Key points are not available for this paper at this time.
Understanding how (targeted) drugs act on cellular systems is crucial for drug discovery and repurposing as wellas personalized drug recommendations in precision medicine. To address this, a variety of high-throughput drugassays are regularly employed, most notably phenotypic cell viability screens, activity-, affinity- or thermal stability-based drug-target binding assays, and proteome-wide drug-response profiling of post-translational modifications(PTMs). In such screens, it is often unclear whether or not a response significantly differs from a curve withoutregulation, making large-scale interpretation cumbersome. For example, treating potency and effect size estimatesfrom random and true curves with the same level of confidence leads to incorrect hypotheses and issues in trainingmachine learning models. Here, we present CurveCurator, an open-source software with interactive dashboard that provides reliableconcentration-response characteristics. It does this by computing p-values and false discovery rates based on arecalibrated F-statistic and a novel thresholding procedure. To demonstrate its broad utility, CurveCurator wasapplied to three concentration-response data sets (>450k curves) from different assays. Drug-phenotype, drug-target binding, and drug-PTM response were linked by taking advantage of CurveCurator's functionalities, such as,regulation classification, robust potency estimation, and interactive dashboards. This unified approach enableselucidation of the full cellular MoAs of drugs by leveraging the potency dimension of regulated curves. As anexample, for Afatinib in the cell line A431, the drug effect was traced from its inhibition of the carcinoma's maindriver (drug-target binding) to the shut-down of key downstream survival signals (drug-PTM response) and theeventual reduction in cell growth (drug-phenotype). CurveCurator is the first tool that provides reliable p-values for assessing the statistical likelihood of regulation inconcentration–response experiments. The objective categorization of concentration–response curves combinedwith an interactive dashboard accelerates data analysis and fills the need for a helpful tool in times of ever-increasing experimental throughput.
The et al. (Fri,) studied this question.