Skip to content

Chapter 11: Introduction to Next-Generation Sequencing (NGS)

📥 Download Lecture Slides (PPTX)

11.1 The Sequencing Revolution

For decades, Sanger sequencing was the gold standard. It was reliable but slow and expensive, sequencing one small piece of DNA at a time. The Human Genome Project took over 10 years and cost billions of dollars using this method.

Next-Generation Sequencing (NGS) changed everything. Instead of one-by-one, NGS technologies sequence millions or billions of DNA fragments simultaneously. This massive parallelism has caused the cost of sequencing to plummet faster than the cost of computing (a trend that outpaces Moore's Law).

11.2 The Core Principle: Massive Parallelism

Illustration of NGS Workflow

While different NGS platforms (like Illumina, PacBio, or Oxford Nanopore) have unique chemistries, the general workflow is similar:

  1. Fragmentation: The genome is broken into millions of small, manageable pieces.
  2. Library Preparation: Special adapters are attached to the ends of these fragments.
  3. Sequencing: The fragments are loaded onto a flow cell (a glass slide) and sequenced in parallel. For Illumina, this involves taking a picture each time a new, fluorescently-tagged nucleotide is added to the growing strand.
  4. Data Output: The machine outputs the sequence data for each fragment into a specific file format.

11.3 The FASTQ File: Sequences and Quality

The standard output file from most NGS machines is the FASTQ file. It's like a FASTA file, but with a crucial addition: a quality score for each base.

A FASTQ record has four lines: 1. @SEQ_ID: The sequence identifier, starting with an @. 2. GATTACA...: The raw sequence of bases. 3. +: A separator line, sometimes repeating the ID. 4. #>>?A?...: The quality string. Each character represents a quality score for the corresponding base in line 2.

11.4 Phred Quality Scores

The characters in the quality string are not random; they are ASCII characters that encode a Phred quality score (Q score). The score relates to the probability of an error in the base call.

  • Q10: 1 in 10 chance of error (90% accuracy)
  • Q20: 1 in 100 chance of error (99% accuracy)
  • Q30: 1 in 1,000 chance of error (99.9% accuracy)

Q30 is generally considered the benchmark for high-quality data.

11.5 Bioinformatics in Action: Parsing FASTQ with Biopython

Just like SeqIO can parse FASTA files, it can also handle FASTQ files, automatically interpreting the quality scores for you.

from Bio import SeqIO

# Assume we have a file named 'reads.fastq'

# SeqIO.parse takes the filename and the format name
for record in SeqIO.parse("reads.fastq", "fastq"):

    # The record object has the sequence and ID...
    print(f"ID: {record.id}")
    print(f"Sequence: {record.seq}")

    # ...and it also has the quality scores as a list of integers!
    # This is the raw Phred score for the first base.
    first_base_quality = record.letter_annotations["phred_quality"]
    print(f"Quality of first base: {first_base_quality}")

    # Let's find the average quality of this read
    avg_quality = sum(record.letter_annotations["phred_quality"]) / len(record.seq)
    print(f"Average read quality: {avg_quality:.2f}")
    print("---")
    # We'll just look at the first record for this example
    break

This ability to programmatically check the quality of your data is the first step in any NGS analysis pipeline. Low-quality reads are often trimmed or discarded before moving on to assembly or alignment.

Summary

NGS allows for massive parallel sequencing, generating huge amounts of data quickly and cheaply. This data is stored in FASTQ files, which contain both the sequence and a Phred quality score for each base. We use tools like Biopython to parse these files and assess data quality before analysis.

11.6 Common NGS Data Types and File Formats

  • FASTQ: Raw reads + quality.
  • FASTA: Assembled sequences or reference genomes (no quality scores).
  • SAM/BAM/CRAM: Aligned reads (SAM is human-readable; BAM is compressed binary; CRAM is compressed with reference-based compression).
  • VCF: Variant calls (SNPs, indels) and annotations.

A standard short-read (Illumina) pipeline for variant discovery or RNA-Seq analysis typically follows:

  1. Raw data QC: FastQC, then aggregate with MultiQC.
  2. Trimming/filtering: TrimGalore/fastp to remove adapters and low-quality bases.
  3. Alignment: BWA-MEM2 for short reads, minimap2 for long reads or transcriptome alignment.
  4. Post-processing: samtools for sorting/indexing; mark duplicates with Picard.
  5. Variant calling: GATK Best Practices pipeline, FreeBayes, or DeepVariant (machine-learning based).
  6. Annotation: SnpEff, VEP to add gene and functional context.

11.8 Reproducibility: Workflows, Containers, and Data Provenance

Reproducible analysis is essential. Recommendations:

  • Use workflow managers: Snakemake, Nextflow, or Cromwell to encode pipelines as reproducible workflows.
  • Containerize tools: use Docker or Singularity/Apptainer to lock tool versions and dependencies.
  • Keep manifests: record conda/pip environments or provide Dockerfile/Singularity images.
  • Use MultiQC and automated reports to capture QC metadata.

11.9 Emerging Methods and Notes

  • Long-read sequencing (Oxford Nanopore, PacBio) enables isoform-level RNA analysis, direct RNA sequencing, and improved structural variant detection.
  • Single-cell sequencing requires specialized preprocessing (UMI handling, cell barcodes) and downstream tools (CellRanger, Scanpy, Seurat).
  • Privacy and ethics: human sequence data often requires controlled-access storage and adherence to consent agreements.

11.10 Quick example Snakemake rule (alignment)

rule align:
  input:
    reads="{sample}.fastq.gz",
    ref="ref/genome.fa"
  output:
    bam="{sample}.bam"
  shell:
    """
    bwa-mem2 mem {input.ref} {input.reads} | samtools sort -o {output.bam}
    samtools index {output.bam}
    """

This small rule demonstrates how to pin commands into a reproducible workflow. Workflow files should be version-controlled alongside configuration and environment manifests.