What Is NGS Testing and How Does It Work?

Next-generation sequencing, or NGS, is a technology that reads millions of DNA (or RNA) fragments simultaneously, producing a detailed readout of genetic information in hours rather than the days or weeks older methods required. Where first-generation sequencing laboriously decoded small stretches of DNA one at a time, modern NGS platforms can sequence the entire genome of even the most complex organisms in a single day.¹ That speed and scale have made NGS the backbone of clinical genetic testing, from identifying cancer mutations to screening a fetus for chromosomal abnormalities.

How NGS Works, Step by Step

The technology sounds complicated, but the logic is straightforward: break DNA into manageable pieces, read all those pieces at once, then use software to stitch the reads back together. Every NGS workflow follows roughly the same sequence, though the details vary by platform and application.

Library Preparation

“Library” is lab shorthand for a collection of DNA fragments ready to be sequenced. The process starts with extracting DNA from whatever sample you have, whether that is a blood draw, a tumor biopsy, or a swab. That DNA is then fragmented into shorter stretches, typically a few hundred base pairs long. Short synthetic sequences called adapters are attached to the ends of each fragment, essentially giving the sequencing machine something to grab onto.² A size-selection step often follows, filtering out fragments that are too short or too long. The quality of this preparation matters enormously: artifacts introduced here, such as DNA damage from oxidation during sample handling, can masquerade as real mutations later.³

Sequencing by Synthesis

The most widely used NGS chemistry, employed by Illumina instruments, is called sequencing by synthesis. The idea: watch DNA being copied one nucleotide at a time, and record which base (A, C, G, or T) gets added at each position. Each of the four nucleotide types carries a different colored fluorescent tag. Crucially, the nucleotides are also chemically modified so that after one is incorporated, the copying reaction pauses. The machine photographs the color, identifies the base, then chemically removes both the fluorescent tag and the blocking group in a single step so the next nucleotide can be added.⁴ This cycle of incorporate-read-cleave repeats hundreds of times for each fragment, and because millions of fragments are being read in parallel across the surface of a flow cell, the machine generates an enormous volume of data per run.

Variations on this chemistry exist. Some approaches combine modified nucleotides that allow the chain to keep growing with fluorescent terminators that stop the chain, reading the sequence from the termination events rather than every single incorporation.⁵ The underlying principle stays the same: build the complementary strand, detect which base was added, and move on.

Bioinformatics and Variant Calling

Raw sequencing data is useless without software to interpret it. A bioinformatics pipeline takes the millions of short reads coming off the machine, checks their quality, aligns them against a reference genome, and then identifies where the patient’s DNA differs from that reference. Those differences are called variants. A pipeline converts raw sequencing output into clinically meaningful information.⁶

Variant calling and annotation are not trivial problems. The software needs to distinguish real mutations from sequencing errors, classify each variant by its likely clinical significance, and attach standardized nomenclature so clinicians everywhere are speaking the same language. When researchers tested three widely used annotation tools against a manually verified set of 298 variants, accuracy ranged from about 93% to nearly 100%, meaning even small tool differences can lead to a handful of variants being misclassified.⁷ Many clinical-grade pipelines now chain multiple tools together, using one program for variant calling, another for annotation, and yet another for assigning pathogenicity labels according to professional guidelines.⁸ For larger structural changes in the genome, such as deletions or duplications spanning multiple genes, specialized tools classify pathogenicity and match the correct category for about four in five manually evaluated variants, with the remainder flagged as uncertain and routed to a human expert.⁹

Short Reads Versus Long Reads

Not all NGS platforms read DNA the same way. The most established machines, particularly those from Illumina, produce short reads, typically 150 to 300 base pairs each. These are highly accurate, fast, and relatively cheap per base. Short-read sequencing still dominates clinical and research labs for those reasons.¹⁰

Long-read technologies, led by Oxford Nanopore and Pacific Biosciences, read stretches of DNA that can be tens of thousands of bases long. That extra length is a big advantage when the genome contains repetitive regions, which are common. Short reads often cannot span a repeat, leaving ambiguity about how the pieces fit together. Long reads clear those hurdles, producing more complete genome assemblies with fewer gaps.¹¹ In microbial genomics, for instance, long-read assemblies were composed of fewer fragments and contained a higher proportion of important marker genes compared to short-read assemblies of the same organisms.¹²

The trade-off is that long reads have historically been less accurate per individual base. Creative workarounds exist, such as computationally chopping long reads into shorter fragments and running them through short-read analysis pipelines, which improved accuracy in population-level variant calling.¹¹ In practice, many labs now use both technologies depending on the question: short reads for rapid, cost-effective clinical tests, and long reads when the genome region in question is structurally complex.

