Figure 9

Quantitative cell morphology phenotypes are predicted significantly better than random expectation. In contrast, genes whose disruption decreases population co-efficient of variance (CV) were not predictable. (a) A histogram plotting the distribution of the area under the receiver operating characteristic (ROC) curve (AUC) values for 562 quantitative morphological phenotypes shows a significantly higher proportion of high AUC values than for 1,000 size-matched random gene sets. (b) Separate analyses of phenotypes associated with morphologic features and phenotypes associated with cell-to-cell variability in the morphologic features reveals asymmetry in predictability. Sets of genes whose disruption causes the 40 largest or smallest mean values of a morphological feature (middle plots) are significantly more predictable than random gene sets (left side). By contrast, although the sets of genes whose disruption most increase the CV tend to be predictable (high AUC), those that most decrease the CV are not (low AUC). Box-and-whisker plots are drawn as in Figure 3. (c) A comparison of the median phenotypic CVs observed for deletion strains versus replicate analyses of wild-type cells shows that deletion strains with the most reduced CVs are essentially wild-type-like in character, whereas those with the most increased CVs show significantly more cell-to-cell variability than wild-type cells. These latter knockout strains carry deletions for genes predominantly involved in maintaining genomic integrity. This trend is therefore likely to have arisen from nonclonal genetic variation in these strains, recapitulating the classic mutator phenotype.