Predicting cellular responses to perturbation across diverse contexts with State
9 affiliations
- 1Arc Institute, Palo Alto, CA, USA
- 2Arc Institute, Palo Alto, CA, USA; University of California, Berkeley, Berkeley, CA, USA
- 3Arc Institute, Palo Alto, CA, USA; University of Pennsylvania, Philadelphia, PA, USA
- 4Arc Institute, Palo Alto, CA, USA; University of California, San Francisco, San Francisco, CA, USA
- 5Arc Institute, Palo Alto, CA, USA; Stanford University, Stanford, CA, USA
- 6Arc Institute, Palo Alto, CA, USA; Yale University, New Haven, CT, USA
- 7Myllia Biotechnology GmbH, Vienna, Austria
- 8Stanford University, Stanford, CA, USA
- 9Arc Institute, Palo Alto, CA, USA. Electronic address: [email protected]
Abstract
While machine learning models offer potential for predicting transcriptomic effects of perturbation, they currently struggle to generalize across cellular contexts. Here, we introduce State, a machine learning model that predicts perturbation effects while accounting for cellular heterogeneity within and across experiments. State is trained using single-cell gene expression data to predict perturbation effects across sets of cells. State improved discrimination of effects on large datasets by more than 30% and identified differentially expressed genes across genetic, signaling, and chemical perturbations with significantly improved accuracy compared with baselines. Its cell embeddings trained on observational data from 167 million cells enable the identification of strong perturbations in cellular contexts where no perturbations were observed during training. We further introduce Cell-Eval, a comprehensive evaluation framework that can be used to evaluate future models. Overall, the performance and flexibility of State set the stage for scaling the development of AI models of cell state.
Related
- Decoding common and rare noncoding variant effects across cellular and developmental contexts
- Decoding sequence determinants of gene expression in diverse cellular and disease states
- scDRP: disentangled representation learning for predicting single-cell responses to perturbations and estimating individual treatment effects
- Temporal Integration of Weak Acoustic Stress Leads to Delayed Cellular Responses
- Divergent transcriptional responses of conserved cellular machinery to warming and phosphorus limitation in freshwater phytoplankton
BibTeX
@article{adduri2026predicting,
title = {{Predicting cellular responses to perturbation across diverse contexts with State}},
author = {Abhinav Adduri and Dhruv Gautam and Beatrice Bevilacqua and Mohsen Naghipourfar and Alishba Imran and Rohan Shah and Noam Teyssier and Rishi Verma and Christopher Carpenter and Basak Eraslan and Francis Chalissery and Rajesh Ilango and Vishvak Subramanyam and Chiara Ricci-Tam and Sanjay Nagaraj and Aidan F. Winters and Mingze Dong and Stefanie Fellinger and Adam Krejčí and Tilmann Bürckstümmer and Sravya Tirukkovular and Jeremy Sullivan and Brian S. Plosky and Nicholas D. Youngblut and Jure Leskovec and Luke A. Gilbert and Silvana Konermann and Patrick D. Hsu and Alexander Dobin and Dave P. Burke and Hani Goodarzi and Yusuf Roohani},
journal = {Cell},
year = {2026},
month = aug,
doi = {10.1016/j.cell.2026.07.052},
url = {https://doi.org/10.1016/j.cell.2026.07.052},
note = {CC BY},
}
RIS
TY - JOUR
TI - Predicting cellular responses to perturbation across diverse contexts with State
AU - Abhinav Adduri
AU - Dhruv Gautam
AU - Beatrice Bevilacqua
AU - Mohsen Naghipourfar
AU - Alishba Imran
AU - Rohan Shah
AU - Noam Teyssier
AU - Rishi Verma
AU - Christopher Carpenter
AU - Basak Eraslan
AU - Francis Chalissery
AU - Rajesh Ilango
AU - Vishvak Subramanyam
AU - Chiara Ricci-Tam
AU - Sanjay Nagaraj
AU - Aidan F. Winters
AU - Mingze Dong
AU - Stefanie Fellinger
AU - Adam Krejčí
AU - Tilmann Bürckstümmer
AU - Sravya Tirukkovular
AU - Jeremy Sullivan
AU - Brian S. Plosky
AU - Nicholas D. Youngblut
AU - Jure Leskovec
AU - Luke A. Gilbert
AU - Silvana Konermann
AU - Patrick D. Hsu
AU - Alexander Dobin
AU - Dave P. Burke
AU - Hani Goodarzi
AU - Yusuf Roohani
PY - 2026
DA - 2026/08/31
JO - Cell
DO - 10.1016/j.cell.2026.07.052
UR - https://doi.org/10.1016/j.cell.2026.07.052
AB - While machine learning models offer potential for predicting transcriptomic effects of perturbation, they currently struggle to generalize across cellular contexts. Here, we introduce State, a machine learning model that predicts perturbation effects while accounting for cellular heterogeneity within and across experiments. State is trained using single-cell gene expression data to predict perturbation effects across sets of cells. State improved discrimination of effects on large datasets by more than 30% and identified differentially expressed genes across genetic, signaling, and chemical perturbations with significantly improved accuracy compared with baselines. Its cell embeddings trained on observational data from 167 million cells enable the identification of strong perturbations in cellular contexts where no perturbations were observed during training. We further introduce Cell-Eval, a comprehensive evaluation framework that can be used to evaluate future models. Overall, the performance and flexibility of State set the stage for scaling the development of AI models of cell state.
SN - 0092-8674
SN - 1097-4172
ER -