Assessing the Accuracy of Ancestral Protein Reconstruction Methods

Abstract

The phylogenetic inference of ancestral protein sequences is a powerful technique for the study of molecular evolution, but any conclusions drawn from such studies are only as good as the accuracy of the reconstruction method. Every inference method leads to errors in the ancestral protein sequence, resulting in potentially misleading estimates of the ancestral protein's properties. To assess the accuracy of ancestral protein reconstruction methods, we performed computational population evolution simulations featuring near-neutral evolution under purifying selection, speciation, and divergence using an off-lattice protein model where fitness depends on the ability to be stable in a specified target structure. We were thus able to compare the thermodynamic properties of the true ancestral sequences with the properties of “ancestral sequences” inferred by maximum parsimony, maximum likelihood, and Bayesian methods. Surprisingly, we found that methods such as maximum parsimony and maximum likelihood that reconstruct a “best guess” amino acid at each position overestimate thermostability, while a Bayesian method that sometimes chooses less-probable residues from the posterior probability distribution does not. Maximum likelihood and maximum parsimony apparently tend to eliminate variants at a position that are slightly detrimental to structural stability simply because such detrimental variants are less frequent. Other properties of ancestral proteins might be similarly overestimated. This suggests that ancestral reconstruction studies require greater care to come to credible conclusions regarding functional evolution. Inferred functional patterns that mimic reconstruction bias should be reevaluated.

Constraining the timing of the Great Oxidation Event within the Rubisco phylogenetic tree

“…We characterized the support for these ancestors by binning their posterior probabilities of states into 10% sized bins and counting the proportion of ancestral sites in each bin. We also generated alternate versions of the ancestral sequences by randomly sampling from their posterior distributions to generate five alternate ancestors for every node, as described (Williams, Pollock, Blackburne, & Goldstein, 2006). …”

Reconstructed Ancestral Enzymes Impose a Fitness Cost upon Modern Bacteria Despite Exhibiting Favourable Biochemical Properties

“…Notwithstanding these caveats, this correlation remains interesting and certainly worthy of further investigation. Williams et al (2006) have previously reported that maximum likelihood (ML) inference methods, such as those used in the inference of LeuB, may have a tendency to gradually overestimate the stability of ancestral proteins along the phylogenetic gene tree. ML is known to be biased in its estimation of ancestral characters.…”

“…A second known bias with ML methods is the tendency to incorporate into ancestral sequences the states that have the highest equilibrium frequency in the employed substitution model (Yang 2006). Thus, it has been suggested that this bias may result in a preponderance of hydrophobic residues in ancestral sequences, as hydrophobic residues tend to have high equilibrium frequencies in substitution models (Gaucher et al 2008;Williams et al 2006). This, in turn, could affect the properties of an ancestral protein.…”

“…In addition, increased protein stability is often accompanied by a trade-off in terms of catalytic activity (Somero 1995), although this appears not to be the case for many inferred ancestral enzymes. From an evolutionary perspective, assuming that the increased thermostability of ancestral proteins is not the result of a bias in the in silico reconstruction method (Williams et al 2006; see ''Result and Discussion'' section), the marginal stability of contemporary proteins may imply that an unknown selection pressure has caused the stability of proteins to decrease during evolution, or simply that stability is a neutral trait (Bloom et al 2007;Taverna and Goldstein 2002;Zeldovich et al 2007). For example, as the majority of mutations are destabilising rather than stabilising, it has been argued that proteins only retain high stability if it is evolutionarily beneficial (Taverna and Goldstein 2002).…”

Superoxide dismutase 1 is positively selected to minimize protein aggregation in great apes

“…Computational ancestral protein reconstruction is sensitive to errors when amino acids less likely to be in the ancestral protein still figure prominently in extant sequences. This can affect 7 the predicted chemical properties and lead to overestimated stability of ancestors because more likely amino acids tend to be more stable [49]. As we use the Bayesian approach this is a minor issue in our work, as also confirmed from the observed opposite trend towards increased stability of extant great ape proteins.…”