Sanger Sequencing: Reading DNA Code Explained

Sanger sequencing reads DNA by building copies of it that stop at a known letter, then sorting those copies by length. The method, published by Frederick Sanger and colleagues in 1977, turned the order of A, C, G, and T from something chemists could only infer into something a laboratory could read off a gel. It underpinned the Human Genome Project, and nearly five decades later it remains a routine check in many labs. This article explains how it works, what it can and cannot do, and where it sits next to modern sequencers.

Key Takeaways

• Sanger sequencing copies a DNA template with a polymerase that occasionally inserts a nucleotide that cannot be extended, so each copy stops at a known base
• Sorting the stopped copies by length reveals the sequence, one letter per size step
• The 1977 method replaced the same group's 1975 "plus and minus" approach and was described as faster and more accurate
• The Human Genome Project relied on Sanger sequencing and, by 2003, had covered about 92 percent of the genome
• Today the method reads a few hundred bases per reaction at very high accuracy, which suits confirmation work and small targets rather than whole genomes

Why reading DNA was so hard before 1977

By the mid-1950s, researchers knew that DNA was a double helix and that its four bases carried information (the path to that structure is told in The History of DNA: From Miescher to the Double Helix). What they could not do was read the order of those bases. Knowing that a gene exists is like knowing a book is on the shelf; the sequence is the text inside it.

Sanger had already solved a harder-looking problem on the protein side. He worked out the exact order of amino acids in insulin by 1955, a decade-long effort that earned him the 1958 Nobel Prize in Chemistry, as Wikipedia's biography of Sanger summarizes. Insulin, the hormone he sequenced, is explained in Insulin and Diabetes Explained. Proteins use 20 building blocks, which gives each position a distinctive chemical signature. DNA uses four chemically similar ones, and a single gene can run to thousands of positions. Telling a long run of A, C, G, and T apart by chemistry alone is like trying to read a page written in one font, one size, with no spaces and no punctuation.

The strategy that worked for insulin did not scale. Break the molecule into pieces, identify each piece, and reassemble the order. For a protein of 51 amino acids that is feasible. For a stretch of DNA thousands of bases long, the fragments are too similar and the puzzle has too many pieces.

Sanger's route into DNA went through a different idea: let an enzyme do the copying and make the copy reveal the sequence. In 1975, working with Alan Coulson, he described the "plus and minus" method, which used DNA polymerase (the enzyme cells use to copy DNA) to synthesize new strands from a short starting piece called a primer. Two years later the same group published a refinement that they reported to be faster and more accurate, and that refinement became the standard.

How chain termination works, one step at a time

The 1977 paper, by Sanger, Nicklen, and Coulson, described the new approach as resembling the earlier plus and minus method but using chemically altered versions of the normal nucleotide building blocks. The abstract on PubMed Central says these analogues, 2',3'-dideoxy and arabinonucleoside forms, act as chain-terminating inhibitors of DNA polymerase, and that the technique was applied to the bacteriophage ΦX174 and was faster and more accurate than the plus and minus methods.

The key chemistry is a missing hook. DNA polymerase (whose normal job is described in DNA Replication and Genetics Explained) builds a strand by attaching each incoming nucleotide to a small chemical group, a hydroxyl, on the 3' end of the strand so far. A dideoxynucleotide (ddNTP) lacks that hydroxyl. Once the polymerase adds one, there is nothing for the next nucleotide to attach to, and that copy stops.

Think of a train assembled one car at a time, where each new car couples to the hook on the back of the last. Normal cars have a hook at both ends. A ddNTP is a car with no rear hook. Whenever one gets coupled on, the train ends there, and you know exactly what kind of car sits at the end.

The ingredients in the reaction

Component	What it does
Template DNA	The strand to be read; the polymerase copies it
Primer	A short piece of DNA that pairs with the template and marks the starting point
DNA polymerase	The enzyme that extends the primer by adding nucleotides one at a time
Normal nucleotides (dNTPs)	The building blocks that let the chain keep growing
Chain-terminating nucleotides (ddNTPs)	Rare, modified building blocks that stop the chain when incorporated

The ddNTPs are present in small amounts compared with the normal nucleotides. That ratio matters. If terminators were common, every copy would stop within a few bases. Because they are rare, each copy keeps growing until, by chance, it picks up a terminator at some position. Across millions of template molecules in the tube, every position gets hit by a terminator in some molecule.

