# Next-Generation Sequencing: Platforms, Pipelines, and Quality Metrics

## Introduction

Next-generation sequencing (NGS) technologies have transformed clinical genetics by enabling massively parallel DNA sequencing at decreasing cost. Understanding the platforms, bioinformatics pipelines, and quality metrics is essential for clinical geneticists who must interpret reports generated from these technologies and recognize their limitations.

## Sequencing Platforms

### Short-Read Platforms

Illumina (sequencing by synthesis) is the dominant platform in clinical laboratories. DNA fragments are clonally amplified on a flow cell via bridge amplification, and fluorescently labeled reversible terminators are incorporated one nucleotide at a time. Read lengths are typically 150-300 bp paired-end. Platforms range from benchtop instruments (MiSeq, NextSeq) to high-throughput systems (NovaSeq, NovaSeq X). The error rate is approximately 0.1-0.5% per base, consisting primarily of substitution errors. Ion Torrent (semiconductor sequencing) detects hydrogen ions released during nucleotide incorporation and is prone to homopolymer errors (insertions/deletions in repeat regions). It is used in some clinical panels, particularly in oncology.

### Long-Read Platforms

Pacific Biosciences (PacBio) HiFi uses single-molecule real-time (SMRT) sequencing with circular consensus sequencing to produce high-fidelity reads of 10-25 kb with greater than 99.9% accuracy. It excels at structural variant detection, phasing, and repeat expansions. Oxford Nanopore Technologies (ONT) passes DNA through a protein nanopore, with current changes identifying bases. Ultra-long reads exceeding 100 kb (and even megabase-scale) are achievable. Native DNA sequencing enables direct methylation detection without bisulfite conversion. The per-read error rate is approximately 1-5% but is improving rapidly and can be compensated by depth of coverage.

| Feature | Illumina (Short-Read) | PacBio HiFi (Long-Read) | Oxford Nanopore (Long-Read) |
|---|---|---|---|
| Read length | 150–300 bp | 10–25 kb | >100 kb possible |
| Per-base accuracy | 99.5–99.9% | >99.9% (HiFi) | 95–99% (improving) |
| Error type | Substitutions | Rare (residual indels) | Indels in homopolymers |
| Throughput | Very high | Moderate-high | Moderate |
| Methylation detection | Requires bisulfite conversion | Yes (kinetics-based) | Yes (native, direct) |
| SV/repeat detection | Limited | Excellent | Excellent |
| Clinical use | Standard panels, exome, genome | Emerging (SVs, repeats, phasing) | Emerging (rapid diagnostics) |
| Cost per Gb | Low | Moderate | Low-moderate |

![Comparison diagram of Illumina sequencing-by-synthesis, PacBio SMRT, and Oxford Nanopore sequencing mechanisms](images/ngs-platform-comparison.png)

## Library Preparation

### Key Steps

DNA extraction and quality assessment includes quantity measurement by fluorometry, quality assessment via fragment size electrophoresis, and purity determination using the 260/280 ratio. Fragmentation uses enzymatic or mechanical (sonication) methods for short-read sequencing but is not required for long-read native libraries. Adapter ligation adds platform-specific adapters that enable binding to the flow cell or SMRT cell and contain index sequences for multiplexing. Target enrichment, when applicable, employs either hybridization capture (biotinylated probes hybridize to targets followed by streptavidin pulldown, used for exome and panel sequencing) or amplicon-based methods (PCR amplification of target regions, which is faster but more prone to allelic dropout and amplification bias). Multiplexing uses index (barcode) sequences to allow pooling of multiple samples per sequencing run.

## Bioinformatics Pipeline

### Primary Analysis

Base calling converts raw signal data to nucleotide sequences with quality scores. Demultiplexing assigns reads to individual samples based on index sequences. The output is FASTQ files containing reads and per-base quality scores on the Phred scale.

### Secondary Analysis

Read alignment maps reads to a reference genome (GRCh37/hg19 or GRCh38/hg38) using aligners such as BWA-MEM2 for short reads or minimap2 for long reads. Post-alignment processing includes duplicate marking (Picard/sambamba) and base quality score recalibration (BQSR). Variant calling employs different tools by variant type: GATK HaplotypeCaller or DeepVariant for SNVs and small indels; ExomeDepth, XHMM, or cn.MOPS for copy number variants; and Manta, DELLY, or Sniffles for structural variants. The output is VCF (Variant Call Format) files.

