Generalizable Morphological Profiling of Cells by Interpretable Unsupervised Learning

The intersection of advanced microscopy and machine learning is transforming the field of cell biology, enabling a more quantitative and data-driven approach. Traditional methods of morphological profiling, which rely on manual feature extraction, are often time-consuming, labor-intensive, and susceptible to human bias. Deep learning offers a promising alternative, but its effectiveness is hindered by its "black-box" operation and its dependence on extensive labeled data. MorphoGenie addresses these challenges by introducing an unsupervised deep-learning framework for single-cell morphological profiling. By integrating high-fidelity image reconstructions with disentangled representation learning, MorphoGenie facilitates discovery of a compact, interpretable latent space. This allows for the extraction of biologically meaningful features without human annotation, overcoming the "curse of dimensionality" inherent in manual methods. MorphoGenie stands out in three key attributes: High-fidelity Image Reconstruction: MorphoGenie utilizes a hybrid architecture that capitalizes on the unparalleled strengths of the variant of variational Autoencoders (VAEs) and generative Adversarial Networks (GANs) to achieve interpretable, high-quality cell image generation. Interpretability: MorphoGenie adopts a VAE-based method to learn a compact, interpretable, and transferable disentangled representation for single-cell morphological analysis. In contrast to the prior work, we propose a novel technique for interpreting the learned representation by extracting handcrafted features from reconstructed images produced by latent traversals, facilitating the discovery of biologically meaningful inferences, especially the heterogeneities of cell types and lineages. Generalizability: MorphoGenie is widely adaptable across various imaging modalities and experimental conditions, promoting cross-study comparisons and reusable morphological profiling results. The model generalizes to unseen single-cell datasets and different imaging modalities while providing explanations for its predictions. Overall, MorphoGenie could spearhead new strategies for conducting comprehensive morphological profiling and make biologically meaningful discoveries across a wide range of imaging modality.