name: bio-single-cell-preprocessing description: Quality control, filtering, and normalization for single-cell RNA-seq using Seurat (R) and Scanpy (Python). Use for calculating QC metrics, filtering cells and genes, normalizing counts, identifying highly variable genes, and scaling data. Use when filtering, normalizing, and selecting features in single-cell data. tool_type: mixed primary_tool: Seurat measurable_outcome: Execute skill workflow successfully with valid output within 15 minutes. allowed-tools:

• read_file
• run_shell_command

Single-Cell Preprocessing

Quality control, filtering, normalization, and feature selection for scRNA-seq data.

Scanpy (Python)

Required Imports

import scanpy as sc
import numpy as np

Calculate QC Metrics

# Calculate mitochondrial gene percentage
adata.var['mt'] = adata.var_names.str.startswith('MT-')
sc.pp.calculate_qc_metrics(adata, qc_vars=['mt'], percent_top=None, log1p=False, inplace=True)

# Key metrics added to adata.obs:
# - n_genes_by_counts: genes detected per cell
# - total_counts: total UMI counts per cell
# - pct_counts_mt: percentage mitochondrial

Visualize QC Metrics

import matplotlib.pyplot as plt

sc.pl.violin(adata, ['n_genes_by_counts', 'total_counts', 'pct_counts_mt'], jitter=0.4, multi_panel=True)
sc.pl.scatter(adata, x='total_counts', y='pct_counts_mt')
sc.pl.scatter(adata, x='total_counts', y='n_genes_by_counts')

Filter Cells and Genes

# Filter cells by QC metrics
sc.pp.filter_cells(adata, min_genes=200)
sc.pp.filter_cells(adata, max_genes=5000)

# Filter by mitochondrial percentage
adata = adata[adata.obs['pct_counts_mt'] < 20, :].copy()

# Filter genes
sc.pp.filter_genes(adata, min_cells=3)

print(f'After filtering: {adata.n_obs} cells, {adata.n_vars} genes')

Store Raw Counts

# Store raw counts before normalization
adata.raw = adata.copy()
# Or use layers
adata.layers['counts'] = adata.X.copy()

Normalization

# Library size normalization (normalize to 10,000 counts per cell)
sc.pp.normalize_total(adata, target_sum=1e4)

# Log transform
sc.pp.log1p(adata)

Highly Variable Genes

# Identify highly variable genes (default: top 2000)
sc.pp.highly_variable_genes(adata, n_top_genes=2000, flavor='seurat_v3', layer='counts')

# Visualize
sc.pl.highly_variable_genes(adata)

# Check results
print(f'Highly variable genes: {adata.var.highly_variable.sum()}')

Subset to HVGs (Optional)

# Keep only highly variable genes for downstream analysis
adata_hvg = adata[:, adata.var.highly_variable].copy()

Scaling (Z-score)

# Scale to unit variance and zero mean
sc.pp.scale(adata, max_value=10)

Regress Out Confounders

# Regress out unwanted variation (e.g., cell cycle, mitochondrial)
sc.pp.regress_out(adata, ['total_counts', 'pct_counts_mt'])

Complete Preprocessing Pipeline

import scanpy as sc

adata = sc.read_10x_mtx('filtered_feature_bc_matrix/')

# QC
adata.var['mt'] = adata.var_names.str.startswith('MT-')
sc.pp.calculate_qc_metrics(adata, qc_vars=['mt'], inplace=True)

# Filter
sc.pp.filter_cells(adata, min_genes=200)
sc.pp.filter_genes(adata, min_cells=3)
adata = adata[adata.obs['pct_counts_mt'] < 20, :].copy()

# Store raw
adata.raw = adata.copy()

# Normalize
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)

# HVGs
sc.pp.highly_variable_genes(adata, n_top_genes=2000)

# Scale
adata = adata[:, adata.var.highly_variable].copy()
sc.pp.scale(adata, max_value=10)

Seurat (R)

Required Libraries

library(Seurat)
library(ggplot2)