Cancer Testing

Oncology is where NGS has arguably had its biggest clinical impact. Tumors accumulate genetic mutations, and identifying those mutations can guide treatment choices. Instead of testing one gene at a time, an NGS panel can screen dozens or even hundreds of cancer-related genes simultaneously. In hereditary cancer assessment, this means a single blood draw can be tested against a panel of susceptibility genes, catching mutations that would be missed by single-gene testing.¹³

In a study of nearly 500 breast cancer patients with clinical features suggesting hereditary risk, about one in five carried a harmful germline mutation across 16 cancer susceptibility genes. More than a third of those mutations were in genes other than BRCA1 and BRCA2, the ones most commonly tested in older single-gene approaches.¹⁴ Without the multi-gene panel, those patients would have received a reassuring but incomplete negative result. Separate work has confirmed that commercially available NGS panels are highly efficient at detecting BRCA1 and BRCA2 mutations from routine clinical DNA samples.¹⁵

NGS has also enabled liquid biopsy, in which tumor DNA circulating in the bloodstream is captured and sequenced instead of requiring a tissue biopsy. Circulating tumor DNA reflects the real-time genetic profile of the cancer, offering a less invasive way to track how a tumor is evolving and whether it is responding to treatment.¹⁶ Because a tumor can be genetically different from one spot to the next, a tissue biopsy sometimes misses mutations present elsewhere in the same cancer. Liquid biopsy partly addresses this sampling bias, though it has its own limitations: in one study of lung cancer patients, mutations in plasma DNA were detected in roughly 60% of cases, meaning a negative liquid biopsy does not rule out a mutation entirely. The authors recommended parallel testing of both tissue and blood for the most reliable results.¹⁷

Rare and Genetic Diseases

For families navigating a suspected genetic disorder in a child, NGS-based testing can end what is sometimes called the diagnostic odyssey: years of inconclusive tests and specialist referrals. Two flavors of NGS are used in this setting. Whole-exome sequencing reads only the protein-coding portions of the genome (about 1–2% of total DNA but home to most disease-causing mutations), while whole-genome sequencing reads everything. A cost-effectiveness analysis in children with suspected genetic disorders found that whole-genome sequencing had a diagnostic yield of about 64%, compared with about 58% for whole-exome sequencing.¹⁸ That roughly six-percentage-point gap may sound small, but for parents who have already exhausted other options, it can be the difference between an answer and continued uncertainty.

Prenatal Screening

Non-invasive prenatal testing, often called NIPT, is one of the most familiar applications of NGS for the general public. During pregnancy, fragments of fetal DNA circulate in the mother’s blood. By sequencing those fragments, labs can detect extra or missing chromosomes, screening for conditions like Down syndrome without the miscarriage risk associated with amniocentesis. Large-scale implementations of NIPT use Illumina sequencing platforms and process tens of thousands of pregnancies.¹⁹ To improve accuracy, labs measure the fraction of fetal DNA in the sample and use statistical methods to flag abnormalities on chromosomes 21, 18, and 13.²⁰ It is worth remembering that NIPT is a screening test, not a diagnostic one. A positive result still needs confirmation through amniocentesis or chorionic villus sampling.

Infectious Disease Diagnosis

Traditional methods for identifying an infection often depend on knowing, or at least guessing, what you are looking for. Culture a sample, test it against a panel of known pathogens, wait for results. Metagenomic NGS flips that approach: instead of targeting specific organisms, it sequences all the genetic material in a clinical sample and then sorts out what belongs to the patient and what belongs to a pathogen. This makes it possible to detect bacteria, viruses, fungi, and parasites simultaneously, without any prior assumptions about which infection might be present.²¹

This matters most for mysterious or hard-to-diagnose infections. A patient with encephalitis of unknown origin, for example, may have already tested negative for every common viral panel. Metagenomic NGS can identify unusual or unexpected pathogens that would never have made it onto a standard testing panel.²² A meta-analysis pooling results across multiple studies found that metagenomic NGS had a pooled sensitivity of about 75% and specificity of about 68% for detecting infections, with overall performance classified as excellent by standard metrics.²³ Those numbers reflect the approach’s strength in casting a wide net, though a specificity of 68% means false positives remain a real concern, particularly because human samples inevitably contain background microbial DNA that can be mistaken for a pathogen.

Known Limitations and Technical Pitfalls

NGS is powerful, but it has blindspots. One well-documented issue is GC bias: regions of DNA that are unusually rich or unusually poor in the bases G and C tend to be sequenced less reliably. At strong GC bias, some stretches of the genome are barely covered by reads, while others are over-represented, and both extremes degrade the quality of the final assembly.²⁴ For clinical labs, this means certain genes or genomic regions may need supplementary testing by other methods to ensure nothing is missed.

