arboreto

Gene regulatory network inference identifies transcription factor-target relationships from expression data. Arboreto uses GRNBoost2 and GENIE3 algorithms to produce a DataFrame with TF, target, and importance columns from your expression matrix, supporting both single-cell and bulk RNA-seq datasets with optional Dask-based scaling.

Both GRNBoost2 and GENIE3 are algorithms for inferring gene regulatory networks from expression data. Arboreto lets you compare both approaches on your dataset; GRNBoost2 typically offers faster computation while GENIE3 provides an alternative statistical framework. Your choice depends on dataset size and computational resources available.

Yes. Arboreto integrates Dask for distributed computing, allowing you to scale GRN inference across local or distributed clusters. This enables processing of large expression matrices without loading entire datasets into memory, critical for genome-scale single-cell RNA-seq analyses.

Arboreto requires expression data as a pandas DataFrame with observations (cells or samples) as rows and genes as columns. The Skill enforces this orientation strictly, requires a seed parameter for reproducibility, and accepts optional lists of transcription factors to restrict inference scope.

Arboreto is designed for use within pySCENIC pipelines to infer gene regulatory networks from expression data. It produces standardized TF-target-importance output compatible with downstream regulon analysis and cell-state annotation steps in single-cell RNA-seq workflows.

Before using arboreto, prepare your expression matrix as a pandas DataFrame with genes as columns and observations as rows, ensure genes are properly annotated, optionally filter lowly-expressed genes, and decide whether to provide a transcription factor list to focus inference. Set a random seed for reproducible results.