### Tertiary Analysis

Variant annotation adds gene, transcript, consequence, population frequency, and in-silico prediction information using tools such as VEP (Variant Effect Predictor), ANNOVAR, or SnpEff. Variant filtration narrows candidates by quality, frequency, consequence, and inheritance pattern. Variant interpretation applies ACMG/AMP classification (pathogenic, likely pathogenic, VUS, likely benign, benign). Clinical reporting generates the final report with interpreted variants.

![Flowchart of the NGS bioinformatics pipeline from raw data through variant calling and clinical interpretation](images/ngs-bioinformatics-pipeline.png)

## Quality Metrics

### Sequencing Quality

The Phred quality score (Q score) quantifies base-calling confidence: Q30 means a 1 in 1,000 error probability (99.9% accuracy). Clinical laboratories typically require more than 80% of bases at Q30. Cluster density and pass filter rate indicate optimal loading for Illumina platforms. The overall per-base error rate across the run is monitored.

### Coverage Metrics

Mean depth of coverage indicates the average number of reads at each position; clinical exome typically achieves 100-150x and gene panels 200-500x. Coverage uniformity measures the percentage of target bases at a specified minimum depth, with the clinical standard being greater than 95% of target at 20x or above for constitutional testing; somatic or mosaic detection may require 500x or more. The on-target rate quantifies the proportion of sequenced reads mapping to the intended target regions.

### Variant Quality

Genotype quality (GQ) expresses confidence in the called genotype. Variant allele fraction (VAF) is expected at approximately 50% for heterozygous and approximately 100% for homozygous in constitutional samples. Strand bias assesses whether variants are supported by reads from both forward and reverse strands. Mapping quality reflects confidence that the read aligns to the correct genomic location.

### Laboratory-Level QC

Known positive controls with characterized variants assess analytical sensitivity. No-template controls detect contamination. Reproducibility metrics track inter-run and intra-run concordance. Turnaround time from sample receipt to report is typically 2-6 weeks for clinical testing.

![Dashboard visualization showing key NGS quality metrics including coverage distribution, Q-score histogram, and on-target rates](images/ngs-quality-dashboard.png)

## Clinical Considerations

Analytical sensitivity varies by variant type: SNVs achieve greater than 99%, small indels approximately 95-99%, and CNVs and structural variants are platform-dependent. GC-rich regions and pseudogenes are challenging for short-read sequencing. Sanger confirmation of reportable variants is still performed by many clinical labs, though some have validated NGS as standalone. Reference genome choice (GRCh37 versus GRCh38) affects variant coordinates and must be documented in reports.

## Clinical Pearls

Coverage depth alone is insufficient; coverage uniformity across the target region is equally critical for clinical reliability. Short-read NGS has significant blind spots including GC-extreme regions, segmental duplications, and structural variants, which may require complementary technologies. Understanding the bioinformatics pipeline is essential for interpreting negative results, as a gene may be on the panel but poorly covered at the exon level. Quality metrics should be reviewed on every clinical report, and low coverage in a gene of interest may warrant reflex to an alternative assay.

## References

1. Goodwin S, McPherson JD, McCombie WR. Coming of age: ten years of next-generation sequencing technologies. *Nature Reviews Genetics*. 2016;17(6):333-351.
2. Rehm HL, Bale SJ, Bayrak-Toydemir P, et al. ACMG clinical laboratory standards for next-generation sequencing. *Genetics in Medicine*. 2013;15(9):733-747.
3. Roy S, Coldren C, Karunamurthy A, et al. Standards and guidelines for validating next-generation sequencing bioinformatics pipelines. *Journal of Molecular Diagnostics*. 2018;20(1):4-27.
4. Wenger AM, Peluso P, Hall RJ, et al. Accurate circular consensus long-read sequencing improves variant detection and assembly of a human genome. *Nature Biotechnology*. 2019;37(10):1155-1162.