession numbers given in Table 1.…”