Abstract
Aberrations in 3D cell morphogenesis are linked to diseases such as cancer. Yet there is little systems-level understanding of cell shape determination in 3D, largely because there is a paucity of data-driven methods to quantify and describe 3D cell shapes. We have addressed this need using unsupervised geometric deep learning to learn shape representations of over 95,000 melanoma cells imaged by 3D high-throughput light-sheet microscopy. We used a dynamic graph convolutional foldingnet autoencoder with improved deep embedded clustering to simultaneously learn lower-dimensional representations and classes of 3D cell shapes. We describe a landscape of 3D cell morphology using deep learning-derived 3D quantitative morphological signatures (3DQMS) across different substrate geometries, following treatment by different clinically relevant small molecules and systematic gene depletion in high-throughput. By data integration, we predict modes of action for different small molecules providing mechanistic insights and blueprints for biological re-engineering. Finally, we provide explainability and interpretability for deep learning models.
Competing Interest Statement
C.D has a licensed granted patent on OPM.
Footnotes
matt.devries{at}icr.ac.uk
lucas.dent{at}icr.ac.uk
nathan.curry08{at}imperial.ac.uk
l.rowe-brown19{at}imperial.ac.uk
v.bousgouni{at}icr.ac.uk
adam{at}adamltyson.com
christopher.dunsby{at}imperial.ac.uk
chris.bakal{at}icr.ac.uk
Substantial additions and revisions to figures and the manuscript as a whole. Our most notable additions to our manuscript include a study where we have performed a comprehensive set of 168 gene depletions in ∼30,000 single cells. We undertook systematic siRNA depletion of most human RhoGTPases, and their regulators (RhoGEFs and RhoGAPs) and presented a high-throughput phenotypic screen using geometric deep learning of 3D cell shape.