Ramon Pfaendler | Scientist
Botnar Institute of Immune Engineering

Festival of Biologics Day 3 @ 09:20

Self-supervised deep learning for unbiased morphological phenotyping across cell types

Characterising cellular phenotypic heterogeneity is central to understanding how morphological and molecular states emerge at the single-cell level. High-content screening provides a data-rich window into this heterogeneity, but extracting biologically meaningful phenotypes from complex imaging data remains challenging. Self-supervised learning offers an unbiased strategy to learn representations directly from unlabelled images, enabling phenotype discovery without predefined annotations.

We present scDINO, a self-supervised vision transformer framework adapted to multi-channel high-content microscopy, and its successor scDINOv2, which provides a unified framework for benchmarking advances in DINO-based representation learning. Across more than 1.3 million primary human immune cells spanning seven cell types, scDINOv2 learns representations without access to cell-type labels that substantially outperform conventional image-analysis approaches, while capturing biologically meaningful features of cellular morphology and lineage-marker expression and enabling efficient, scalable inference. Applied to primary human T cells, scDINO enabled the unsupervised de novo discovery of a previously unrecognised T cell phenotype. Furthermore, in a high-content perturbation screen in human induced pluripotent stem cells, scDINO identified discrete morphotypes embedded within continuous morphological trajectories associated with pluripotency exit and cellular fate commitment.

Together, these results establish self-supervised imaging as a powerful framework for scalable, unbiased representation learning and de novo phenotype discovery in high-content cellular screens.

Ramon Pfaendler, Scientist, Botnar Institute of Immune Engineering

×

Ramon Pfaendler / Self-supervised deep learning for unbiased morphological phenotyping across cell types

Ramon Pfaendler, Scientist,
Botnar Institute of Immune Engineering

Ramon grew up in Steinach at the Lake of Constance and later moved to Zurich, where he obtained his MSc in Systems Biology from the University of Zurich before completing his PhD in Berend Snijder’s group at ETH Zurich. A quantitative stem cell biologist at heart, his PhD research focused on...

CloseMore