No evidence that synonymous mutations in yeast genes are mostly deleterious

In a recent paper1, Shen et al. reported that most mutations in the coding regions of 21 yeast genes were strongly deleterious, and that the distributions of fitness effects were similar for synonymous and nonsynonymous mutations. Taken at face value, these results would conflict with well-established findings from a broad range of fields and approaches. Here, we argue that these results arose from a lack of appropriate controls for the impacts of background genetic effects in edited strains. A re-examination of the data in Shen et al. strongly suggests that it is entirely consistent with the expectation that most nonsynonymous and nearly all synonymous mutations have no detectable effects on fitness. We present analyses which show that the data is inconsistent with the proposed explanation that pervasive fitness effects of synonymous mutations arise from their effects on mRNA levels, that the sequence-based fitness assay overestimates fitness effects compared to the growth-based fitness assay, and that the observed wide fitness distributions for nonsense mutations are consistent with ‘off-target’ effects or other uncontrolled sources of biological variation contributing to measured fitness. We conclude by discussing the essential controls and other experimental design considerations that are required to produce interpretable results regarding the fitness effects of mutations in large-scale screens.