How to Read DNA Fragments: A Beginner’s Look at DNA

Reading DNA fragments means determining the exact order of the four chemical letters (A, T, C, and G) that make up a stretch of DNA. Scientists do this through a set of techniques collectively called DNA sequencing, and the process has evolved from painstaking manual methods in the late 1970s to machines that can decode billions of letters in a single run. The basic idea, though, has stayed surprisingly consistent: break DNA into manageable pieces, sort or label those pieces in ways that reveal their sequence, and then stitch the results back together with software. What follows is a walk through each stage of that process, from the molecule itself to the real-world problems fragment reading now solves.

What DNA Actually Looks Like at the Molecular Level

DNA is a long, thread-like molecule built from two strands wound around each other in the famous double helix. Each strand is a chain of smaller units called nucleotides, and every nucleotide contains one of four bases: adenine (A), thymine (T), cytosine (C), or guanine (G). The two strands are held together by pairing rules: A always pairs with T, and C always pairs with G. This pairing was a key insight of the Watson and Crick model published in 1953, building on earlier chemical analyses showing that the proportions of bases in DNA followed predictable ratios.¹

One detail that matters a lot for sequencing is directionality. Each strand has a chemical “head” end (called 5′) and a “tail” end (called 3′), and the two strands run in opposite directions. This isn’t just a structural curiosity. When a single strand of DNA is threaded through a tiny pore, its orientation changes how it moves and how much electrical current flows around it. Experiments threading single-stranded DNA through a protein pore showed that a strand entering tail-first moves about twice as slowly as one entering head-first, and the electrical current flowing past the two orientations differs by roughly 30%.² That asymmetry turns out to be useful: it is one of the physical principles behind nanopore sequencing, which we’ll get to later.

Breaking DNA Into Fragments

A human genome contains over three billion base pairs. No sequencing method reads the whole thing in one shot. Instead, the DNA has to be broken into smaller fragments first, and how you break it matters. There are three common approaches: physical shearing (sonication, which uses sound waves), enzymatic digestion (using proteins that cut DNA at specific sequences), and tagmentation (using an enzyme called transposase that simultaneously cuts the DNA and attaches adapter sequences needed for sequencing).³

Enzymatic cutting with restriction enzymes is one of the oldest tools in molecular biology, and it is beautifully precise. Each restriction enzyme recognizes a specific short DNA sequence and cuts there. The enzyme EcoRV, for instance, cuts specifically within the six-letter sequence GATATC.⁴ Some enzymes recognize longer, rarer sequences and therefore cut DNA into much bigger pieces. NotI, for example, recognizes eight letters (GCGGCCGC) and makes far fewer cuts across a genome.⁵ Scientists pick enzymes strategically depending on the fragment size they need.

Sorting Fragments by Size With Gel Electrophoresis

Once DNA is cut into fragments, you often need to separate those fragments by size before you can read them. The workhorse method for this is gel electrophoresis. You load DNA samples into a slab of agarose gel (a jelly-like material), then apply an electric field across it. Because DNA’s backbone carries a negative charge, fragments migrate toward the positive electrode. Smaller fragments slip through the gel’s pores more easily and travel farther; larger ones lag behind. The distance a fragment travels is inversely related to the logarithm of its molecular weight, so you can estimate a fragment’s size by comparing it to a set of known standards run alongside your sample.⁶

This may sound like a crude sieve, but it is remarkably informative. By staining the gel with a dye that glows under ultraviolet light, you can see bands representing fragments of different sizes. A forensic lab might use this to check whether PCR amplification worked before moving on to more precise analysis. A researcher cloning a gene might cut the desired band right out of the gel and purify the DNA from it. Gel electrophoresis doesn’t tell you the sequence of letters in a fragment, but it tells you how big the fragment is and whether your sample looks clean, making it an essential first step.

Sanger Sequencing and Reading a Chromatogram

For decades, the gold standard for actually reading the letters in a DNA fragment was Sanger sequencing, named after Frederick Sanger, who published the method in 1977. The key insight was adding special modified nucleotides, called dideoxynucleotides, to a DNA-copying reaction. Normal nucleotides let the growing chain keep extending, but a dideoxynucleotide acts as a chain terminator: once it gets incorporated, the chain stops.⁷ By running many copies of the reaction simultaneously, you end up with fragments of every possible length, each ending at a different position in the sequence. Sorting those fragments by size reveals the order of bases.

The method became far more practical when each of the four dideoxynucleotides was tagged with a different fluorescent dye, so all four reactions could run in a single tube and a laser could read the colors as fragments passed through a detector.⁸ The output is a chromatogram: a graph with colored peaks, each peak representing one base in the sequence. A clean chromatogram shows tall, evenly spaced, single-color peaks. In practice, though, chromatograms can be messy. Problems during PCR amplification can produce noisy peaks, artifacts, or confusing double peaks that might represent a true genetic variant (heterozygosity), a mixed sample, or simply a technical error.⁹ Learning to tell these apart is one of the trickier skills in a genetics lab.

