repository·master·Indexed 41 days ago
https://github.com/colomemaria/episcanpyA Python-based toolkit for analyzing single-cell epigenomic data, such as scATAC-seq and scBS-seq, extending the Scanpy framework. It provides modules for building count matrices (epi.ct), preprocessing and quality control (epi.pp), downstream analysis and clustering (epi.tl), and visualization (epi.pl). The toolkit utilizes anndata.AnnData as its primary data structure for storing methylation and ATAC-seq matrices.
EpiScanpy is a Python toolkit designed for the analysis of epigenomic single-cell data. It specifically supports:
EpiScanpy is built as an epigenomic extension of Scanpy, the widely used tool for single-cell RNA-seq analysis. Users familiar with Scanpy will find the workflows and data structures conceptually similar.
anndata.AnnData object as its primary data structure for storing methylation and ATAC-seq matrices. This structure stores observations (n_obs, e.g., cells) and variables (n_vars, e.g., expression, methylation, or chromatin features).If you want to install the latest version directly from the source code, use pip to install from the GitHub repository.
pip install git+https://github.com/colomemaria/epiScanpyThe episcanpy.api.tl (Tools) module contains functions for downstream analysis, clustering, and dimensionality reduction.
tl.tsne: t-SNE embedding.tl.umap: UMAP embedding.tl.diffmap: Diffusion maps.tl.pca: Principal Component Analysis.tl.louvain: Louvain community detection.tl.leiden: Leiden clustering.tl.kmeans: K-means clustering.tl.hc: Hierarchical clustering.tl.getNClusters: Determine number of clusters.tl.rank_features: Rank features.tl.top_feature_genes: Identify top feature genes.tl.var_features_to_genes: Map variable features to genes.tl.geneactivity: Calculate gene activity.tl.find_genes_in_features: Find genes within specific features.tl.load_markers: Load marker genes.tl.identify_cluster: Identify cluster markers.tl.dpt: Diffusion Pseudotime.tl.draw_graph: Visualize graphs.tl.diffmap: Diffusion maps.tl.ARI: Adjusted Rand Index.tl.AMI: Adjusted Mutual Information.tl.homogeneity: Homogeneity score.tl.silhouette: Silhouette coefficient.You can install the latest stable version of episcanpy via pip, or install the latest development version directly from the GitHub repository if you need specific updates.
# to install the latest stable version of episcanpy
pip install episcanpy
# to install the latest development version of episcanpy
pip install git+https://github.com/colomemaria/epiScanpy@update_load_featuresAll preprocessing and quality control functions are located in the epi.pp module. The input for these functions is an anndata.AnnData object.
Use these functions to understand feature distribution and coverage:
epi.pp.commoness_features(adata, **processing_params)epi.pp.coverage_cells(adata, **processing_params)To remove low-quality data:
epi.pp.filter_cells(adata, min_features=10)epi.pp.filter_features(adata, min_cells=10)To reduce the feature space to the most variable features:
epi.pl.cal_var(adata)epi.pp.select_var_feature(adata, max_score=0.2, nb_features=50000)# Quality control examples
epi.pp.commoness_features(adata, **processing_params)
epi.pp.coverage_cells(adata, **processing_params)
# Filtering
epi.pp.filter_cells(adata, min_features=10)
epi.pp.filter_features(adata, min_cells=10)
# Feature selection
epi.pl.cal_var(adata)
epi.pp.select_var_feature(adata, max_score=0.2, nb_features=50000)Use this tutorial to learn how to process and cluster count matrices specifically from the Cusanovich mouse scATAC-seq atlas.
https://nbviewer.jupyter.org/github/colomemaria/epiScanpy/blob/master/docs/tutorials/Cusanovich2018_BoneMarrow_data_processing_diffmap.htmlThis tutorial demonstrates how to build count matrices for different feature spaces (such as windows or promoters) across various cytosine contexts for single-cell DNA methylation data.
https://nbviewer.jupyter.org/github/colomemaria/epiScanpy/blob/master/docs/tutorials/bld_count_matrix_methylation_tutorial.htmlThe episcanpy.api.pp (Preprocessing) module handles data cleaning, normalization, and imputation.
pp.imputation_met: Impute missing methylation data.pp.load_met_noimput: Load methylation data without imputation.pp.readandimputematrix: Read and impute a methylation matrix.pp.filter_cells, pp.filter_features, pp.coverage_cells, pp.coverage_features, pp.commonness_features, pp.density_features.pp.normalize_total, pp.normalize_per_cell, pp.log1p, pp.binarize.pp.pca, pp.regress_out, pp.neighbors.pp.subsample, pp.downsample_counts, pp.sparse, pp.snap2anndata (convert to AnnData format).To install episcanpy using Conda, first ensure you have a working Python 3.5 or 3.6 environment. You should install the necessary dependencies (seaborn, scikit-learn, statsmodels, numba, python-igraph, and louvain), create a dedicated environment for scanpy, and then install episcanpy from the annadanese channel.
Follow these steps in order:
# Install dependencies
conda install seaborn scikit-learn statsmodels numba
conda install -c conda-forge python-igraph louvain
# Create a scanpy environment
conda create -n scanpy python=3.6 scanpy
# Install episcanpy from the annadanese channel
conda install -c annadanese episcanpyA tutorial is available for building custom count matrices using standard 10x single cell ATAC outputs (TSV or CellRanger outputs) and performing the corresponding data processing.
https://nbviewer.jupyter.org/github/colomemaria/epiScanpy/blob/master/docs/tutorials/Tutorial_building_custom_count_matrices_from_tsv_or_cellranger_outputs.ipynbYou can build count matrices from ATAC-seq BAM files using different sets of annotations, such as enhancers. The tutorial demonstrates how to handle these files to prepare data for downstream analysis.
https://nbviewer.jupyter.org/github/colomemaria/epiScanpy/blob/master/docs/tutorials/ATAC_bld_ct_mtx_tutorial.html