Sample quality also matters more than many people realize. DNA that has been stored improperly, extracted from old formalin-fixed tissue, or processed under suboptimal conditions can introduce artifacts that look like real mutations. Labs have adopted protective measures, such as buffer-exchanging all incoming DNA samples to remove potential contaminants and adding compounds that guard against oxidation during the physical shearing of DNA.³

Interpretation can be just as tricky as the chemistry. A sequencing run may flag thousands of variants, many of which are benign or of uncertain significance. Deciding which variants actually matter requires databases of known pathogenic mutations, population frequency data, and often a skilled geneticist or genetic counselor reviewing the results. Automation helps, but as the annotation-tool benchmarking studies show, even well-validated software disagrees on a small fraction of calls. In clinical genetics, that small fraction can translate to real patients getting ambiguous or incorrect answers.

Cost and Access

A common misconception is that NGS is prohibitively expensive. In some contexts, it actually saves money. When researchers compared the total cost of NGS-based testing against traditional sequential PCR-based testing in patients with metastatic lung cancer, NGS came out cheaper: roughly $4,900 per patient versus about $6,600 for the PCR approach, because a single NGS panel replaced multiple rounds of individual tests.²⁵ Those figures were from a US commercial insurance perspective; under Medicare the numbers were lower for both strategies, but NGS was still the least costly option.

The bigger access barrier is reimbursement, not list price. A survey of oncology practices across the Asia-Pacific region found that government reimbursement for NGS-based testing was inconsistent regardless of a country’s income level. Even some high-income nations did not offer government funding for either tissue-based or liquid biopsy sequencing. Among the barriers cited, test cost was mentioned by every respondent, followed by limited budgets and regulatory restrictions.²⁶ Insurance coverage and patient income were also flagged as major equity concerns. In other words, the technology exists and works, but who can actually get it depends heavily on where they live and how their health system is set up.

Secondary Findings and the Ethics of Knowing Too Much

When you sequence a large portion of someone’s genome, you sometimes find things you were not looking for. A test ordered to investigate a child’s developmental delay might incidentally reveal a mutation linked to adult-onset cancer risk. These are called secondary or incidental findings, and they raise genuinely difficult ethical questions. Should the lab report a finding unrelated to the original reason for testing? Does a patient have the right to opt out of learning about such findings? What about when the patient is a child, and the finding has implications for a parent?²⁷

Professional guidelines have converged on a few principles. Most policies worldwide recommend that labs look for clinically actionable, clearly pathogenic variants, and that patients receive pre-test counseling explaining what kinds of results might come back, including the possibility of unexpected findings. A systematic review of 63 policies from different countries found that the vast majority addressed how secondary findings should be disclosed, and most recommended that patients have a genuine choice about whether to learn about them.²⁸ In practice, consent forms often include a checkbox for opting in or out, and genetic counselors play an essential role in walking patients through what those choices mean before the test is run.

Keeping Genomic Data Safe

An NGS run on a single person can produce hundreds of gigabytes of raw data, and the information it encodes is uniquely identifying. Unlike a stolen credit card number, your genome cannot be reissued. That makes storage and privacy an active area of concern. Researchers have developed encryption schemes specifically for genomic data that allow secure cloud storage while still permitting limited computation on the encrypted data, so that, for instance, a clinician could run a query without the cloud provider ever seeing the raw sequence.²⁹ Other approaches focus on privacy-preserving storage formats that protect the sensitive portions of a genome file while keeping the rest accessible for clinical use.³⁰ These are still largely research tools rather than standard clinical infrastructure, but as genomic testing becomes routine, the pressure to solve storage and privacy at scale is growing quickly.

Beyond Human Health

Although clinical testing gets the most attention, NGS has spread far beyond the hospital. In agriculture, sequencing crop genomes has become a gateway to understanding and improving the plants that feed the world. Food crops including cereals, tuber crops, vegetables, and fruits have been sequenced, helping breeders identify genes for disease resistance, drought tolerance, and nutritional quality.³¹ Environmental scientists use metagenomic sequencing to catalog microbial communities in soil, water, and air, tracking how ecosystems respond to pollution or climate change. Wildlife biologists sequence endangered species to assess genetic diversity. Forensic labs use NGS to analyze degraded DNA samples that older methods could not handle. The underlying technology is the same in each case; what changes is the question being asked.

Where NGS Is Headed

The technology continues to evolve rapidly. RNA sequencing, which reads the messages cells actively produce rather than the static DNA blueprint, has matured to the point where researchers can measure gene activity in individual cells and even map where in a tissue those genes are being turned on. New long-read and direct RNA sequencing methods are revealing aspects of RNA biology, from folding patterns to molecular interactions, that short-read methods could not capture.³² On the clinical side, the push is toward faster turnaround, lower costs, and broader access. Portable long-read sequencers already fit in a coat pocket and have been used in remote field settings and outbreak investigations. As sequencing costs continue to fall and bioinformatics tools get more reliable, the question is shifting from “can we sequence this?” to “what should we do with the answer?”