A worked example: reading eight letters

Suppose the new strand being built should read ATCGGTAC from its start. The template strand it copies reads 3'-TAGCCATG-5'. In the original method, the experimenter ran four separate tubes, each containing all four normal nucleotides plus one kind of terminator. Here is what each tube produces, in fragment lengths counted in bases:

Tube	Terminator	Fragment lengths that appear	Why
1	ddATP (stops at A)	1 and 7	Positions 1 and 7 of ATCGGTAC are A
2	ddTTP (stops at T)	2 and 6	Positions 2 and 6 are T
3	ddCTP (stops at C)	3 and 8	Positions 3 and 8 are C
4	ddGTP (stops at G)	4 and 5	Positions 4 and 5 are G

Now sort all eight fragments by length and read the terminator for each size: 1 is A, 2 is T, 3 is C, 4 is G, 5 is G, 6 is T, 7 is A, 8 is C. The sequence spells out ATCGGTAC. Real reactions produce hundreds of lengths rather than eight, but the logic is identical.

Sorting the fragments by length

The sorting step relies on gel electrophoresis, which pushes DNA through a porous gel with an electric field. Shorter fragments slip through the pores faster than longer ones, so after a run the fragments line up by size like runners strung out along a track by speed. The gel in the original method had to separate fragments that differ in length by a single nucleotide, which is the hard part: the summary of the method on Wikipedia's Sanger sequencing page names resolving one-base differences as the main obstacle that limits how long a read can be.

In the first version, the polymerase was supplied with a radioactive label so the fragments would expose X-ray film, and researchers read the film by eye, lane by lane from the shortest fragment to the longest. Four lanes side by side, one per base, produced a ladder pattern that the reader climbed from bottom to top. Each rung gave one letter.

From four tubes to one: how automation changed the method

The manual approach worked, but it was slow. Reading film by hand is tedious, and four reactions per sequence multiplies the pipetting. The next step was to replace radioactivity with color.

In 1985 and 1986, Lloyd Smith, Leroy Hood, and colleagues described fluorescently labeled primers and terminators, work cited on the Sanger sequencing page and published in Nature in 1986 as "Fluorescence detection in automated DNA sequence analysis." If each of the four terminators carries a different dye, all four reactions can run in a single tube, and a laser and detector can identify the color of each fragment as it passes. Applied Biosystems introduced a commercial automated instrument using slab gels and fluorescent labels in March 1987, according to the history cited on that page.

Slab gels were later replaced by capillary electrophoresis, in which fragments travel through hair-thin tubes filled with a polymer. Capillary arrays allowed dozens or hundreds of samples to run in parallel and made reading automatic. This combination, dye-labeled terminators plus capillary arrays, is the form of Sanger sequencing most laboratories use today.

The output is no longer a film but a chromatogram, a plot of colored peaks. Each peak is one base, the color says which base, and the height says how strong the signal was. Software called a base caller turns the trace into a text sequence and attaches a quality score to each position.

What a Sanger read looks like and how reliable it is

A single Sanger reaction produces one read: a continuous stretch of sequence from the primer onward. The summary on the Sanger sequencing page gives typical read lengths of 500 to 600 bases with a practical maximum around 800, and it states direct sequencing is limited to roughly 300 to 1,000 nucleotides per reaction. A university core facility at the University of Texas Medical Branch advertises reads of at least 750 bases at 99.9 percent accuracy or better, with quality scores supplied for each base. A 2008 review in Nature Biotechnology by Shendure and Ji, cited on the same Wikipedia page, puts accuracy at about 99.99 percent.

Property	Typical value	Source
Read length	500 to 600 bases typical; about 800 maximum	Wikipedia summary of the literature
Accuracy	About 99.9 to 99.99 percent	UTMB core facility; Shendure and Ji (2008)
Reaction price at a university core	$5.50 to $7.50 per reaction	UTMB core facility
Turnaround at that core	48 hours	UTMB core facility
Template needed	A PCR product of 1 kilobase or less at 25 ng/µl, or a plasmid at up to 100 ng/µl	UTMB core facility

Prices vary by institution and volume, so treat the dollar figures as an example of the scale rather than a standard rate.

