Single Cell RNA Sequencing Pipeline: How to Build a Complete Scanpy Analysis (2026 Guide)

Single Cell RNA Sequencing Pipeline: How to Build a Complete Scanpy Analysis (2026 Guide)

Single cell RNA sequencing (scRNA-seq) has revolutionized our understanding of cellular heterogeneity—and Scanpy remains the most trusted open-source Python framework for analyzing it. In 2026, researchers across immunology, oncology, and developmental biology rely on Scanpy to transform raw sequencing data into biologically meaningful insights. This guide walks you through a complete, reproducible pipeline using the PBMC 3k dataset, from loading to annotation.

Step 1: Data Loading and Quality Control

Begin by loading your 10x Genomics or Drop-seq data using scanpy.read_10x_mtx(). Apply strict quality control: filter out low-quality cells with < 500 genes or >10% mitochondrial reads. Use sc.pp.calculate_qc_metrics() to assess cell viability and remove doublets. High cell viability (>85%) is critical for robust downstream analysis.

Step 2: Normalization, Feature Selection, and Dimensionality Reduction

Normalize counts using log-normalization (sc.pp.normalize_total() + sc.pp.log1p()). Identify highly variable genes (HVGs) with sc.pp.highly_variable_genes()—typically 1,000–3,000 genes. Apply PCA for linear dimensionality reduction, retaining the first 20–30 components. Then, compute UMAP or t-SNE embeddings for non-linear visualization. UMAP is preferred for preserving global structure and scalability with large datasets.

Step 3: Clustering and Cell Type Annotation

Cluster cells using the Leiden algorithm (sc.tl.leiden()) with resolution tuned between 0.4–1.2 to balance granularity. Identify marker genes per cluster via differential expression (sc.tl.rank_genes_groups()). Annotate cell types using reference-based tools like SingleR or CellTypist, which automate labeling using published scRNA-seq atlases. Avoid manual annotation bias by cross-validating with known markers (e.g., CD3E for T cells, CD19 for B cells).

Step 4: Integrating Multimodal Insights

While Scanpy handles RNA data natively, integrate DNA mutation data via external tools like Seurat’s integration pipeline or CellChat for multi-omics. For somatic mutation overlay, tools like MOFA+ or Harmony can align RNA and DNA latent spaces—enabling cancer researchers to pinpoint mutation-driven subpopulations within RNA-defined clusters.

Step 5: Reproducibility and Scalability

Use Jupyter notebooks with version control (Git) and containerization (Docker) to ensure reproducibility. Leverage cloud platforms like Google Cloud Life Sciences or AWS Batch for scaling to 100K+ cells. Institutions now adopt Scanpy pipelines as standard infrastructure, with training modules embedded in graduate bioinformatics curricula worldwide.

In summary, a Scanpy-based scRNA-seq pipeline—when paired with best practices in QC, clustering, and annotation—delivers accurate, scalable, and interpretable results. As data volumes grow, transparency and reproducibility are no longer optional; they’re essential for clinical translation.