LLMs-Universal-Life-Science-and-Clinical-Skills- scGPT_Agent

skill: id: biomedical.foundation_models.scgpt_agent version: 1.0.0 name: scGPT category: foundation_models/scgpt_agent description: short: Agent for scGPT single-cell foundation model operations. long: "Agent for scGPT single-cell foundation model operations.\n\nProvides cell
\ type annotation, perturbation prediction, and batch\nintegration using the
\ scGPT transformer model trained on 33M+ cells.\n\nExample:\n >>> agent
\ = scGPTAgent(model="scgpt-whole-human")\n >>> result = agent.annotate_cell_types(adata)\n
\ >>> print(f"Found {len(result.cell_types)} cell types")" use_cases: - Auto-generated from source code capabilities:

• name: load_model description: "Load the scGPT model.\n\nReturns:\n True if successful, False
  \ otherwise" inputs: [] outputs:
  • name: result type: bool description: Result of the operation
• name: annotate_cell_types description: "Annotate cell types in single-cell data.\n\nArgs:\n adata: AnnData
  \ object with scRNA-seq data\n config: Annotation configuration\n\nReturns:\n
  \ AnnotationResult with cell type predictions" inputs:
  • name: adata type: Any description: Input parameter adata required: true
  • name: config type: Optional[AnnotationConfig] description: Input parameter config required: true outputs:
  • name: result type: AnnotationResult description: Result of the operation
• name: predict_perturbation description: "Predict cellular response to gene perturbation.\n\nArgs:\n adata:
  \ AnnData object with baseline expression\n config: Perturbation configuration\n
  \nReturns:\n PerturbationResult with predicted changes" inputs:
  • name: adata type: Any description: Input parameter adata required: true
  • name: config type: Optional[PerturbationConfig] description: Input parameter config required: true outputs:
  • name: result type: PerturbationResult description: Result of the operation
• name: integrate_batches description: "Integrate multiple single-cell datasets.\n\nArgs:\n adata_list:
  \ List of AnnData objects to integrate\n config: Integration configuration\n
  \nReturns:\n Integrated AnnData object (or mock result)" inputs:
  • name: adata_list type: List[Any] description: Input parameter adata_list required: true
  • name: config type: Optional[IntegrationConfig] description: Input parameter config required: true outputs:
  • name: result type: Any description: Result of the operation
• name: get_cell_embeddings description: "Extract cell embeddings from scGPT.\n\nArgs:\n adata: AnnData
  \ object\n layer: Which transformer layer to use ("last", "mean", or
  \ layer number)\n\nReturns:\n Cell embedding matrix (n_cells x n_dims)" inputs:
  • name: adata type: Any description: Input parameter adata required: true
  • name: layer type: str description: Input parameter layer required: true outputs:
  • name: result type: Any description: Result of the operation
• name: gene_expression_prediction description: "Predict expression of masked genes.\n\nArgs:\n adata: AnnData
  \ object\n genes_to_predict: List of gene names to predict\n use_neighbors:
  \ Whether to use cell neighbors for prediction\n\nReturns:\n Dictionary mapping
  \ gene names to predicted expression arrays" inputs:
  • name: adata type: Any description: Input parameter adata required: true
  • name: genes_to_predict type: List[str] description: Input parameter genes_to_predict required: true
  • name: use_neighbors type: bool description: Input parameter use_neighbors required: true outputs:
  • name: result type: Dict[Any] description: Result of the operation
• name: get_model_info description: Get information about the loaded model. inputs: [] outputs:
  • name: result type: Dict[Any] description: Result of the operation metadata: created: '2026-01-12' author: BioKernel Auto-Generator