Where reads go wrong

Quality is not uniform across a read. The first 15 to 40 bases after the primer are often low quality because of how the primer binds, and quality tends to deteriorate after about 700 to 900 bases, according to the limitations summarized on the Sanger sequencing page. Sequence-dependent differences in how the polymerase incorporates each dye also cause uneven peak heights and shapes. The UTMB facility asks users to flag templates that are more than 70 percent GC or AT, because the chemistry has to be adjusted for them.

A more interesting problem appears when the template is not a single molecule type. If a sample contains two versions of a sequence, as it does when a person carries one normal and one variant copy of a gene, the chromatogram at that position shows two overlapping peaks. Readers learn to recognize these. The same effect makes mixtures of unrelated sequences unreadable, which is why the method is usually applied to a purified PCR product or a cloned plasmid rather than to a raw sample.

The PCR step that prepares most templates is covered in the PCR article linked at the end of this page.

The first genomes: from ΦX174 to the human mitochondrion

Sanger sequencing arrived in a decade in which sequencing itself was in flux, and a few details of the early history are commonly told wrongly. The table below lists the dated milestones from the sources used in this article.

Year	Milestone
1958	Sanger receives the Nobel Prize in Chemistry for the structure of insulin
1975	Sanger and Coulson publish the plus and minus method
1976	Fiers and colleagues complete the 3,569-nucleotide RNA genome of bacteriophage MS2
1977	Sanger's group reports about 5,375 nucleotides of bacteriophage ΦX174 DNA
1977	Sanger, Nicklen, and Coulson publish the chain-terminator method
1980	Sanger shares the Nobel Prize in Chemistry with Walter Gilbert and Paul Berg
1981	Human mitochondrial DNA sequenced: 16,569 base pairs
1982	Bacteriophage lambda DNA sequenced: 48,502 base pairs
1986	Fluorescent dye-based automated sequencing described
1987	First commercial automated sequencer introduced
1990 to 2003	Human Genome Project

Two corrections deserve space. First, the famous ΦX174 genome of 1977 was not read with the chain-terminator method. The Nature paper's abstract describes a sequence of approximately 5,375 nucleotides determined with the "plus and minus" method. The chain-terminator method appeared later that year and used ΦX174 as a test case. Second, ΦX174 was the first complete DNA genome, but not the first complete genome of any kind: the RNA genome of the bacteriophage MS2, 3,569 nucleotides long, was finished in 1976 by Walter Fiers's group, with the final paper in Nature on April 8, 1976. Current reference sequences for ΦX174 list 5,386 nucleotides, so the figure moved slightly as the sequence was corrected over the years.

The ΦX174 sequence produced a surprise. The abstract of the 1977 paper reports that two pairs of genes are encoded by the same stretch of DNA read in different reading frames. Overlapping genes had not been expected, and the finding showed that a small genome can pack information more tightly than anyone had assumed. It also showed that reading sequence directly turns up discoveries that no amount of indirect chemistry would have predicted.

By 1981 the method scaled to the 16,569 base pairs of the human mitochondrial genome and by 1982 to the 48,502 base pairs of bacteriophage lambda. Those are tiny by today's standards, but lambda is nine times the size of ΦX174, and each step proved that the approach could keep pace with larger targets. Phages are viruses that infect bacteria, and many of the early genomes were phages for a practical reason: small genome, easy to grow, easy to purify. For how such viruses reproduce, see How Viruses Work: From Infection to Replication.

How Sanger sequencing built the Human Genome Project

The Human Genome Project ran from 1990 to 2003. In April 2003, the consortium announced an essentially complete human genome sequence, and according to the National Human Genome Research Institute, the project used Sanger DNA sequencing, refined through several technical innovations. The 2003 sequence covered 92 percent of the genome, up from an earlier 2000 draft that covered about 90 percent, and it still had fewer than 400 gaps. Complex and highly repetitive regions were the hardest to read. The Telomere-to-Telomere consortium filled the remaining gaps in March 2022.

