A multimodal perturbation atlas defines the phenotypic resolution of cellular morphology.

Microscopy is uniquely suited to this goal: it is non-destructive and can be applied to living cells in their native context. Yet its phenotypic resolving power remains incompletely characterized relative to molecular assays.

Here, we present a multimodal perturbation atlas of 1,000 pooled CRISPR knockouts in A549 cells, profiled by fluorescence microscopy (42 live, 13 fixed markers), label-free quantitative phase imaging of the same live cells (at single timepoints), and single-cell RNA sequencing (scRNA-seq). We develop deep learning frameworks to interpret the rich cell-biological signatures in these ~65M single-cell profiles. At matched reagent cost, phase imaging exceeds the phenotypic resolution of both fluorescence imaging and scRNA-seq, and more reliably recovers higher-order pathway organization.