Abstract
Genetic drift is an important evolutionary force of strength inversely proportional to N e , the effective population
size. The impact of drift on genome diversity and evolution is known to vary among species, but quantifying this effect is a
difficult task. Here we assess the magnitude of variation in drift power among species of animals via its effect on the mutation
load – which implies also inferring the distribution of fitness effects of deleterious mutations. To this aim, we analyze the
non-synonymous (amino-acid changing) and synonymous (amino-acid conservative) allele frequency spectra in a large sample
of metazoan species, with a focus on the primates vs. fruit flies contrast. We show that a Gamma model of the distribution of
fitness effects is not suitable due to strong differences in estimated shape parameters among taxa, while adding a class of
lethal mutations essentially solves the problem. Using the Gamma + lethal model and assuming that the mean deleterious
effects of non-synonymous mutations is shared among species, we estimate that the power of drift varies by a factor of at least
500 between large- N e and small- N e species of animals, i.e., an order of magnitude more than the among-species variation in
genetic diversity. Our results are relevant to Lewontin’s paradox while further questioning the meaning of the N e parameter in
population genomics.