Calculate QC Metrics

# Calculate mitochondrial percentage
seurat_obj[['percent.mt']] <- PercentageFeatureSet(seurat_obj, pattern = '^MT-')

# View QC metrics
head(seurat_obj@meta.data)

Visualize QC Metrics

# Violin plots
VlnPlot(seurat_obj, features = c('nFeature_RNA', 'nCount_RNA', 'percent.mt'), ncol = 3)

# Scatter plots
plot1 <- FeatureScatter(seurat_obj, feature1 = 'nCount_RNA', feature2 = 'percent.mt')
plot2 <- FeatureScatter(seurat_obj, feature1 = 'nCount_RNA', feature2 = 'nFeature_RNA')
plot1 + plot2

Filter Cells

# Filter by QC metrics
seurat_obj <- subset(seurat_obj,
    subset = nFeature_RNA > 200 &
             nFeature_RNA < 5000 &
             percent.mt < 20)

cat('After filtering:', ncol(seurat_obj), 'cells\n')

Normalization (Log Normalization)

# Standard log normalization
seurat_obj <- NormalizeData(seurat_obj, normalization.method = 'LogNormalize', scale.factor = 10000)

Normalization (SCTransform)

# SCTransform - recommended for most workflows
# Combines normalization, scaling, and HVG selection
seurat_obj <- SCTransform(seurat_obj, vars.to.regress = 'percent.mt', verbose = FALSE)

Find Variable Features

# Identify highly variable features (if not using SCTransform)
seurat_obj <- FindVariableFeatures(seurat_obj, selection.method = 'vst', nfeatures = 2000)

# Visualize
top10 <- head(VariableFeatures(seurat_obj), 10)
plot1 <- VariableFeaturePlot(seurat_obj)
plot2 <- LabelPoints(plot = plot1, points = top10, repel = TRUE)
plot2

Scaling

# Scale data (if not using SCTransform)
all.genes <- rownames(seurat_obj)
seurat_obj <- ScaleData(seurat_obj, features = all.genes)

# Or scale only variable features (faster)
seurat_obj <- ScaleData(seurat_obj)

Regress Out Confounders

# Regress out unwanted variation during scaling
seurat_obj <- ScaleData(seurat_obj, vars.to.regress = c('percent.mt', 'nCount_RNA'))

Complete Preprocessing Pipeline (Log Normalization)

library(Seurat)

counts <- Read10X(data.dir = 'filtered_feature_bc_matrix/')
seurat_obj <- CreateSeuratObject(counts = counts, min.cells = 3, min.features = 200)

# QC
seurat_obj[['percent.mt']] <- PercentageFeatureSet(seurat_obj, pattern = '^MT-')

# Filter
seurat_obj <- subset(seurat_obj,
    subset = nFeature_RNA > 200 & nFeature_RNA < 5000 & percent.mt < 20)

# Normalize
seurat_obj <- NormalizeData(seurat_obj)

# HVGs
seurat_obj <- FindVariableFeatures(seurat_obj, nfeatures = 2000)

# Scale
seurat_obj <- ScaleData(seurat_obj)

Complete Preprocessing Pipeline (SCTransform)

library(Seurat)

counts <- Read10X(data.dir = 'filtered_feature_bc_matrix/')
seurat_obj <- CreateSeuratObject(counts = counts, min.cells = 3, min.features = 200)

# QC
seurat_obj[['percent.mt']] <- PercentageFeatureSet(seurat_obj, pattern = '^MT-')

# Filter
seurat_obj <- subset(seurat_obj,
    subset = nFeature_RNA > 200 & nFeature_RNA < 5000 & percent.mt < 20)

# SCTransform (does normalization, HVG, and scaling)
seurat_obj <- SCTransform(seurat_obj, vars.to.regress = 'percent.mt', verbose = FALSE)

QC Thresholds Reference

Method Comparison

normalize_total

log1p

NormalizeData

SCTransform

highly_variable_genes

FindVariableFeatures

scale

ScaleData

regress_out

ScaleData(vars.to.regress)

SCTransform(vars.to.regress)

Related Skills

• data-io - Load data before preprocessing
• clustering - PCA and clustering after preprocessing
• markers-annotation - Find markers after clustering