scCBGM: Single-Cell Editing via Concept Bottlenecks

Alma Andersson, Aya Abdelsalam Ismail, Edward De Brouwer, Doron Haviv, Tommaso Biancalani, Kyunghyun Cho, Gabriele Scalia, Aicha Bentaieb, Hector Corrada Bravo
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:2716-2754, 2026.

Abstract

Understanding cellular phenotypes and how they respond to perturbations is critical for disease biology and therapeutic design. Single-cell RNA sequencing enables characterization at cellular resolution, yet the combinatorial space of conditions makes exhaustive experimental mapping infeasible. We introduce single-cell Concept Bottleneck Generative Models (scCBGM), a framework for interpretable and precise counterfactual editing of individual cells. scCBGM adapts concept bottleneck architectures for single-cell data through decoder skip connections and a cross-covariance penalty that promotes disentanglement without dimensional constraints. We extend the framework to flow matching models, enabling concept-guided editing in both encoding-decoding and generation regimes. To enable rigorous evaluation, we develop a synthetic benchmark with ground-truth counterfactuals. Across multiple real datasets, scCBGM demonstrates superior performance in combinatorial generalization and counterfactual prediction, supported by cell-level validation on synthetic data and population-level benchmarks on real datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-andersson26a, title = {sc{CBGM}: Single-Cell Editing via Concept Bottlenecks}, author = {Andersson, Alma and Ismail, Aya Abdelsalam and De Brouwer, Edward and Haviv, Doron and Biancalani, Tommaso and Cho, Kyunghyun and Scalia, Gabriele and Bentaieb, Aicha and Corrada Bravo, Hector}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {2716--2754}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/andersson26a/andersson26a.pdf}, url = {https://proceedings.mlr.press/v306/andersson26a.html}, abstract = {Understanding cellular phenotypes and how they respond to perturbations is critical for disease biology and therapeutic design. Single-cell RNA sequencing enables characterization at cellular resolution, yet the combinatorial space of conditions makes exhaustive experimental mapping infeasible. We introduce single-cell Concept Bottleneck Generative Models (scCBGM), a framework for interpretable and precise counterfactual editing of individual cells. scCBGM adapts concept bottleneck architectures for single-cell data through decoder skip connections and a cross-covariance penalty that promotes disentanglement without dimensional constraints. We extend the framework to flow matching models, enabling concept-guided editing in both encoding-decoding and generation regimes. To enable rigorous evaluation, we develop a synthetic benchmark with ground-truth counterfactuals. Across multiple real datasets, scCBGM demonstrates superior performance in combinatorial generalization and counterfactual prediction, supported by cell-level validation on synthetic data and population-level benchmarks on real datasets.} }
Endnote
%0 Conference Paper %T scCBGM: Single-Cell Editing via Concept Bottlenecks %A Alma Andersson %A Aya Abdelsalam Ismail %A Edward De Brouwer %A Doron Haviv %A Tommaso Biancalani %A Kyunghyun Cho %A Gabriele Scalia %A Aicha Bentaieb %A Hector Corrada Bravo %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-andersson26a %I PMLR %P 2716--2754 %U https://proceedings.mlr.press/v306/andersson26a.html %V 306 %X Understanding cellular phenotypes and how they respond to perturbations is critical for disease biology and therapeutic design. Single-cell RNA sequencing enables characterization at cellular resolution, yet the combinatorial space of conditions makes exhaustive experimental mapping infeasible. We introduce single-cell Concept Bottleneck Generative Models (scCBGM), a framework for interpretable and precise counterfactual editing of individual cells. scCBGM adapts concept bottleneck architectures for single-cell data through decoder skip connections and a cross-covariance penalty that promotes disentanglement without dimensional constraints. We extend the framework to flow matching models, enabling concept-guided editing in both encoding-decoding and generation regimes. To enable rigorous evaluation, we develop a synthetic benchmark with ground-truth counterfactuals. Across multiple real datasets, scCBGM demonstrates superior performance in combinatorial generalization and counterfactual prediction, supported by cell-level validation on synthetic data and population-level benchmarks on real datasets.
APA
Andersson, A., Ismail, A.A., De Brouwer, E., Haviv, D., Biancalani, T., Cho, K., Scalia, G., Bentaieb, A. & Corrada Bravo, H.. (2026). scCBGM: Single-Cell Editing via Concept Bottlenecks. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:2716-2754 Available from https://proceedings.mlr.press/v306/andersson26a.html.

Related Material