The relationship between <i>h</i> and <i>s</i> inferred from <i>Arabidopsis</i> does not fit the human nonsynonymous SFS.

<p>Top: Comparison of dominance parameters for models considered, including a fully additive model, a model with a monotonic decay in <i>h</i> that is within 1.92 LL units of the MLE, and the model from Huber et al. [<a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011198#pgen.1011198.ref025" target="_blank">25</a>] estimated for <i>Arabidopsis</i>. Middle: Model fit in comparison to the additive model. Note that the additive model and monotonic models have similar log-likelihoods though the Huber et al. [<a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011198#pgen.1011198.ref025" target="_blank">25</a>] model has a much worse log-likelihood. Bottom: Discrete DFE parameters estimated for each dominance model when fit to the human nonsynonymous SFS. Y-axis shows the proportion of mutations in each bin of the discrete DFE.</p> <p>(PDF)</p>