Identification of multivariate phenotypes most influenced by mutation: Drosophila serrata wings as a case study

The distribution of pleiotropic mutational effects impacts phenotypic adaptation. However, small effect sizes and high sampling error of covariances hinder investigations of the factors influencing this distribution. Here, we explored the potential for shared information across traits affected by the same mutations to counter sampling error, allowing robust characterization of patterns of mutational input. Exploiting a published dataset representing 12 samples of the same mutation accumulation experiment in Drosophila serrata, we inferred robust signals of mutational effects from the concordance across samples. Krzanowski's common subspace analysis identified a multivariate wing trait with statistically supported mutational variance in all samples. Importantly, this multivariate trait was aligned with the major axis of among-line (mutational) variance within most population samples. That is, despite considerable heterogeneity among samples in individual (co)variance parameter estimates, the predominant pattern of correlated mutational effects was identified in each dataset. 2 other multivariate traits were statistically supported across most samples. Smaller effect sizes (lower mutational variance) with concomitant larger sampling error or other factors (e.g., microenvironmental dependence of effects) may reduce the robustness of estimated mutational input for these traits. Overall, our results suggest that multivariate analyses of mutation accumulation experiments can detect the true signal of pleiotropic mutation, and that sampling error does not preclude such studies from extending our knowledge of pleiotropic mutational effects.