Learning single-cell perturbation responses using neural optimal transport.

Autor: Bunne C; Department of Computer Science, ETH Zurich, Zürich, Switzerland.; AI Center, ETH Zurich, Zürich, Switzerland., Stark SG; Department of Computer Science, ETH Zurich, Zürich, Switzerland.; AI Center, ETH Zurich, Zürich, Switzerland.; Medical Informatics Unit, University of Zurich Hospital, Zürich, Switzerland.; Swiss Institute of Bioinformatics, Zurich, Switzerland., Gut G; Department of Molecular Life Sciences, University of Zurich, Zürich, Switzerland., Del Castillo JS; Department of Molecular Life Sciences, University of Zurich, Zürich, Switzerland., Levesque M; Department of Dermatology, University of Zurich Hospital, University of Zurich, Zürich, Switzerland., Lehmann KV; Department of Computer Science, ETH Zurich, Zürich, Switzerland. kjlehmann@ukaachen.de.; Cancer Research Center Cologne-Essen, Site: Center Integrated Oncology Aachen, Aachen, Germany. kjlehmann@ukaachen.de., Pelkmans L; Department of Molecular Life Sciences, University of Zurich, Zürich, Switzerland. lucas.pelkmans@mls.uzh.ch., Krause A; Department of Computer Science, ETH Zurich, Zürich, Switzerland. krausea@ethz.ch.; AI Center, ETH Zurich, Zürich, Switzerland. krausea@ethz.ch., Rätsch G; Department of Computer Science, ETH Zurich, Zürich, Switzerland. gunnar.raetsch@inf.ethz.ch.; AI Center, ETH Zurich, Zürich, Switzerland. gunnar.raetsch@inf.ethz.ch.; Medical Informatics Unit, University of Zurich Hospital, Zürich, Switzerland. gunnar.raetsch@inf.ethz.ch.; Swiss Institute of Bioinformatics, Zurich, Switzerland. gunnar.raetsch@inf.ethz.ch.; Department of Biology, ETH Zurich, Zürich, Switzerland. gunnar.raetsch@inf.ethz.ch.
Jazyk: angličtina
Zdroj: Nature methods [Nat Methods] 2023 Nov; Vol. 20 (11), pp. 1759-1768. Date of Electronic Publication: 2023 Sep 28.
DOI: 10.1038/s41592-023-01969-x
Abstrakt: Understanding and predicting molecular responses in single cells upon chemical, genetic or mechanical perturbations is a core question in biology. Obtaining single-cell measurements typically requires the cells to be destroyed. This makes learning heterogeneous perturbation responses challenging as we only observe unpaired distributions of perturbed or non-perturbed cells. Here we leverage the theory of optimal transport and the recent advent of input convex neural architectures to present CellOT, a framework for learning the response of individual cells to a given perturbation by mapping these unpaired distributions. CellOT outperforms current methods at predicting single-cell drug responses, as profiled by scRNA-seq and a multiplexed protein-imaging technology. Further, we illustrate that CellOT generalizes well on unseen settings by (1) predicting the scRNA-seq responses of holdout patients with lupus exposed to interferon-β and patients with glioblastoma to panobinostat; (2) inferring lipopolysaccharide responses across different species; and (3) modeling the hematopoietic developmental trajectories of different subpopulations.
(© 2023. The Author(s).)
Databáze: MEDLINE
Externí odkaz: