Presented at SLAS, San Diego, CA, USA
Drug research strives to increase the physiological relevancy of experimental models in early discovery, e. g. by using material from human donors for HCS. Using such physiologically relevant cell models can however introduce challenges due to the limited amount of material available for experimentation and the variation between donors.
In this poster, we show how our recently introduced Deep Learning-based high content analysis software remedies this situation by successfully transferring knowledge obtained from previous experiments to a new experimental setting, reducing the time taken for assay and image analysis optimization. It also robustly identifies relevant phenotypes under varying experimental conditions, important for screens combining material from multiple donors. We illustrate the approach with data from a human renal assay and show that within a few minutes knowledge is being transferred between screening modalities, crossing the boundaries between screening modality and between different human donors. In comparing the results with conventional approaches typical for current high-content screening projects, we show a significant improvement in speed and result quality. This new approach gives unmatched ease and performance and allows a much faster delivery on R&D projects, thereby significantly reducing cycle times with human-derived models.