albert-ying/autonomous-lab/single-cell-rna-seq

Single-cell RNA-seq Analysis

When to use

• Processing 10X Genomics scRNA-seq data
• Multi-sample integration and batch correction
• Cell type identification using marker databases
• Condition-specific differential expression within cell types

Standard workflow

1. Data loading: Read 10X MTX format (barcodes, features, matrix)
2. QC filtering: Remove low-quality cells (low gene count, high mito%)
3. Normalization: Library size normalization + log transformation
4. Integration: Batch correction across samples if needed
5. Clustering: PCA → neighbors → Leiden clustering
6. Cell type annotation: Use known marker genes or reference databases
7. Differential expression: Compare conditions within each cell type

Cell type annotation strategy

import pandas as pd
# Load marker database (e.g., CellMarker)
markers_db = pd.read_excel('Cell_marker_Seq.xlsx')
# Get cluster markers
sc.tl.rank_genes_groups(adata, 'leiden', method='wilcoxon')
# For each cluster, compare top markers against database
# Assign cell type based on best marker overlap

Differential expression between conditions

# For each cell type cluster
for cluster in adata.obs['leiden'].unique():
    mask = adata.obs['leiden'] == cluster
    sub = adata[mask].copy()
    # Compare pre vs post exercise
    sc.tl.rank_genes_groups(sub, 'condition', method='wilcoxon',
                            groups=['post'], reference='pre')
    result = sc.get.rank_genes_groups_df(sub, group='post')

Key decisions

• Use raw counts for DE analysis (not scaled data)
• Apply Benjamini-Hochberg correction for multiple testing
• Report both significant and non-significant results per cluster