The method that made this possible is the one described above, run at enormous scale. A genome is far longer than a single read, so the strategy is to break it into overlapping pieces, sequence each piece, and use computers to assemble the overlaps into a continuous sequence. A single read covers a few hundred bases; the human genome has about three billion. That gap in scale is why the project needed industrial automation, and why the capillary instruments described earlier mattered.

The cost story is easier to see in a table. These are estimates from NHGRI, and the agency cautions that figures from different sources are not always directly comparable because they count different things, such as labor, equipment, and data analysis.

Period	Estimated cost	What it covers
1990 to 2003	About $2.7 billion in US funding	The whole project, including mapping and technology development
2003 (hypothetical)	About $50 million	A second reference genome with the methods then available
2006	About $14 million	An individual draft genome
Mid-2015	Just above $4,000	An individual draft genome with next-generation sequencing
Late 2015	Below $1,500	The same approach, draft quality

Source: NHGRI, "The Cost of Sequencing a Human Genome", last updated November 1, 2021.

The drop from the first reference to a 2015 genome came largely from methods that read millions of short fragments at once rather than one reaction at a time. For what the project found, and how the numbers changed afterward, see The Human Genome Project Explained and The Human Genome Revolution: From $3 Billion to $200 and What We Found.

The scale gap: The first complete DNA genome, ΦX174 in 1977, held about 5,375 nucleotides. The human genome is more than 500,000 times larger. Sanger sequencing bridged that gap because automation multiplied the number of reactions running at once, not because any single read got longer.

Where Sanger sequencing is still used

If newer methods are so much cheaper per base, why does anyone still run Sanger reactions? Because cost per base is not the only cost. For a small target, a single reaction at a core facility costs a few dollars, takes two days, and returns a clean text file with no computational pipeline to build.

Verifying DNA constructs. Researchers who build a plasmid, edit a gene, or amplify a region by PCR typically sequence the result to confirm that it contains what they intended. The UTMB service accepts exactly these templates: plasmids and PCR products. It is one of the most common everyday uses of the method.

Targeted clinical confirmation. Clinical laboratories use Sanger to confirm variants found by broader tests and to test a known family variant in relatives. How much confirmation is needed is an active debate (see the next section).

Pathogen work on a defined target. The method has been used in protocols for sequencing the SARS-CoV-2 S gene, described by Daniels and colleagues in 2021, and in the norovirus outbreak surveillance network CaliciNet, described by Vega and colleagues in 2011, both listed among applications on the Sanger sequencing page. Both look at a defined piece of a genome rather than the entire thing.

Fragment sizing. The same capillary instruments that read sequence can also measure the length of labeled DNA fragments, which is how short tandem repeat (STR) analysis is done. The Sanger sequencing page lists STR analysis among applications of the instrument platform. Forensic DNA profiling, including the work that has overturned wrongful convictions, depends on repeat lengths measured this way, not on reading the sequence base by base.

For how personalized genetics ties these tests to treatment decisions, see Personalized Medicine & AI-Driven Genomics.

Does a next-generation result still need a Sanger check?

The old rule of thumb was simple: any variant found by next-generation sequencing (NGS), the massively parallel methods that replaced Sanger for large projects, had to be confirmed by Sanger before it went into a patient report. The reasoning was that NGS makes more random errors. Two studies show why this is no longer a settled question.

A 2016 study by Ambry Genetics in The Journal of Molecular Diagnostics examined 20,000 hereditary cancer panels covering 47 genes. The company reported that without Sanger confirmation, 1.3 percent of NGS variant calls would have been false positives, mostly in complex genomic regions, according to its press release. Filtering them out by quality scores alone would have missed about 2 percent of real variants, according to a summary of the same work.

A 2019 study led by Stephen Lincoln, with co-authors including Justin Zook, Marc Salit, and Matthew Lebo, in the same journal looked at data from two clinical laboratories. It combined five Genome in a Bottle reference samples and more than 80,000 clinical specimens, about 200,000 variant calls with orthogonal confirmation data, including 1,662 false positives. The authors found that a combination of quality criteria flagged all of the false positives while minimizing the number of true variants sent for confirmation, and that each laboratory should set and validate its own thresholds. The full text is on PubMed Central.

