| As a replacement for laborious visual assays, automated image analysis can ease comparison between images, but computational approaches also have the broader potential to allow investigation of otherwise intractable classes of phenotypic variation. I here describe computational methods that ease high throughput screening and basic cell biological analysis, and show that these represent progress toward strategies for profiling complex phenotypic response to drug treatment.; First, I describe a set of methods to ease analysis of microtubule dynamics in fluorescence speckle microscopy (FSM)-labelled Xenopus egg extract spindles, a system that is complicated by extract heterogeneity and unconstrained spindle movement. Further difficulties in FSM studies of internal spindle dynamics come from high speckle turnover and overlapping speckle flows. I describe automated techniques for correcting spindle drift and determining the movement of speckle ensembles, which we have used to examine microtubule flux during spindle assembly and perform a quantitative analysis of the relationship between flux and activity of the spindle kinesin Eg5.; I next describe the automation of otherwise impractical image-based high-throughput screens. In the first, a screen for small molecule modulators of the activity of the Eg5 inhibitor monastrol, we identified new mitotic kinesin inhibitors and two potent monastrol suppressors of unknown mechanism. We further observed a complex interaction with microtubule inhibitors that offers clues about the mechanisms underlying spindle morphogenesis and illustrates the complexity of phenotypic drug-response.; In another screen, we developed a centrosome counting assay that allowed the identification of new inhibitors of centrosome duplication and connects a number of unexpected pathways to this process. We observed that a simple cytometric profiling strategy based on multidimensional phenotypic analysis groups compounds with similar mechanisms, suggesting a strategy for finding compounds with undesirable targets.; Finally, we expand this approach by presenting a method for high-throughput cytological profiling by microscopy. We generated quantitative, multi-dimensional measures of individual cell states over wide ranges of drug-induced perturbations. Profiling dose-dependent phenotypic effects of drugs, we successfully categorized blinded drugs and suggested targets for drugs of uncertain mechanism. These methods should prove directly helpful for the tasks of predicting the target and toxicity of new drugs. |
