Branch-site Method Research Articles

Coevolution of insulin-like growth factors, insulin and their receptors and binding proteins in New World Monkeys

Previous work has shown that the evolution of both insulin-like growth factor 1 (IGF1) and insulin shows an episode of accelerated change on the branch leading to New World Monkeys (NWM). Here the possibility that this is accompanied by a corresponding episode of accelerated evolution of IGF1 receptor (IGF1R), insulin receptor (IR) and/or IGF binding proteins (IGFBPs) was investigated. Analysis of receptor sequences from a range of primates and some non-primate mammals showed that accelerated evolution did indeed occur on this branch in the case of IGF1R and IR, but not for the similar insulin receptor–related receptor (IRRR) which does not bind insulin or IGF1. Marked accelerated evolution on this branch was also seen for some IGFBPs, but not the mannose 6-phosphate/IGF2 receptor or epidermal growth factor receptor. The rate of evolution slowed before divergence of the lineages leading to the NWM for which sequences are available (Callithrix and Saimiri). For the IGF1R and IR, the accelerated evolution was most marked for the extracellular domains (ectodomains). Application of the branch-site method showed dN/dS ratios significantly greater than 1.0 for both receptor ectodomains and for IGFBP1, and allowed identification of residues likely to have been subject to selection. These residues were concentrated in the N-terminal half of the IGF1R ectodomain but the C-terminal half of the IR ectodomain, which could have implications for the formation of hybrid receptors. Overall the results suggest that adaptive coevolution of IGF1, insulin and their receptors and some IGFBPs occurred during the evolution of NWM. For the most part, the residues that change on this branch could not be associated with specific functional aspects (ligand binding, receptor dimerization, glycosylation) and the physiological significance of this coevolution remains to be established.

Performance of Standard and Stochastic Branch-Site Models for Detecting Positive Selection among Coding Sequences

The branch-site model is a widely popular approach that accommodates for the lineage- and the site-specific heterogeneity of natural selection regimes among coding sequences. This model relies on prior knowledge of the (foreground) lineage(s) evolving under positive selection at some sites. Unfortunately, such prior information is not always available in practice. A more recent technique (Guindon S, Rodrigo A, Dyer K, Huelsenbeck J. 2004. Modeling the site-specific variation of selection patterns along lineages. Proc Natl Acad Sci USA 101:12957-12962) alleviates this issue by explicitly modeling the variability of selection patterns using a stochastic process. However, the performance of this approach for deciding whether a set of homologous sequences evolved under positive selection at some point has not been assessed yet. This study compares the sensitivity and specificity of tests for positive selection derived from both the standard and the stochastic approaches using extensive simulations. We show that the two methods have low proportions of type I errors, that is, they tend to be conservative when testing the null hypothesis of no positive selection if sequences truly evolve under neutral or negative selection regimes. Also, the standard approach is more powerful than the stochastic one when the prior knowledge on foreground lineages is correct. When this prior is incorrect, however, the stochastic approach outperforms the standard model in a broad range of conditions. Additional comparisons also suggest that the stochastic branch-site method compares favorably with the recently proposed mixed-effects model of evolution of Murrell et al. (Murrell B, Wertheim JO, Moola S, Weighill T, Scheffler K, Pond SLK. 2012. Detecting individual sites subject to episodic diversifying selection. PLoS Genet. 8:e1002764). Altogether, our results show that the standard branch-site model is well suited to confirmatory analyses, whereas the stochastic approach should be preferred over the standard or the mixed-effects ones for exploratory studies.

