Material Insights · Understanding the Material World

Genes: DNA, inheritance and DNA technology: The human genome and the tools for reading and copying DNA

All courses / Understanding Science / Module 14: Genes: DNA, inheritance and DNA technology Part 5 of 6

The human genome and the tools for reading and copying DNA

Once DNA's code was known, three tools turned it into a technology: enzymes to cut it, a reaction to copy it and machines to read it. Together they let people read a whole human genome in a day for the price of a phone.

5.1 The human genome has about 3.1 billion base pairs and about 20,000 protein-coding genes, and most of it does not code for protein

The Human Genome Project began in 1990 as an international effort. It reported working drafts in 2001 and, according to NHGRI, was completed in April 2003, at a cost of roughly three billion dollars. NHGRI gives about 3 billion base pairs and an estimated 20,000 to 25,000 genes. That is no more than a roundworm of 1,000 cells. People are not complex because of more genes. They have more regulation, more ways to splice each gene, and more interactions.

Only about 1 to 1.5 per cent of the genome codes for protein. Another share is genes for functional RNAs and the switches that control reading. About half is repeats, including copies of mobile elements that spread through the genome over evolutionary time. How much of the rest is useful is still debated. Comparing mammals suggests that something like a tenth of the genome is held steady by selection.

The 2003 version was called complete, but about 8 per cent, mostly repeats, was unresolved, because short reads could not be assembled across them. In 2022 the Telomere-to-Telomere consortium published T2T-CHM13, the first gapless sequence of a human genome, at 3.055 billion base pairs. It added nearly 200 million bases of new sequence, including 99 predicted protein-coding genes. India's GenomeIndia project, started by the Department of Biotechnology in 2020, finished sequencing 10,000 genomes from across the country in 2024 and released the data in January 2025. It matters because many Indian communities have married mostly within the group for centuries and carry variants rare elsewhere.

Two unrelated people are about 99.9 per cent the same. That sounds small but 0.1 per cent of 3.1 billion is 3 million differences (calculated), which is plenty to tell every person apart.

Data Worked estimates: what the numbers mean
QuestionWorkingResult
Information in one set of chromosomes3.1 x 10^9 bases x 2 bits (4 letters = 2 bits)6.2 x 10^9 bits / 8 = 775 MB (calculated), a little more than one 700 MB CD
Coding DNA in 20,000 genes20,000 x about 1,400 bases (a protein of about 450 amino acids)28 million bases (calculated)
Share of genome that codes28 million / 3,100 millionabout 0.9% (calculated), consistent with 1 to 1.5%
Differences between two people3.1 x 10^9 x 0.001about 3 million (calculated)

Ratios of this kind are order-of-magnitude checks. Real gene sizes vary widely, and the human average coding length is near 1.4 to 1.5 kilobases.

Source: National Human Genome Research Institute (NIH), A Brief Guide to Genomics, https://www.genome.gov/about-genomics/fact-sheets/A-Brief-Guide-to-Genomics (about 3 billion base pairs; an estimated 20,000 to 25,000 genes, coding on average for about three proteins each; Human Genome Project completed April 2003)

5.2 Restriction enzymes cut DNA at chosen sequences, and cut pieces can be sorted, joined and copied

In the 1950s and 1960s workers noticed that a virus grown in one strain of bacteria did badly in another, as if the second bacterium cut up foreign DNA. Werner Arber explained it by enzymes that cut DNA at particular sequences, Hamilton Smith isolated one in 1970, and Daniel Nathans used such enzymes to map a virus. The three shared the 1978 Nobel Prize in Physiology or Medicine. The cell protects its own DNA by marking its own sites with methyl tags. The node 'Restriction enzymes' tells the story.

A restriction enzyme recognises a short, palindromic sequence, one that reads the same on both strands in opposite directions. EcoRI, from E. coli, recognises GAATTC and cuts each strand between G and A, leaving short single-stranded overhangs on both ends. These 'sticky ends' pair with any other piece cut by the same enzyme. DNA ligase seals the joins. A gene from one organism can therefore be spliced into a plasmid, a small DNA ring that bacteria copy and share. Stanley Cohen and Herbert Boyer did this in 1973 and gave the first recombinant DNA organisms.

Pieces are sorted by gel electrophoresis. DNA is negatively charged, so in an electric field it moves through a jelly of agarose towards the positive electrode, and short pieces move faster than long ones. The pattern of bands is a fingerprint of the cuts.

