Molecular Clock: Testing

Abstract The molecular clock hypothesis originally rested on the assumption of rate constancy across lineages of a phylogeny, which would produce an approximately steady rate of accumulation of deoxyribonucleic acid or amino acid changes through time. This assumption has been questioned on the basis of increasingly large data sets, which have shown significant variability of rates in evolutionary lineages. To address this issue, tests have been developed to examine whether rates of molecular evolution vary significantly among taxonomic groups or phylogenetic lineages. Two major types of tests exist: those based on comparisons of genetic distances and those based on likelihood ratios. The first ones compare genetic distances between two species (or groups of species) relative to an outgroup; the latter ones compare maximum likelihood values for the same phylogeny calculated with and without the constant rate assumption. In those cases where the rate constancy assumption is violated, modern molecular clocks (relaxed clocks) are now being applied to implement the rate heterogeneity in the time estimation process. Key Concepts: The original formulation of the molecular clock hypothesis refers to the approximately steady rate of accumulation of DNA or amino acid changes through time. The assumption underlying the strict molecular clock is that evolutionary rates across lineages are constant. Molecular clock tests have been developed to assess the validity of the rate constancy assumption for DNA and amino acid sequences. Rate testing relies on a measure of genetic distance (branch length) for two or more lineages relative to an outgroup or on likelihood ratio tests. Distance‐based rate tests compare rates between two lineages or groups of lineages using the number of nucleotide or amino acid differences between them. Likelihood ratio tests use the chi‐square statistic to compare the likelihoods of a tree calculated with and without the rate constancy assumption. Unequal rates can be modelled by relaxed molecular clocks assuming two alternative models: ancestor and descendant lineages are autocorrelated or uncorrelated. Local rate testing using Bayes Factors is used to distinguish between autocorrelated and uncorrelated rate variation in Bayesian‐based relaxed clock methods. If evolutionary rates are heterogeneous among lineages, the violating lineages can be pruned from the data set or divergence times can be estimated with relaxed clock methods. Molecular clocks can be used to time divergences for which rare or incomplete palaeontological records exist.