(1256-A) Deep Learning-based Morphological Profiling as a Screening Method for Human Macrophage Polarization Modulators
Monday, February 5, 2024
12:00 PM – 1:00 PM EST
Location: Exhibit Halls AB
Abstract: Macrophage biology underlies many disease states in both immunology and oncology. The identification of novel medicines to treat these disease areas has slowed, potentially due to the lack of identified druggable targets to investigate. We hope to address this gap by employing novel morphological profiling techniques on primary human macrophages to identify modulators that have been missed by previous efforts. Specifically, this workflow utilizes M0, M1, and M2 macrophages as control states, followed by cell painting and high throughput imaging. Analysis is done with traditional image segmentation-based techniques for quality control prior to applying a deep learning convolutional neural network (CNN). Features extracted from the CNN generate meaningful signatures of biological states which may facilitate prediction of mechanisms of action (MoAs). As a proof of concept, we used this pipeline to demonstrate both genetic and pharmacological inhibition of JAK1 reduce the appearance of M1 morphology in GM-CSF / LPS/ IFN induced macrophages. Encouraged by these results, we screened a library containing 1579 compounds with known mechanisms of action and 3840 compounds selected for chemical diversity. Examination of morphological profiles with annotated activity enabled nomination of potential targets that alter macrophage morphology. Simultaneously, an arrayed CRISPR screen of 800 gene knockouts was performed to identify novel gene targets that may play a role in macrophage morphology. The results identified several new potential target mechanisms that alter macrophage biology. Taken together, these data demonstrate that morphological profiling can be applied to primary human cells to identify novel small molecules and genetic targets for the modulation of macrophage polarization.