Put together, the two studies point to a middle position. High-quality calls may not need confirmation if a laboratory has validated its own cutoffs, but low-quality calls and variants in difficult regions still do. Sanger sequencing keeps a role here not because it is fast but because it works by a different mechanism, so its errors are unlikely to match the errors of the first method.

Where Sanger sequencing falls short

Every limitation of Sanger sequencing follows from one fact: each reaction reads one stretch of DNA from one primer.

Short reads. A read of a few hundred bases cannot span a long repeat. NHGRI names complex and highly repetitive DNA as the hardest material to sequence, and it is a main reason the 2003 human sequence still had gaps.

One target at a time. To read a gene of 5,000 bases, a laboratory designs several primers and runs several reactions. To read an entire genome, it would need millions. Methods that read huge numbers of fragments in parallel make this trade-off the other way, giving up read length for volume.

Needs a defined template. The reaction needs a primer, so you must already know something about the sequence at the starting point. It cannot discover an unknown piece of DNA from scratch without extra preparation.

Mixed signals. As described earlier, mixtures of templates blur the peaks. A variant present in a small fraction of molecules, such as a cancer mutation in a tissue sample that is mostly normal cells, can hide in baseline noise.

Structural changes are hard to see. When a laboratory amplifies a region and reads it, it sees only the piece it amplified. A large deletion, a duplication, or a rearrangement may fall outside the read or be masked by a normal copy on the other chromosome. This is why diagnostic workflows pair Sanger with methods designed to count copies and detect rearrangements.

The comparison with parallel methods looks like this.

Feature	Sanger sequencing	Next-generation sequencing
Reads per run	One per reaction	Millions or more in parallel
Read length	Hundreds of bases	Generally shorter per read
Best at	Small targets, confirmation, construct checks	Whole genomes, exomes, panels, expression studies
Setup	Primer for the target; minimal software	Library preparation and a computational pipeline
Detects low-frequency variants	Poorly	Better, with enough depth
Typical cost structure	A few dollars per reaction	Lower cost per base, higher cost per project

The same principles of parallel reading carry into techniques such as Single-Cell RNA Sequencing, which assigns millions of short reads back to individual cells, and Liquid Biopsy, which looks for tumor DNA in blood.

Common misconceptions about Sanger sequencing

"Sanger sequencing is obsolete." It is not the right tool for a genome, but it is the right tool for many small questions. A core facility that charges about $5.50 to $7.50 per reaction and returns results in 48 hours is hard to beat for checking a plasmid.

"The ΦX174 genome in 1977 proved the chain-terminator method." The sequence published in Nature that year used the plus and minus method. The chain-terminator paper used ΦX174 as a test, which is a related but different claim.

"Forensic DNA identification is Sanger sequencing." Standard forensic profiling measures the lengths of short repeats, not the order of bases, though it runs on the same family of capillary instruments.

"A high accuracy figure means every base is right." An accuracy near 99.99 percent means about one error in 10,000 bases on average, and the errors concentrate in the first few dozen and last few hundred bases of a read. A 700-base read will usually be clean in the middle and unreliable at both ends. Good practice is to read in both directions and compare.

"The Human Genome Project finished the genome." The 2003 sequence covered about 92 percent and left gaps in repetitive regions. A complete, gap-free human genome came with the Telomere-to-Telomere consortium in 2022.

"Sanger sequencing needs a slab gel and radioactive film." That describes the 1970s version. Modern instruments use fluorescent dyes, a laser, and capillary tubes, and the operator receives a text file and a quality score for each base.

FAQ

Q: How does Sanger sequencing work in simple terms? A: A polymerase copies a DNA template, but the reaction includes a small amount of modified nucleotides that stop the copy wherever they are inserted. Each stopped fragment ends at a known base. Sorting the fragments by length, using a gel or a capillary, lets software read the sequence from shortest to longest.

Q: Why is Sanger sequencing still used if newer methods exist? A: It is inexpensive and fast for small targets. A reaction at a university core facility can cost $5.50 to $7.50 and return in two days, and the output is a simple sequence file. It also works by a different mechanism than parallel methods, so it can serve as an independent check on a result.

