Dynamic graph prior-guided latent adaptation for generalizable single-cell perturbation response prediction.
- 1School of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, China.
- 2Zhongguancun Academy, Beijing 100094, China.
Abstract
Motivation
Predicting transcriptional responses to drug perturbations in unseen cellular contexts is a central challenge in single-cell perturbation modeling. This challenge creates a need for models that combine flexible state transitions with structured biological prior knowledge. Existing data-driven models often struggle to generalize across unseen cell types, while many prior-based models rely on static biological graphs without explicitly conditioning prior influence on perturbation-specific target support.Results
We propose DGP-LA, a dynamic graph-prior latent adaptation framework that refines biological priors and injects them as gated latent corrections within a cell-state-conditioned transition model. DGP-LA couples a transition backbone with perturbation-conditioned sparse graph refinement. It propagates gene-aligned target support over the refined graph and selectively injects the resulting topology-aware signal into the hidden representation. On Tahoe and NeurIPS, DGP-LA improves Pearson Delta relative to State by 8.7% and 35.8% for generalization across unseen cell types. In zero-shot drug perturbation response prediction on unseen cell types, it improves Pearson Delta relative to CRISP by 79.1% and 67.6%. The model also improves DE-gene recovery, response discrimination, and Hallmark pathway recovery, supporting adaptive latent integration of biological priors for more generalizable and biologically interpretable perturbation modeling.Availability
DGP-LA's source code and reproducibility materials are available at https://github.com/bieyl/DGP-LA and archived on Zenodo at https://doi.org/10.5281/zenodo.22155398.
Related
- A systematic comparison of single-cell perturbation response prediction models
- Conditional Monge Gap enables generalizable single-cell perturbation modelling
- An open structure-guided RNA foundation model for generalizable prediction
- Dynamic perturbation analysis: a modular rewiring framework for decoding cancer stage progression and identifying therapeutic vulnerabilities.
- D-SPIN constructs regulatory network models from scRNA-seq that reveal organizing principles of perturbation response
BibTeX
@article{bie2026dynamic,
title = {{Dynamic graph prior-guided latent adaptation for generalizable single-cell perturbation response prediction.}},
author = {Yunlong Bie and Peishan Fan and Qijin Yin and Li Xiao},
journal = {Bioinformatics (Oxford, England)},
year = {2026},
month = oct,
doi = {10.1093/bioinformatics/btag747},
url = {https://europepmc.org/article/MED/42865067},
note = {CC BY},
}
RIS
TY - JOUR
TI - Dynamic graph prior-guided latent adaptation for generalizable single-cell perturbation response prediction.
AU - Yunlong Bie
AU - Peishan Fan
AU - Qijin Yin
AU - Li Xiao
PY - 2026
DA - 2026/10/01
JO - Bioinformatics (Oxford, England)
DO - 10.1093/bioinformatics/btag747
UR - https://europepmc.org/article/MED/42865067
AB - Motivation Predicting transcriptional responses to drug perturbations in unseen cellular contexts is a central challenge in single-cell perturbation modeling. This challenge creates a need for models that combine flexible state transitions with structured biological prior knowledge. Existing data-driven models often struggle to generalize across unseen cell types, while many prior-based models rely on static biological graphs without explicitly conditioning prior influence on perturbation-specific target support.Results We propose DGP-LA, a dynamic graph-prior latent adaptation framework that refines biological priors and injects them as gated latent corrections within a cell-state-conditioned transition model. DGP-LA couples a transition backbone with perturbation-conditioned sparse graph refinement. It propagates gene-aligned target support over the refined graph and selectively injects the resulting topology-aware signal into the hidden representation. On Tahoe and NeurIPS, DGP-LA improves Pearson Delta relative to State by 8.7% and 35.8% for generalization across unseen cell types. In zero-shot drug perturbation response prediction on unseen cell types, it improves Pearson Delta relative to CRISP by 79.1% and 67.6%. The model also improves DE-gene recovery, response discrimination, and Hallmark pathway recovery, supporting adaptive latent integration of biological priors for more generalizable and biologically interpretable perturbation modeling.Availability DGP-LA's source code and reproducibility materials are available at https://github.com/bieyl/DGP-LA and archived on Zenodo at https://doi.org/10.5281/zenodo.22155398.
SN - 1367-4803
SN - 1367-4811
ER -