The first medical product was human insulin made in E. coli, from 1978, which reached the market in 1982. In 1974 leading researchers called a voluntary pause on some experiments, and in 1975 a meeting at Asilomar in California replaced it with safety guidelines, an early case of researchers setting rules for their own work.

Data Worked estimate: how often does a recognition site turn up by chance?
Site lengthNumber of possible sequencesExpected spacing along random DNACuts in a 3.1 billion base genome
4 bases4^4 = 256one site per 256 bases3.1 x 10^9 / 256 = about 12 million (calculated)
6 bases (EcoRI)4^6 = 4,096one per 4,100 basesabout 760,000 (calculated)
8 bases4^8 = 65,536one per 65,500 basesabout 47,000 (calculated)

Real DNA is not random, so counts differ. A 3,000-base plasmid expects 3,000 / 4,096 = 0.7 six-base sites (calculated), so it often has 0 or 1, which is what makes plasmids useful as carriers.

Source: Hewitt, Lyons, Suchocki and Yeh, Conceptual Integrated Science, 2nd edition (Pearson, 2013), ch. 16, 'Genetics'
  • Hewitt, Lyons, Suchocki and Yeh, Conceptual Integrated Science, 2nd edition (Pearson, 2013) · ch. 16, 'Genetics'
  • Standard histories of genetics, as summarised in Encyclopaedia Britannica (Gregor Mendel; Rosalind Franklin; Francis Crick; James Watson; DNA; Meselson-Stahl) and the Nobel Prize site (Physiology or Medicine 1962, 1968, 1978 and 1980; Chemistry 1993) · dates and findings for Mendel (read 1865, published 1866, rediscovered 1900), Sutton and Boveri (1902 to 1903), Morgan (1910), Griffith (1928), Avery, MacLeod and McCarty (1944), Hershey and Chase (1952), Chargaff, Franklin and Gosling (Photo 51, 1952), Watson and Crick (Nature, 25 April 1953), Meselson and Stahl (1958), Crick and others (1961), Nirenberg and Matthaei (1961), Jacob and Monod (1961), Luria and Delbruck (1943), Cohen and Boyer (1973), Sanger (1977), Mullis (1983). Checked in November 2026: Mendel read 8 February and 8 March 1865, printed 1866 (Villanova Mendel text); Photo 51 taken May 1952 and published in Nature on 25 April 1953 beside Watson and Crick (King's College London); Meselson and Stahl at Caltech, PNAS 44 (1958) (Arizona State University Embryo Project); Mullis developed PCR in 1983 at Cetus, Nobel 1993 (Britannica)

5.3 PCR copies a chosen stretch of DNA a billion-fold in an afternoon, and the cost of reading DNA has fallen faster than computing costs

Kary Mullis conceived the polymerase chain reaction in 1983 while working at Cetus. It is a copying machine for DNA. You mix the DNA with two short primers that match the ends of the stretch you want, free nucleotides and a polymerase. Heat the mixture to about 95 degrees C, so the strands separate. Cool to about 55 to 60 degrees, so primers attach at their matches. Warm to 72, so the polymerase builds the second strands. Each cycle takes a minute or two and doubles the number of copies.

The early version lost its enzyme at every heating step. The fix was Taq polymerase, from Thermus aquaticus, a bacterium found in hot springs in Yellowstone, whose enzyme works at 72 degrees and survives 95. Primers of 20 bases are specific, because 4^20 is about a million million (calculated), much larger than a genome of 3 billion letters, so a given 20-base match turns up once by chance at most. Mullis won the 1993 Nobel Prize in Chemistry. PCR is now used for diagnosing infections (the pandemic tests of 2020 used RT-PCR, in which RNA is first copied to DNA), forensic work, ancient DNA, and checking seed lots for GM traits. Because one molecule can be amplified, contamination is a real hazard.

Reading DNA also moved on. Frederick Sanger's method of 1977 stops copies at a chosen base to sort fragments by length. Machines using it, first run in parallel, read the first human genome. From about 2005, 'next-generation' machines read millions of short pieces at once.

NHGRI's data on cost per genome show a fall from over 95 million dollars in 2001 to about 525 dollars in 2022. That is 95 million / 525 = a factor of about 180,000 in 21 years (calculated).

Data Worked estimate: PCR copies, if every cycle doubles and if only part of the molecules copy
CyclesPerfect (2^n)90% efficient (1.9^n)80% efficient (1.8^n)
101,024 (calculated)613 (calculated)357 (calculated)
201,048,576 (calculated)376,000 (calculated)127,000 (calculated)
301,073,741,824 (about 1 billion) (calculated)about 230 million (calculated)about 45 million (calculated)

Real PCR levels off after about 30 to 40 cycles as primers and nucleotides run out.

