scMaize: A Single-Cell Foundation Model and Integrated Atlas for Maize

Single-cell transcriptomics has resolved cell type-specific gene expression in plants, yet maize still lacks an integrated reference and species-specific foundation models. We present scMaize, which combines scMaizeAtlas, an integrated atlas of 385,675 cells from 20 projects and 66 samples across seven tissues with hierarchical annotation, with two Transformer-based foundation models pretrained on this atlas. scMaizeExp serves as an expression-only baseline, whereas scMaizeGO incorporates Gene Ontology (GO) functional embeddings as an inductive bias. Although the two models showed comparable global expression prediction accuracy, the GO prior improved rank-order prediction, strengthened attention toward functionally coherent gene modules, and enhanced embedding organization. scMaizeGO achieved 86.0% cell type classification accuracy and 97.1% tissue classification accuracy. Evaluation on independent maize, rice, and Arabidopsis datasets demonstrated the transferability of scMaizeGO representations, while few-shot fine-tuning enabled accurate cross-species classification using a limited fraction of labeled cells. Perturbation analysis further showed that the model captured treatment-associated cellular states, and expression projection identified condition-responsive genes enriched in established stress pathways. An online platform (https://www.scmaize.com) provides atlas exploration, model access, and zero-code analysis tools. Together, scMaize provides an integrated resource and computational framework for transferable and perturbation-aware representation learning in crop single-cell genomics. HIGHLIGHTSO_LIscMaizeAtlas integrates 385,675 cells from 20 maize single-cell projects. C_LIO_LIscMaizeGO incorporates Gene Ontology priors into maize-specific pretraining. C_LIO_LIGO priors improve rank-order prediction, attention coherence and embeddings. C_LIO_LIFew-shot tuning enables cross-species cell-type classification with limited labels. C_LIO_LIExpression projection reveals stress-responsive genes in root cell states. C_LI