Latent causal diffusions for single-cell perturbation modeling.
7 affiliations
- 1Department of Computer Science, Eidgenössische Technische Hochschule (ETH) Zürich, Zürich 8092, Switzerland.
- 2Laboratory for Information and Decision Systems, Massachusetts Institute of Technology (MIT), Cambridge, MA 02139.
- 3Eric and Wendy Schmidt Center, Broad Institute of MIT and Harvard, Cambridge, MA 02142.
- 4School of Computer and Communication Sciences, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne 1015, Switzerland.
- 5School of Life Sciences, Swiss Institute for Experimental Cancer Research, EPFL, Lausanne 1015, Switzerland.
- 6Max Planck Institute for Intelligent Systems, Tübingen 72076, Germany.
- 7European Laboratory for Learning and Intelligent Systems (ELLIS) Institute, Tübingen 72076, Germany.
Abstract
Perturbation screens hold the potential to systematically map regulatory processes at single-cell resolution, yet modeling and predicting transcriptome-wide responses to perturbations remains a major computational challenge. Existing methods often underperform simple baselines, fail to disentangle measurement noise from biological signal, and provide limited insight into the causal structure governing cellular responses. Here, we present the latent causal diffusion (LCD), a generative model that frames single-cell gene expression as a stationary diffusion process observed under measurement noise. LCD outperforms established approaches in predicting the distributional shifts of unseen perturbation combinations in single-cell RNA-sequencing screens while simultaneously learning a mechanistic dynamical system of gene regulation. To interpret these learned dynamics, we develop an approach we call causal linearization via perturbation responses (CLIPR), which yields an approximation of the direct causal effects between all genes modeled by the diffusion. CLIPR provably identifies causal effects under a linear drift assumption and recovers causal structure in both simulated systems and a genome-wide perturbation screen, where it clusters genes into coherent functional modules and resolves causal relationships that standard differential expression analysis cannot. The LCD-CLIPR framework bridges generative modeling with causal inference to predict unseen perturbation effects and map the underlying regulatory mechanisms of the transcriptome.
Related
- Dynamic graph prior-guided latent adaptation for generalizable single-cell perturbation response prediction.
- iGPSC-DB: a comprehensive database for an inferred single-cell genetic perturbation atlas
- A systematic comparison of single-cell perturbation response prediction models
- Conditional Monge Gap enables generalizable single-cell perturbation modelling
- scLAP: A Liver-Adapted Contrastive Shared-Latent Framework for Pseudo-Pairing Single-Cell Transcriptomes and Chromatin Accessibility Profiles
BibTeX
@article{lorch2026latent,
title = {{Latent causal diffusions for single-cell perturbation modeling.}},
author = {Lars Lorch and Jiaqi Zhang and Charlotte Bunne and Andreas Krause and Bernhard Schölkopf and Caroline Uhler},
journal = {Proceedings of the National Academy of Sciences of the United States of America},
year = {2026},
month = oct,
doi = {10.1073/pnas.2602630123},
url = {https://europepmc.org/article/MED/42848419},
note = {CC BY},
}
RIS
TY - JOUR
TI - Latent causal diffusions for single-cell perturbation modeling.
AU - Lars Lorch
AU - Jiaqi Zhang
AU - Charlotte Bunne
AU - Andreas Krause
AU - Bernhard Schölkopf
AU - Caroline Uhler
PY - 2026
DA - 2026/10/08
JO - Proceedings of the National Academy of Sciences of the United States of America
DO - 10.1073/pnas.2602630123
UR - https://europepmc.org/article/MED/42848419
AB - Perturbation screens hold the potential to systematically map regulatory processes at single-cell resolution, yet modeling and predicting transcriptome-wide responses to perturbations remains a major computational challenge. Existing methods often underperform simple baselines, fail to disentangle measurement noise from biological signal, and provide limited insight into the causal structure governing cellular responses. Here, we present the latent causal diffusion (LCD), a generative model that frames single-cell gene expression as a stationary diffusion process observed under measurement noise. LCD outperforms established approaches in predicting the distributional shifts of unseen perturbation combinations in single-cell RNA-sequencing screens while simultaneously learning a mechanistic dynamical system of gene regulation. To interpret these learned dynamics, we develop an approach we call causal linearization via perturbation responses (CLIPR), which yields an approximation of the direct causal effects between all genes modeled by the diffusion. CLIPR provably identifies causal effects under a linear drift assumption and recovers causal structure in both simulated systems and a genome-wide perturbation screen, where it clusters genes into coherent functional modules and resolves causal relationships that standard differential expression analysis cannot. The LCD-CLIPR framework bridges generative modeling with causal inference to predict unseen perturbation effects and map the underlying regulatory mechanisms of the transcriptome.
SN - 0027-8424
SN - 1091-6490
ER -