SAVE: A Generalizable Framework for Multi-Condition Single-Cell Generation with Gene Block Attention

arXiv cs.AI / 4/21/2026

📰 NewsTools & Practical UsageModels & Research

共有:

Key Points

SAVE introduces a generalizable generative framework for multi-condition single-cell gene expression using conditional Transformers.
Instead of treating genes as independent tokens, SAVE groups semantically related genes into blocks and uses Gene Block Attention to capture higher-order dependencies among gene modules.
A Flow Matching mechanism and a condition-masking strategy improve flexible simulation and enable generalization to unseen combinations of biological and technical conditions.
Across multiple benchmarks—conditional generation, batch effect correction, and perturbation prediction—SAVE achieves better generation fidelity and extrapolative generalization than state-of-the-art methods, particularly in low-resource and combinatorially held-out scenarios.
The authors provide public code via GitHub, supporting reproducibility and broader adoption for virtual cell synthesis and biological interpretation.

Abstract

Modeling single-cell gene expression across diverse biological and technical conditions is crucial for characterizing cellular states and simulating unseen scenarios. Existing methods often treat genes as independent tokens, overlooking their high-level biological relationships and leading to poor performance. We introduce SAVE, a unified generative framework based on conditional Transformers for multi-condition single-cell modeling. SAVE leverages a coarse-grained representation by grouping semantically related genes into blocks, capturing higher-order dependencies among gene modules. A Flow Matching mechanism and condition-masking strategy further enhance flexible simulation and enable generalization to unseen condition combinations. We evaluate SAVE on a range of benchmarks, including conditional generation, batch effect correction, and perturbation prediction. SAVE consistently outperforms state-of-the-art methods in generation fidelity and extrapolative generalization, especially in low-resource or combinatorially held-out settings. Overall, SAVE offers a scalable and generalizable solution for modeling complex single-cell data, with broad utility in virtual cell synthesis and biological interpretation. Our code is publicly available at https://github.com/fdu-wangfeilab/sc-save