Q: What is the difference between Sanger and next-generation sequencing? A: Sanger sequencing produces one read per reaction, typically several hundred bases long. Next-generation methods read huge numbers of fragments at once, each generally shorter, which makes them far cheaper for entire genomes. Sanger suits small targets and confirmation; parallel methods suit discovery.

Q: How long a stretch of DNA can Sanger sequencing read? A: Typical reads are 500 to 600 bases, with a practical maximum near 800, and direct sequencing is limited to roughly 300 to 1,000 nucleotides per reaction. Quality is lowest in the first 15 to 40 bases and tends to fall after 700 to 900 bases.

Q: Was the Human Genome Project done with Sanger sequencing? A: Yes. NHGRI states that the project used Sanger DNA sequencing, improved through several technical innovations. The 2003 sequence covered 92 percent of the genome, and the remaining gaps were filled by the Telomere-to-Telomere consortium in 2022.

Q: Can Sanger sequencing find every kind of mutation? A: No. It reads single-base changes and small insertions or deletions in the region that was amplified. Large deletions, duplications, and rearrangements may be missed, and low-frequency variants can disappear into background noise, so clinical workflows pair it with other tests.

Q: Who invented Sanger sequencing and when? A: Frederick Sanger, working with Alan Coulson and a colleague named Nicklen, developed it in the 1970s. The plus and minus method appeared in 1975 and the chain-terminator method in 1977. Sanger shared the 1980 Nobel Prize in Chemistry with Walter Gilbert and Paul Berg, and died on November 19, 2013.

The takeaway

The cleverness of Sanger sequencing is that it never reads a base directly. It arranges for the chemistry to leave a record: a pile of fragments whose lengths encode the order and whose last letters encode the identity. Once that record exists, reading it is a matter of sorting, a problem that gels, lasers, and capillary tubes solve well.

That idea explains both its decline and its persistence. For anything that needs millions of reads, it was displaced by methods that sort fragments in parallel. For anything small, specific, and important, a short, accurate, independently produced read remains difficult to improve on. When a genetic result will affect a person's care, the question has shifted from whether Sanger is still relevant to which results are reliable enough to skip it, and the best evidence says each laboratory has to answer that for itself.

References

Sanger, F., & Coulson, A.R. (1975). "A Rapid Method for Determining Sequences in DNA by Primed Synthesis with DNA Polymerase." Journal of Molecular Biology, 94(3), 441-448.

Sanger, F., Air, G.M., Barrell, B.G., et al. (1977). "Nucleotide sequence of bacteriophage φX174 DNA." Nature, 265, 687-695.

Sanger, F., Nicklen, S., & Coulson, A.R. (1977). "DNA Sequencing with Chain-Terminating Inhibitors." Proceedings of the National Academy of Sciences, 74(12), 5463-5467. Abstract on PubMed Central

Fiers, W., Contreras, R., Duerinck, F., et al. (1976). "Complete nucleotide sequence of bacteriophage MS2 RNA: primary and secondary structure of the replicase gene." Nature, 260, 500-507.

Smith, L.M., Sanders, J.Z., Kaiser, R.J., et al. (1986). "Fluorescence detection in automated DNA sequence analysis." Nature, 321, 674-679.

Shendure, J., & Ji, H. (2008). "Next-generation DNA sequencing." Nature Biotechnology, 26(10), 1135-1145.

International Human Genome Sequencing Consortium. (2004). "Finishing the euchromatic sequence of the human genome." Nature, 431, 931-945.

Lincoln, S.E., Truty, R., Lin, C.-F., Zook, J.M., et al. (2019). "A Rigorous Interlaboratory Examination of the Need to Confirm Next-Generation Sequencing-Detected Variants with an Orthogonal Method in Clinical Genetic Testing." The Journal of Molecular Diagnostics, 21(2), 318–329. PubMed Central

National Human Genome Research Institute. Human Genome Project Completion FAQ and The Cost of Sequencing a Human Genome.

---

Read next:

• DNA Structure: Watson, Crick & Franklin's X-ray Crystallography Explained
• PCR Invention: Copying DNA Millions of Times Explained
• Cloning Explained: How Dolly the Sheep Was Made