
Resources: Posters
Paired Single-Cell Imaging and Transcriptomics Provide a Functional Alignment Layer for Virtual Cell Models
May 30, 2026
Cold Spring Harbor Laboratory - AI in Biology
Keywords: Virtual cell models, DINOv2, morphology-transcriptomics integration, phagocytosis, cellular senescence

This research introduces a functional alignment layer that pairs longitudinal live-cell imaging with endpoint transcriptomics to train virtual cell foundation models on cellular function, not just gene expression. Utilizing the R3200 Platform and CellCage technology alongside DINOv2 morphological embeddings, the study identifies genes driving microglial phagocytosis and preadipocyte senescence that are invisible to unsupervised transcriptomic clustering alone. By anchoring supervised models on same-cell imaging readouts, the work establishes a new framework for building function-predictive foundation models in biology.