Reply to Nozawa et al.: Complementary statistical methods support positive selection of a duplicated UV opsin gene in Heliconius

Statistical methods used to test for positive selection have a long history and continue to evolve (1–4). In their letter, Nozawa et al. (5) question our use of the branch-site method in our recent paper in PNAS (6). As experimental biologists, we welcome all methods that facilitate the detection of interesting parts of the genome for functional exploration. In this case, a statistically significant result using the branch-site method combined with structural modeling and the identification of a few biochemically relevant substitutions provided us with an incentive for the in vivo physiological characterization of the UV-sensitive rhodopsins in Heliconius. Had we not had that first hint from sequence data alone, it is unlikely that we would have made that functional discovery, which makes the eyes of Heliconius unique compared with all other studied butterflies.

Response to Yang et al.: Problems with Bayesian methods of detecting positive selection at the DNA sequence level

It is unfortunate that Yang et al. (1) misrepresent the contents and conclusions of our recent study (2). Our study was motivated by the recent publication of papers reporting the detection of positive selection in many genes of the human lineage by using the branch-site method (BSM) and others, where the ratio (ω) of nonsynonymous to synonymous nucleotide substitutions is used as a measure of positive selection (e.g., ref. 3). Because the number of nucleotide substitutions per gene (n) in the human lineage after separation from the chimpanzee is very small (n ≈ 4) and BSM depends on the large sample theory, we examined the theoretical basis of BSM in comparison with the small-sample method (SSM) using Fisher's exact test (4), which gives accurate statistical tests for small samples.

Reliabilities of identifying positive selection by the branch-site and the site-prediction methods

Natural selection operating in protein-coding genes is often studied by examining the ratio (omega) of the rates of nonsynonymous to synonymous nucleotide substitution. The branch-site method (BSM) based on a likelihood ratio test is one of such tests to detect positive selection for a predetermined branch of a phylogenetic tree. However, because the number of nucleotide substitutions involved is often very small, we conducted a computer simulation to examine the reliability of BSM in comparison with the small-sample method (SSM) based on Fisher's exact test. The results indicate that BSM often generates false positives compared with SSM when the number of nucleotide substitutions is approximately 80 or smaller. Because the omega value is also used for predicting positively selected sites, we examined the reliabilities of the site-prediction methods, using nucleotide sequence data for the dim-light and color vision genes in vertebrates. The results showed that the site-prediction methods have a low probability of identifying functional changes of amino acids experimentally determined and often falsely identify other sites where amino acid substitutions are unlikely to be important. This low rate of predictability occurs because most of the current statistical methods are designed to identify codon sites with high omega values, which may not have anything to do with functional changes. The codon sites showing functional changes generally do not show a high omega value. To understand adaptive evolution, some form of experimental confirmation is necessary.

Codon-based tests of positive selection, branch lengths, and the evolution of mammalian immune system genes

Using basic probability theory, we show that there is a substantial likelihood that even in the presence of strong purifying selection, there will be a number of codons in which the number of synonymous nucleotide substitutions per site (d (S)) exceeds the number of non-synonymous nucleotide substitutions per site (d (N)). In an empirical study, we examined the numbers of synonymous (b (S)) and non-synonymous substitutions (b (N)) along branches of the phylogenies of 69 single-copy orthologous genes from seven species of mammals. A pattern of b (N) > b (S) was most commonly seen in the shortest branches of the tree and was associated with a high coefficient of variation in both b (N) and b (S), suggesting that high stochastic error in b (N) and b (S) on short branches, rather than positive Darwinian selection, is the explanation of most cases where b (N) is greater than b (S) on a given branch. The branch-site method of Zhang et al. (Zhang, Nielsen, Yang, Mol Biol Evol, 22:2472-2479, 2005) identified 117 codons on 35 branches as "positively selected," but a majority of these codons lacked synonymous substitutions, while in the others, synonymous and non-synonymous differences per site occurred in approximately equal frequencies. Thus, it was impossible to rule out the hypothesis that chance variation in the pattern of mutation across sites, rather than positive selection, accounted for the observed pattern. Our results showed that b (N)/b (S) was consistently elevated in immune system genes, but neither the search for branches with b (N) > b (S) nor the branch-site method revealed this trend.