A new phylogenetic protocol: Dealing with model misspecification and confirmation bias in molecular phylogenetics

Lars S Jermiin; Renee A Catullo; Barbara R Holland

The choice of phylogenetic method implies accepting the assumptions on which the method rests. For example, it is often assumed that the sequences evolved along a single bifurcating tree and that the evolutionary processes operating at the variable sites in the sequences are independent and identically-distributed processes. If model-based molecular phylogenetic methods are chosen, the underlying assumption usually is that the evolutionary processes operating at the variable sites can be approximated accurately by using Markov models that are stationary, reversible, and homogeneous (63-65) over time (the assumption of evolution under SRH conditions). In practice, the choice is one between methods assuming that the underlying evolutionary processes can be modelled using a Markov model of nucleotide or amino-acid substitutions (i.e., distance methods (66-71), likelihood methods (66,68,70-76), analysis. However, if a researcher's expectations are met by the phylogenetic results, it is more likely that a report will be written without a further assessment of what might have gone wrong during the analysis. This tendency-allow prior expectations to influence the interpretation of phylogenetic estimates-is called confirmation bias. Confirmation bias is not discussed in phylogenetics, even though it is well recognized as a critical factor in other disciplines (e.g., psychology and social science (121)), so it is timely that the phylogenetic community takes onboard the serious implications of this.