Ancestral Reconstruction

Ancestral reconstruction is the extrapolation back in time from measured characteristics of individuals (or populations) to their common ancestors. It is an important application of phylogenetics, the reconstruction and study of the evolutionary relationships among individuals, populations, or species to their ancestors. In the context of biology, ancestral reconstruction can be used to recover different kinds of ancestral character states, including the genetic sequence (ancestral sequence reconstruction), the amino acid sequence of a protein, the composition of a genome (e.g., gene order), a measurable characteristic of an organism (phenotype), and the geographic range of an ancestral population or species (ancestral range reconstruction). Nonbiological applications include the reconstruction of the vocabulary or phonemes of ancient languages [1] and cultural characteristics of ancient societies such as oral traditions [2] or marriage practices [3]. Ancestral reconstruction relies on a sufficiently realistic model of evolution to accurately recover ancestral states. No matter how well the model approximates the actual evolutionary history, however, one's ability to accurately reconstruct an ancestor deteriorates with increasing evolutionary time between that ancestor and its observed descendants. Additionally, more realistic models of evolution are inevitably more complex and difficult to calculate. Progress in the field of ancestral reconstruction has relied heavily on the exponential growth of computing power and the concomitant development of efficient computational algorithms (e.g., a dynamic programming algorithm for the joint maximum likelihood [ML] reconstruction of ancestral sequences [4]). Methods of ancestral reconstruction are often applied to a given phylogenetic tree that has already been inferred from the same data. While convenient, this approach has the disadvantage that its results are contingent on the accuracy of a single phylogenetic tree. In contrast, some researchers advocate a more computationally intensive Bayesian approach that accounts for uncertainty in tree reconstruction by evaluating ancestral reconstructions over many trees [5].