sclet: A State-Aware Single-Cell Analysis Toolkit for SingleCellExperiment

sclet is a state-aware analysis toolkit built around SingleCellExperiment, the core Bioconductor container for single-cell data. Rather than treating SingleCellExperiment as a low-level backend hidden behind another framework, sclet takes it as the primary abstraction for modern single-cell and spatial transcriptomics workflows.

The package introduces an explicit Analysis-State Contract (Provenance DAG). In practice, this means the object records active layers, reductions, graphs, identities, and higher-level analysis outputs as part of a coherent workflow state. Upstream assumptions such as batch-corrected layers, dimensional reductions, and derived analysis results are therefore inspectable and reusable instead of being scattered across ad hoc slots or implicit conventions.

In addition to core steps (preprocessing, dimensionality reduction, clustering, annotation, and visualization), sclet provides a broad workflow surface for both single-cell and spatial transcriptomics, with selected Python tools managed through the basilisk sandbox:

• Core Analysis: State-aware preprocessing, dimensionality reduction, clustering, annotation, and provenance-aware workflow management.
• Integration: Multi-backend support (fastMNN, Harmony, scVI).
• Python Interoperability: Safe and isolated execution of scVelo (RNA Velocity), pySCENIC (Gene Regulatory Networks), cellrank (Advanced Fate Mapping), cell2location (Spatial Deconvolution), and celloracle (In-Silico Perturbation).
• Downstream Extensions: Gene set scoring (UCell, AUCell, GSVA), pseudobulk DE, enrichment analysis, cell-cell communication, and Milo differential abundance.
• AI Copilot: An intelligent diagnostic agent (sclet_copilot) powered by the aisdk framework that reads the object’s provenance to perform rigorous cross-chain error control.

For user-facing analysis verbs, sclet uses the Run* naming style throughout, such as RunIntegration(), RunSCENIC(), and RunVelocity(). This is now the public analysis API of the package.

✍️ Authors

Guangchuang YU

School of Basic Medical Sciences, Southern Medical University

https://yulab-smu.top

⏬ Installation

Get the development version from github:

## install.packages("remotes")
remotes::install_github("YuLab-SMU/sclet")

📖 Vignette

For more details, please refer to the online documents:

• vignette