Fig. 6

Cut-offs can be defined based on the distribution of screen data. Screens which result in a high number of growth defects are more accurately described using a bimodal mixture model and are characterised by a ‘central peak’ and a second ‘hit peak’. Bimodal mixture models can be fitted automatically to screens with many growth defects using ScreenGarden and produce a component plot (A) and a fit plot (B) as well as q-values for each LGR. If q ≥ 0.5, the data is predicted to follow the distribution of component 2 and thus LGRs account for predicted growth defects. C Mixture model, Z-transformation and empirically defined LGR cut-offs were compared for the Nop10 Synthetic Physical Interactions dataset. Cut-off definition using a bimodal mixture model predicted more than twice the number of growth defects compared to Z-transformation or LGR-based thresholds