DRIFT: Disentangled Responsive-Invariant Flow Transport for Single-Cell Perturbation Prediction
DRIFT predicts single-cell perturbation responses by flow-matching only a disentangled responsive cell-state block.
DRIFT targets single-cell perturbation prediction, where unpaired sequencing and cell-to-cell variability can confound treatment effects. A variational encoder splits each cell into an invariant block and a responsive block using conditional priors and an information-theoretic invariance constraint. Conditional flow matching then transports only the responsive block, conditioned on the perturbation and the invariant state. The authors report that the method outperforms the strongest published baseline on combinatorial and unseen perturbation prediction benchmarks.
- Separates invariant and responsive cell-state components before transport.
- Flow matching acts only on the responsive block, avoiding prescribed latent shifts.
- Conditional priors and an invariance constraint enforce the disentanglement.
- Outperforms prior methods on combinatorial and unseen perturbation benchmarks.
Full article191 words · extracted from arxiv.org · click to collapse
Predicting cellular responses to perturbations is a central problem in cellular biology, with broad applications in systems biology and drug discovery. This task is challenging because cellular responses can be complex and cell-state dependent, intrinsic cell-to-cell variability can be confounded with perturbation effects, and destructive single-cell RNA sequencing precludes paired measurements of the same cell before and after treatment. Flow matching transports control cells to perturbed states flexibly, but acting on the full cell state can confound perturbation effects with pre-existing cell-to-cell variability. Disentangled approaches separate responsive from invariant components, but model perturbations through prescribed mechanisms, such as latent shifts or graph edits, limiting their flexibility. We address both limitations in a unified framework. A variational encoder disentangles each cell into an invariant block, capturing state unaffected by the perturbation, and a responsive block, capturing state it changes, through conditional priors and an information-theoretic invariance constraint. Conditional flow matching transports only the responsive block, conditioned on the perturbation and invariant state, yielding a flexible, data-driven model of perturbation effects without confounding pre-existing variability. Across several benchmarks, our method outperforms the strongest published method in settings involving combinatorial and unseen perturbation prediction.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.35106