Fundamental Next-Generation Sequencing Technologies

Why This Matters

Next-generation sequencing technologies form the backbone of modern computational genomics—and you'll be tested on more than just platform names. Understanding how each technology generates data determines everything downstream: what computational pipelines you'll use, what biases you'll need to correct for, and which biological questions each platform can actually answer. The differences between short-read and long-read technologies, optical versus non-optical detection, and sequencing-by-synthesis versus ligation-based approaches aren't just technical trivia—they're the foundation for choosing appropriate analysis methods.

When you encounter questions about read alignment, assembly algorithms, or error correction strategies, you're really being asked: do you understand why different sequencing chemistries produce different data characteristics? A platform's read length, error profile, and throughput directly shape the computational challenges you'll face. Don't just memorize that Illumina produces short reads and PacBio produces long reads—know why that matters for genome assembly, structural variant detection, and transcriptome analysis.

---

Short-Read Sequencing Platforms

Short-read technologies revolutionized genomics by enabling massively parallel sequencing at low cost. The trade-off is read length for throughput—these platforms sacrifice the ability to span repetitive regions in exchange for generating billions of reads per run, making them ideal for applications where coverage depth matters more than contiguity.

Illumina Sequencing (Sequencing by Synthesis)

• Reversible dye terminators allow single-nucleotide incorporation per cycle—each base is identified by its fluorescent label before the terminator is cleaved, enabling the next incorporation
• Massively parallel throughput sequences millions to billions of fragments simultaneously on flow cell clusters, making it the dominant platform for most genomics applications
• Short reads (50–300 bp) require sophisticated alignment algorithms but provide excellent coverage uniformity for variant calling, RNA-Seq, and targeted resequencing

Ion Torrent Semiconductor Sequencing

• pH-based detection measures hydrogen ion release during nucleotide incorporation—no optical system required, reducing instrument cost and complexity
• Homopolymer errors arise because multiple identical bases incorporate simultaneously, making signal intensity proportional to run length (a common source of indel errors)
• Rapid turnaround and lower capital costs make it suitable for smaller labs, clinical diagnostics, and targeted panels rather than large-scale whole-genome projects

SOLiD Sequencing

• Ligation-based chemistry uses fluorescently labeled di-base probes rather than polymerase extension, providing inherent error-checking through two-base encoding
• High raw accuracy made it valuable for SNP detection and applications requiring precise variant calls, though throughput limitations reduced adoption
• Short reads (50–75 bp) and complex data encoding required specialized alignment tools (largely superseded by Illumina for most applications)

---

Long-Read Sequencing Platforms

Long-read technologies solve the fundamental limitation of short reads: the inability to span repetitive elements and resolve structural complexity. These platforms sacrifice per-base accuracy (in raw reads) and throughput for the ability to generate reads that can bridge gaps, phase haplotypes, and capture full-length transcripts.

Pacific Biosciences (PacBio) SMRT Sequencing

• Single-molecule real-time detection observes polymerase activity in zero-mode waveguides, capturing nucleotide incorporation as it happens without amplification bias
• Circular consensus sequencing (CCS/HiFi) passes the polymerase around circular templates multiple times, achieving $>99\%$ accuracy through error averaging
• Reads exceeding 15,000 bp enable de novo assembly of complex genomes, full-length isoform sequencing, and direct detection of base modifications like methylation

Oxford Nanopore Sequencing

• Ionic current measurement detects characteristic disruptions as DNA translocates through protein nanopores—the only platform requiring no synthesis or amplification chemistry
• Ultra-long reads (megabase-scale) can span entire repetitive regions, centromeres, and complex structural variants that fragment with any other technology
• Portable MinION devices enable real-time, field-deployable sequencing for pathogen surveillance, environmental sampling, and point-of-care diagnostics

---

Historical and Transitional Technologies

Understanding deprecated platforms helps you interpret legacy datasets and appreciate why current technologies evolved as they did. These methods introduced key innovations that shaped modern sequencing chemistry.

454 Pyrosequencing

• Pyrophosphate detection generates light proportional to nucleotide incorporation—the first commercially successful NGS platform, enabling the "$1000 genome" race
• Longer reads (up to 1,000 bp) made it valuable for de novo assembly and metagenomics before long-read platforms matured
• Homopolymer limitations and higher per-base costs led to discontinuation in 2016, though its bead-based emulsion PCR approach influenced Ion Torrent development

---

Library Preparation Strategies

These aren't sequencing platforms themselves but library construction methods that enhance what any short-read platform can achieve. They transform the information content of sequencing data by providing spatial context beyond individual reads.

Paired-End Sequencing

• Bidirectional reads from both ends of a fragment provide insert size information—critical for detecting indels, improving alignment specificity, and scaffolding assemblies
• Standard fragment sizes (200–500 bp) allow reads to overlap for error correction or remain separated for structural inference, depending on application needs
• Computational assumption of known insert size distribution enables split-read and discordant-pair analysis for structural variant detection

Mate-Pair Sequencing

• Long-range linking circularizes fragments of 2–10 kb before sequencing junction points, connecting distant genomic regions that paired-end libraries cannot span
• Scaffolding applications use mate-pair data to order and orient contigs, bridging gaps caused by repeats longer than standard insert sizes
• Chimeric artifacts from incomplete circularization require computational filtering but provide irreplaceable long-range information for assembly

---

Application-Specific Sequencing Methods

These methods combine NGS with biochemical enrichment or selection to answer specific biological questions. The computational analysis differs fundamentally from whole-genome approaches because you're interpreting enrichment signals, not uniform coverage.

RNA-Seq

• Transcriptome quantification converts RNA to cDNA for sequencing, enabling measurement of gene expression, alternative splicing, and novel transcript discovery
• Library preparation choices (poly-A selection vs. ribosomal depletion, stranded vs. unstranded) determine what RNA populations you capture and how you interpret strand origin
• Computational pipelines require splice-aware alignment, transcript assembly or quantification against references, and statistical models for differential expression that account for count-based data distributions

ChIP-Seq

• Protein-DNA interaction mapping uses antibody pulldown to enrich DNA fragments bound by specific proteins (transcription factors, modified histones), then sequences to identify binding locations
• Peak calling algorithms distinguish true binding sites from background noise, requiring input controls and statistical models for enrichment significance
• Integration challenges arise when combining ChIP-Seq peaks with expression data, motif analysis, and other epigenomic assays to reconstruct regulatory networks

Whole Genome Sequencing (WGS)

• Comprehensive variant detection captures SNPs, indels, and structural variants across the entire genome without target bias, enabling population genetics and clinical diagnosis
• Coverage requirements vary by application—germline variant calling typically needs $30\text{–}50\times$, while somatic mutation detection in heterogeneous tumors may require $100\times$ or more
• Computational infrastructure demands substantial storage (hundreds of GB per sample), efficient alignment pipelines, and variant annotation databases for biological interpretation

---

Quick Reference Table

---

Self-Check Questions

1. Which two platforms use non-optical detection methods, and how do their error profiles differ as a result?

2. You need to assemble a plant genome with large repetitive regions and characterize structural variants. Compare the advantages of PacBio HiFi versus Oxford Nanopore for this application—which would you choose and why?

3. Explain why paired-end sequencing improves alignment specificity compared to single-end reads, and describe a scenario where mate-pair libraries would be necessary instead.

4. A collaborator hands you a dataset and says "analyze this for differential expression." What library preparation details do you need to know before choosing a computational pipeline, and why do these choices matter?

5. If an FRQ asks you to design a study detecting transcription factor binding sites genome-wide, which sequencing application would you use, and what computational steps distinguish its analysis from standard WGS variant calling?