Simple and accurate estimation of ancestral protein sequences

There are a variety of reasons to reconstruct the sequences of ancient proteins, but whatever the reason, the value of the reconstructed protein depends on the accuracy with which the ancient sequence is inferred. This study uses sequences simulated by a sequence-evolution simulation program that compares parsimony, maximum likelihood, and the Bayesian methods of inferring ancestral sequences and concludes that the Bayesian method, as implemented by MRBAYES 3.11, is preferred. Estimated ancestral sequences are of necessity the same length as the alignment on which the underlying phylogeny is based. A highly accurate method for correcting the estimated sequences is introduced, and it is shown that the correction permits inferring the sequences of ancient protein sequences with a very high degree of accuracy.