Abstract
Convergence is the process by which several species independently evolve similar traits. This evolutionary process is not only strongly related to fundamental questions such as the predictability of evolution and the role of adaptation, its study also may provide new insights about genes involved in the convergent character. We focus on this latter question and aim to detect molecular basis of a given phenotypic convergence. After pointing out a number of concerns about detection methods based on ancestral reconstruction, we propose a novel approach combining an original measure of the extent to which a site supports a phenotypic convergence, with a statistical framework for selecting genes from the measure of their sites. First, our measure of “convergence level” outperforms two previous ones in distinguishing simulated convergent sites from non-convergent ones. Second, by applying our detection approach to the well-studied case of convergent echolocation between dolphins and bats, we identified a set of genes which is very significantly annotated with audition-related GO-terms. This result constitutes an indirect evidence that genes involved in a phenotypic convergence can be identified with a genome-wide approach, a point which was highly debated, notably in the echolocation case (the latest articles published on this topic were quite pessimistic). Our approach opens the way to systematic studies of numerous examples of convergent evolution in order to link (convergent) phenotype to genotype.
