scanpy

End-to-end single-cell RNA-seq analysis runs quality control, normalization, dimensionality reduction, Leiden clustering, marker identification, cell annotation, and trajectory inference on an AnnData object to produce interpretable cell embeddings and pseudotime trends.

QC filtering and normalization for scRNA-seq count data involves calculating mitochondrial QC metrics, applying sc.pp preprocessing functions, and filtering AnnData objects to prepare clean inputs for downstream dimensionality reduction and clustering.

10X data and count tables are supported inputs for generating UMAP and Leiden clustering results, alongside .h5ad files, allowing you to produce PCA and t-SNE embeddings directly from common scRNA-seq input formats.

Trajectory inference and pseudotime analysis use PAGA and DPT on processed AnnData objects to map cell state transitions, generating pseudotime trends that support biological interpretation of developmental progressions.

An .h5ad file is not strictly required to identify marker genes, as 10X data and count tables are also accepted, but all inputs are processed into AnnData objects to run rank_genes_groups for differential expression discovery.

After running dimensionality reduction and cell annotation, you can export results through AnnData writes, saving the processed .h5ad object alongside generated UMAP, t-SNE, and PCA plots for persistence and downstream visualization.