Sanger sequencing remains highly accurate and is still widely used for confirming specific mutations, reading individual genes, or validating results from newer methods. Its main limitation is throughput: a single Sanger run reads only about 700 to 1,000 bases, and scaling up to cover entire genomes is slow and expensive.

Next-Generation Sequencing and Short Reads

The throughput problem drove the development of what’s now called next-generation sequencing. Rather than reading one fragment at a time, these platforms read millions of fragments simultaneously. The DNA is broken into short pieces, adapter sequences are attached, and each fragment is amplified into a tiny cluster on a surface. As new bases are added during a copying reaction, the machine detects a signal (usually a flash of light) for each base at each cluster, all at once. This massively parallel approach produces enormous quantities of data in a single run.

The trade-off is that each individual “read” is short, typically 100 to 300 bases. That means you end up with millions of small puzzle pieces that have to be assembled, either by mapping them onto a known reference genome or by fitting overlapping fragments together without a reference. Researchers have demonstrated that short reads alone can be assembled into human genomes, achieving contiguous stretches (called contigs) of several thousand bases and larger scaffolds of hundreds of thousands of bases.¹⁰ Software designed for this assembly task identifies segments shared by multiple reads and aligns them using sophisticated algorithms.¹¹

Short-read platforms produce highly accurate individual reads, making them the go-to choice for detecting small mutations such as single-letter changes or short insertions and deletions. They struggle, however, with repetitive regions of the genome, where the short fragments can’t be placed unambiguously. Imagine trying to reconstruct a jigsaw puzzle from pieces that are all the same shade of blue: if they come from a stretch of repetitive DNA, they could fit in multiple places.

Long-Read Sequencing and Why Read Length Matters

The limitations of short reads pushed the field toward “third-generation” platforms that read much longer stretches of DNA in a single pass, sometimes tens of thousands of bases or more. Two major approaches emerged: single-molecule real-time (SMRT) sequencing and nanopore sequencing.

SMRT sequencing, developed by Pacific Biosciences, watches a single DNA polymerase enzyme as it copies a template strand in real time. The polymerase sits at the bottom of a tiny optical well called a zero-mode waveguide, a structure so small that only the light from the very bottom reaches the detector. Each time the polymerase adds a fluorescently labeled nucleotide, the instrument records its color. This approach produces long reads with less bias related to GC content (regions rich in G and C bases that can cause problems for other methods) and can also detect chemical modifications on the DNA bases directly.¹² Historically, one challenge with SMRT sequencing was inefficient loading of DNA into the waveguides, but newer approaches using nanopore-equipped waveguides have improved loading efficiency by five orders of magnitude and made it nearly independent of fragment length.¹²

Nanopore sequencing takes a different approach. A single strand of DNA is threaded through a protein pore embedded in a membrane. As each base passes through the pore, it partially blocks the flow of ions, changing the electrical current in a way that depends on which bases are inside the pore at that moment.¹³ Computational models then translate these current signals into a base sequence. The method’s appeal is that it can read extremely long fragments, needs very little DNA, and works on a device small enough to fit in your pocket. Researchers are even developing models that can predict current signals from chemically modified or non-standard bases, broadening what nanopore platforms can detect.¹⁴

Long reads solve the repetitive-region problem that plagues short reads, because a single fragment can span the entire repetitive stretch and anchor itself in unique sequence on either side. They also make it easier to detect large structural rearrangements in a genome. The downside has been higher per-read error rates, though this gap has been closing rapidly.

How Errors Creep In and How They Get Caught

No sequencing method is perfect, and understanding where errors come from is part of reading DNA fragments competently. Short-read platforms have very low per-read error rates, but they suffer from systematic errors that tend to recur at specific positions in the genome, and stacking up more reads at those positions doesn’t always fix the problem.¹⁵ Long-read platforms, by contrast, make errors more randomly, meaning that reading the same spot about 37 times is typically enough to smooth them out.¹⁵ The two error profiles are complementary, which is why many large projects use both technologies together.

Another common source of trouble is chimeras: artificial sequences formed when two unrelated DNA fragments get fused during the PCR amplification step. Chimeras look like real sequences but are actually stitched-together artifacts. This is a particular headache in studies that survey microbial communities, where you want to count how many different species are present. A chimera that blends sequence from two species could be mistaken for a third species that doesn’t actually exist. Multiple software tools have been developed to detect and remove chimeras, with some working by comparing suspicious sequences against a reference database and others flagging low-abundance sequences that look like recombinations of more common ones.¹⁶ The interplay between GC content, chimera formation, and sequencing bias is an active area of study, with researchers testing various data-processing pipelines to see which combination delivers the most accurate picture of a real community.¹⁷

