Usage Principles

Input

  • data_name: Name of the dataset.

  • data_dir: Path of an h5ad file where the scCAS data count matrix is stored in the compressed sparse row format. AnnData object of shape n_obs × n_vars. Rows correspond to cells and columns to peaks/regions.

  • if_bi: Whether to binarize the scCAS data count matrix.

  • if_mc_bi: Whether to binarize the metacell-by-region matrix.

  • threshold: Threshold for binarizing metacell-by-region matrix.

  • filter_rate: Proportion for feature selection.

  • chunk_size: Number of cells in each chunk.

  • carousel_resolution: Ratio of the number of cells to that of metacells.

  • base: Export path for EpiCarousel.

  • step: Length of Walktrap community detection.

  • threads: Number of parallel processes.

  • mc_mode: Mode of calculating metacell-by-region matrix.

  • index: (Optional) Ground truth cell type label of single cells for downstream analysis and evaluation.

Output

  • adata: Metacell AnnData object of shape n_obs × n_vars stored in an h5ad file. Rows correspond to metacells and columns to features.

EpiCarousel can also be seamlessly integrated with EpiScanpy, a widely-used Python library for epigenomics single cell analysis:

import episcanpy.api as epi
import epicarousel

# Run EpiCarousel
carousel = epicarousel.core.Carousel(data_name,
                                    data_dir,
                                    if_bi,
                                    if_mc_bi,
                                    threshold,
                                    filter_rate,
                                    chunk_size,
                                    carousel_resolution,
                                    base,
                                    step,
                                    threads,
                                    mc_mode,
                                    index
                                    )
carousel.make_dirs()
carousel.data_split()
carousel.identify_metacells()
carousel.merge_metacells()
carousel.metacell_preprocess()
carousel.metacell_data_clustering()
carousel.result_comparison()
carousel.delete_dirs()

# Load the metacell data as an AnnData object (adata).
carousel.mc_adata