A scale-space filter approach to clustering dose-response curves of novel drug compounds

Automated dose-response curve classification methods need to be developed to keep up with the vast amount of data emanating from high throughput screening assays. We have devised a methodology based on scale-space filtering that allows tracking of salient features of dose-response curves at different scales. A complete and unbiased representation of dose-response data at various scale levels enables us to subsequently employ metrics that robustly classify the different functional categories of drug compounds that may exist in a novel screening assay. This powerful tool is intuitive and yet rigorous in that it distinguishes "important" characteristics of "different" curves. The method is non-heuristic, eliminates subjectivity, and is general enough to be readily employed in a wide variety of applications.