(1130-C) A machine learning-assisted image-based profiling methodology for phenotypic drug discovery
Tuesday, February 6, 2024
12:00 PM – 1:00 PM EST
Location: Exhibit Halls AB
Abstract: Image-based profiling assays have enabled high-throughput phenotypic profiling of cell populations perturbed by various treatments, yet the downstream data interpretation remains a challenge. We have developed a machine learning-assisted phenotypic screening methodology based on an image-based cell painting assay. High quality cell images were acquired with optimized protocols. To establish a robust and automated workflow for image data analysis, quality control, and biological interpretation, we performed a pilot screen of almost 400 compounds targeting kinase proteins. The results demonstrate that this methodology can accurately and automatically capture single-cell morphological profiles induced by various perturbations. Moreover, it could cluster individual cells exposed to compounds with a similar target or mechanism of action based on cytological profiles. This image-based profiling methodology provides an automated and unbiased approach to characterize disease- or treatment-associated cellular phenotypes and offers a new approach to novel hit identification.