IRIS solves a long-standing gap: #imaging and droplet-based sequencing were never truly connected at single-cell resolution. Here, every cell is imaged first (BF + 4 fluorescent channels) → then deterministically barcoded → then sequenced, enabling single cell #phenomics.

As a 1st application, we used IRIS to profile >5k FUCCI-3T3 cells, reconstructing the full continuous #cell-cycle from morphology + RNA; identifying 670 cycling genes. IRIS’s morphology-anchored cell-cycle angle aligns with Seurat/Tricycle but provides smoother, higher-resolution structure.

We uncovered phase-specific TF activity, revealing how DREAM complex repression, FUCCI intensity, and cell-cycle speed are linked. IRIS detects quiescence-primed vs fully quiescent states & shows that slowly cycling cells display stronger DREAM-mediated repression, insights missed by RNA-only tools.

We also show that IRIS enables prediction of transcriptomes directly from images (#ML). Models trained on IRIS data recover gene-level variation, cell-cycle phase, and cell identity from morphology alone.

IRIS also revealed that subtle nuclear morphologies correspond to distinct molecular states, including the previously puzzling T cell stripy nuclear phenotype (collab. w/ @BerendSnijder's lab; Hale et al., Science, 2024), now shown to map to a specific transcriptomic program.