CELLama: Foundation Model for Single Cell and Spatial Transcriptomics by Cell Embedding Leveraging Language Model Abilities

CELLama is created, a framework that harnesses language models to convert cellular data into “sentences” that represent gene expression and metadata, enabling a universal embedding of cells. Unlike most single‐cell foundation models, CELLama supports scalable analysis and offers flexible applications including spatial transcriptomics. Its capacity to automate cell typing with atlas‐scale datasets underscores its practical value in simplifying complex workflows. Abstract Large‐scale single‐cell RNA sequencing (scRNA‐seq) and spatial transcriptomics (ST) have transformed biomedical research into a data‐driven field, enabling the creation of comprehensive atlases. These methodologies facilitate detailed understanding of biology and pathophysiology; however, the complexity and sheer volume of data present analytical challenges, particularly in robust cell typing, integration, and understanding complex spatial relationships of cells. To address these challenges, CELLama (Cell Embedding Leverage Language Model Abilities) develops a framework that leverage language model to transform cell data into “sentences” that encapsulate gene expressions and metadata, enabling universal cell embedding. CELLama, serving as a foundation model, supports flexible applications ranging from cell typing to analysis of spatial contexts, independent of complex dataset‐specific analysis workflows by using a large cell atlas. The results demonstrate that CELLama has significant potential to transform cellular analysis in various contexts, from determining cell types using multi‐tissue atlases and their interactions to unraveling intricate tissue dynamics. CELLama is created, a framework that harnesses language models to convert cellular data into “sentences” that represent gene expression and metadata, enabling a universal embedding of cells. Unlike most single-cell foundation models, CELLama supports scalable analysis and offers flexible applications including spatial transcriptomics. Its capacity to automate cell typing with atlas-scale datasets underscores its practical value in simplifying complex workflows. Abstract Large-scale single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics (ST) have transformed biomedical research into a data-driven field, enabling the creation of comprehensive atlases. These methodologies facilitate detailed understanding of biology and pathophysiology; however, the complexity and sheer volume of data present analytical challenges, particularly in robust cell typing, integration, and understanding complex spatial relationships of cells. To address these challenges, CELLama (Cell Embedding Leverage Language Model Abilities) develops a framework that leverage language model to transform cell data into “sentences” that encapsulate gene expressions and metadata, enabling universal cell embedding. CELLama, serving as a foundation model, supports flexible applications ranging from cell typing to analysis of spatial contexts, independent of complex dataset-specific analysis workflows by using a large cell atlas. The results demonstrate that CELLama has significant potential to transform cellular analysis in various contexts, from determining cell types using multi-tissue atlases and their interactions to unraveling intricate tissue dynamics. Advanced Science, EarlyView.