From Raw Reads to Useful Answers

Generating sequence data is only half the job. Turning millions of short reads or thousands of long reads into meaningful biological findings requires a substantial computational pipeline. In clinical settings, the goal is often to identify “variants,” places where a patient’s DNA differs from the reference genome. This means aligning reads to the reference, calling out differences, and then determining which differences are biologically meaningful and which are noise. Best-practice guidelines for clinical variant calling address everything from the initial alignment step through to final validation, with different strategies recommended for inherited disorders versus cancer, where mutations occur in only a fraction of cells.¹⁸

The software matters as much as the chemistry. Genome assemblers, variant callers, and quality-filtering tools each shape the final result. A read that maps ambiguously might be placed in the wrong location, creating a false variant. A chimeric read that slips past the filter might look like a novel mutation. Clinicians reviewing sequencing results are expected to validate important findings with an independent method, often Sanger sequencing, precisely because its error profile is different from that of next-generation platforms.

Forensic DNA Profiling

One of the highest-profile applications of DNA fragment reading is forensic identification. Rather than sequencing an entire genome, forensic labs focus on short tandem repeats (STRs): stretches of DNA where a short motif (say, AGAT) repeats a variable number of times. Different people carry different numbers of repeats at each STR locus, and examining enough loci produces a profile that is statistically unique. The fragments are amplified by PCR, labeled with fluorescent dyes, and separated by capillary electrophoresis, a refined version of the gel technique that uses a thin tube instead of a slab and offers high precision in sizing.¹⁹ The result is an electropherogram, a graph of peaks where each peak’s position indicates the size of the STR fragment and therefore the number of repeats. Forensic analysts read these graphs in much the same way that geneticists read Sanger chromatograms, watching for artifacts, stutter peaks, and signals from mixed samples.

Environmental DNA and Biodiversity Surveys

You don’t always need to have an organism in hand to read its DNA. Environmental DNA (eDNA) refers to genetic material shed by organisms into soil, water, or sediment. By collecting a water sample from a harbor or a scoop of lake-bottom mud, filtering out the DNA, amplifying target gene regions, and sequencing the resulting fragments, researchers can detect which species are present without ever seeing them. This approach, called eDNA metabarcoding, has proven useful for spotting invasive species early. Surveys of marine sediment and water have detected numerous non-indigenous species, including several that were newly introduced to the area.²⁰

Long-read sequencing is beginning to improve this work. Because short-read metabarcoding relies on short amplified fragments, it sometimes can’t distinguish between closely related species whose short barcode regions are too similar. Nanopore-based long-read metabarcoding of marine vertebrates has shown that longer amplicons improve species-level identification, expanding the molecular toolbox for biodiversity monitoring.²¹ The idea that you can sequence an ocean’s inhabitants from a bottle of seawater was science fiction not long ago. It is now routine enough that conservation agencies are starting to adopt it.

Reading Fragments From the Distant Past

Ancient DNA pushes the limits of fragment reading. DNA left in bones, teeth, or sediment degrades over time through well-characterized chemical processes. The strands break into short pieces, largely because purines (the A and G bases) fall off the backbone and cause strand breaks. On top of that, cytosine bases lose an amino group (a process called deamination), turning them into uracil, which reads as thymine during sequencing. These C-to-T changes pile up at the ends of ancient fragments, creating a distinctive damage pattern.²² Analysis of Neandertal DNA confirmed this: substitutions from miscoded cytosine residues were dramatically overrepresented and clustered at fragment ends, while other substitutions were rare.²³

Paradoxically, this damage signature is now used as a stamp of authenticity. When researchers extract DNA from ancient sediment, they expect to see short fragments and increasing rates of C-to-T damage at fragment ends in older layers. In a study of lake sediment from a Ugandan tropical forest spanning about 2,200 years, fragment lengths ranged from roughly 127 to 175 base pairs and declined with depth, while the characteristic end-damage increased with age.²⁴ If an ancient sample lacks this damage pattern, it is probably contaminated with modern DNA. The fragments are tiny and battered, but the same sequencing methods used on fresh DNA can read them, and the damage itself tells a story about how long the DNA has been sitting in the ground.

When Fragment Size Tells You More Than Sequence

Not every application of fragment reading focuses on decoding the letter sequence. Sometimes the size distribution of DNA fragments is itself the information you want. In prenatal screening, for instance, fragments of fetal DNA circulating in a pregnant person’s blood are shorter on average than fragments from the mother’s own cells. By analyzing the size profile of cell-free DNA, clinicians can enrich for the fetal signal without needing an invasive procedure. In cancer diagnostics, a similar principle applies: tumor-derived DNA fragments in the bloodstream tend to be shorter than fragments from healthy tissue, so fragment-length analysis can help detect or monitor tumors from a simple blood draw.

Even in ancient DNA work, the fragment size distribution carries information beyond what any single sequence read provides. A steep decline in average fragment length with sample depth, correlated with increasing chemical damage, helps researchers build models of DNA decay rates in different environments. The fragments are not just vehicles for sequence data. Their physical characteristics are data too.