Disentangled generative modeling of single-cell perturbation responses across dynamic cellular states

PertVI disentangles baseline mitotic phase, cellular state, and batch effects from treatment-induced shifts in single-cell perturbation data, isolating off-manifold anomalies in orthogonal space.

• Why it matters: Standard models can obscure treatment effects when profiles mix cell division, lineage progression, donor variation, and stress, limiting mechanistic interpretation across multi-condition systems.
• What they did: The method uses a deep generative, disentangled variational inference framework to quantify treatment-driven changes in cell-state composition and infer state-gated transcriptional responses.
• The result: In simulated benchmarks and conditional chemical and genetic screens of primary human Homo sapiens CD8+ and CD4+ T cells, it separates memory-precursor adjuvants from toxic phenocopies and links phenotypic shifting to state-specific impulses rather than broad transcriptional perturbation responses.

AI summary of the abstract · may contain errors · not medical advice · read the paper.