Morphological Subprofile Analysis for Bioactivity Annotation of Small Molecules

Abstract

Fast prediction of mode of action for bioactive compounds would immensely foster bioactivity annotation in compound collections and may early on reveal off-targets in chemical biology research and drug discovery. A variety of target-based assays is available for addressing the modulation of druggable proteins. However, they cannot precisely predict how a compound would influence cellular processes due to polypharmacology. Furthermore, non-protein targets are often not considered. Morphological profiling, e.g., using the Cell Painting assay that monitors hundreds of morphological features upon compound perturbation and staining of cellular components, offers a fast, unbiased assessment of compound activity on various targets and cellular processes in one single experiment. However, due to incomplete bioactivity annotation and unknown activities of reference (landmark) compounds, prediction of bioactivity is not straightforward. Here we introduce the concept of subprofile analysis to map the mode of action for both reference and unexplored compounds. We defined mode-of-action clusters for a group of reference compounds and extracted cluster subprofiles that contain only a subset of morphological features (i.e., subprofiles) to represent a consensus profile. Subprofile analysis allows for assignment of compounds to, currently, ten different targets or modes of action in one single assay and bypasses the need of exploring all biosimilar reference compounds for the generation of target hypothesis. This approach will enable rapid bioactivity annotation of compound collections, particularly of uncharacterized small molecules, and will be extended to further bioactivity clusters in future. The data is public accessible via https://github.com/mpimp-comas/2022_pahl_ziegler_subprofiles and the web app tool http://cpcse.pythonanywhere.com/.