In the Eagle module, we start by focusing on quality-controlled, annotated datasets that contain dimensionality-reduced single-cell transcriptomic data. First, we downsample the dataset to 10,000 cells based on cell type proportions. We then proceed with the standard workflow of the Scanpy toolkit, computing the top 3,000 highly variable genes for each dataset and applying principal component analysis (PCA) for dimensionality reduction with default parameters. We cluster the cells using a resolution of 1.0. Subsequently, we compiled 108 gene sets and calculated gene set scores for each dataset. The scoring approach involved identifying the intersection between the genes in the set and those in the dataset. If the intersection was non-empty, the average expression value of the intersecting genes in a given cell was used as that cell's score for the gene set. These scores were visualised using Scanpy’s sc.pl.umap and sc.pl.violin functions.