Preprocessing and QC - iffatAGheyas/bioinformatics-tutorial-wiki GitHub Wiki

6.1.2 Preprocessing & QC

After sequencing, the first step is to turn raw FASTQ files into a gene × cell count matrix, while performing basic QC to remove empty droplets or low-quality cells.

---

A. Demultiplexing & Alignment

1. 10x Genomics (Cell Ranger)

   • Input: raw BCL (Illumina) or FASTQ directory
   • Command:

     cellranger count \
       --id=SampleA \
       --transcriptome=/path/to/refdata-cellranger \
       --fastqs=raw_data/SampleA/ \
       --sample=SampleA \
       --localcores=8 \
       --localmem=64
     

   • What it does:
     • Demultiplexes BCL → per‐sample FASTQ (if needed)
     • Aligns reads (STAR internally) to the reference
     • Collapses UMIs, filters barcodes (empty‐droplet removal)
     • Generates outs/filtered_feature_bc_matrix/

2. Open‐source: STARsolo

   • Input: paired trimmed FASTQ
   • Command:

     mkdir -p align/STARsolo/SampleA
     STAR \
       --runThreadN 8 \
       --genomeDir ref/STAR_index \
       --readFilesIn trimmed/SampleA_R1.fastq.gz trimmed/SampleA_R2.fastq.gz \
       --readFilesCommand zcat \
       --soloType CB_UMI_Simple \
       --soloCBstart 1 --soloCBlen 16 \
       --soloUMIstart 17 --soloUMIlen 12 \
       --soloFeatures Gene \
       --soloOutFileNames align/STARsolo/SampleA/ Solo.out
     

   • What it does:
     • Aligns reads in splice-aware mode
     • Extracts cell barcodes (CB) & UMIs
     • Produces Solo.out/Gene/filtered directory

---

B. Generating Gene × Cell Matrices

Whether using Cell Ranger or STARsolo, the final filtered matrix is in Matrix Market format:

Outputs (gene × cell count matrices):

• Cell Ranger
  outs/filtered_feature_bc_matrix/

• STARsolo
  align/STARsolo/SampleA/Solo.out/Gene/filtered

  ├── barcodes.tsv.gz    # list of cell barcodes  
  ├── features.tsv.gz    # gene IDs and names  
  └── matrix.mtx.gz      # sparse count matrix (genes × cells)  
  
  

1. Inspect file contents:

   zcat matrix.mtx.gz | head -n 5
   zcat barcodes.tsv.gz | head -n 5
   zcat features.tsv.gz | head -n 5

2. Load into R (Seurat):

library(Seurat)
counts <- Read10X(data.dir = "outs/filtered_feature_bc_matrix/")
seurat_obj <- CreateSeuratObject(counts, project="SampleA")

3.** Load into Python (Scanpy):**

import scanpy as sc
adata = sc.read_10x_mtx(
    "outs/filtered_feature_bc_matrix/",
    var_names='gene_symbols',
    cache=True
)

QC Checks

• Cell count: number of barcodes in barcodes.tsv.gz
• Gene count per cell: nFeature_RNA (Seurat) or adata.obs['n_genes'] (Scanpy)
• UMI count per cell: nCount_RNA or adata.obs['n_counts']
• Mitochondrial fraction: percentage of reads mapping to mitochondrial genes

Use these metrics to filter out low-quality cells before downstream analysis.