Source: Hewitt, Lyons, Suchocki and Yeh, Conceptual Integrated Science, 2nd edition (Pearson, 2013), ch. 16, 'Genetics'
Data Worked estimate: how fast has the cost of reading a genome fallen?
StepWorkingResult
Total fall95,000,000 / 525about 181,000 times
Number of halvings2^17.5 is about 181,00017.5 halvings
Time per halving21 years / 17.51.2 years, about 14 months
Computing, for comparisonone halving in about 2 years in the classic Moore's lawonly about 1,400 times in 21 years

Cost per genome in 2001 and 2022 from NHGRI data. The curve fell sharply around 2007 to 2008 when next-generation machines arrived, faster than Moore's law. The numbers reflect sequencing only, not interpretation.

Source: NHGRI DNA Sequencing Costs data, as reproduced in the DataHub article 'The evolution of DNA sequencing costs, 2001 to 2022', https://datahub.io/blog/the-evolution-of-dna-sequencing-costs-insights-from-2001-to-2022 (median cost of a human genome over US$95 million in 2001, US$525 by 2022; source given as NHGRI, genome.gov)
  • Hewitt, Lyons, Suchocki and Yeh, Conceptual Integrated Science, 2nd edition (Pearson, 2013) · ch. 16, 'Genetics'
  • NHGRI DNA Sequencing Costs data, as reproduced in the DataHub article 'The evolution of DNA sequencing costs, 2001 to 2022' · web
  • The Nobel Prize in Chemistry 2020 (Charpentier and Doudna, for the development of a method for genome editing) and 1993 (Mullis, for the invention of the polymerase chain reaction method) · Motivation 'for the development of a method for genome editing', as quoted by the Royal Swedish Academy of Sciences, https://www.kva.se/en/news/nobelpriset-i-kemi-2020-2/; 1993 prize to Mullis for the invention of the polymerase chain reaction method
  • Standard histories of genetics, as summarised in Encyclopaedia Britannica (Gregor Mendel; Rosalind Franklin; Francis Crick; James Watson; DNA; Meselson-Stahl) and the Nobel Prize site (Physiology or Medicine 1962, 1968, 1978 and 1980; Chemistry 1993) · dates and findings for Mendel (read 1865, published 1866, rediscovered 1900), Sutton and Boveri (1902 to 1903), Morgan (1910), Griffith (1928), Avery, MacLeod and McCarty (1944), Hershey and Chase (1952), Chargaff, Franklin and Gosling (Photo 51, 1952), Watson and Crick (Nature, 25 April 1953), Meselson and Stahl (1958), Crick and others (1961), Nirenberg and Matthaei (1961), Jacob and Monod (1961), Luria and Delbruck (1943), Cohen and Boyer (1973), Sanger (1977), Mullis (1983). Checked in November 2026: Mendel read 8 February and 8 March 1865, printed 1866 (Villanova Mendel text); Photo 51 taken May 1952 and published in Nature on 25 April 1953 beside Watson and Crick (King's College London); Meselson and Stahl at Caltech, PNAS 44 (1958) (Arizona State University Embryo Project); Mullis developed PCR in 1983 at Cetus, Nobel 1993 (Britannica)

5.4 Genetic tests read variants as risks and probabilities, and the same accuracy can mean very different things

A genetic test looks for a particular kind of difference. A karyotype counts and examines whole chromosomes. A targeted test looks for one known variant. A panel checks dozens of genes. Exome sequencing reads the protein-coding 1 to 1.5 per cent and genome sequencing reads all of it. Others check cancer cells, drug-handling variants, newborns or ancestry. The principles below hold for all of them and are not medical advice, which belongs with a doctor or a genetic counsellor.

First, a variant is not a diagnosis. Some variants, such as the one behind sickle-cell disease, nearly always produce the trait. Many raise risk only a little, and a large number are of unknown significance. The share of carriers who show a trait is its penetrance. Second, most common conditions depend on many genes plus environment, so a 'polygenic score' moves odds a little and cannot decide an individual case.

Third, the base rate controls how useful a positive result is. A test with 99 per cent sensitivity and 99 per cent specificity sounds nearly perfect. For a condition affecting 1 person in 1,000, though, most positives are false, as the table shows. This is why screening of whole populations needs a follow-up test.

Fourth, DNA is about families. A result for one person tells you about relatives. The field has therefore developed counselling to explain what a result means before and after the test, rules on consent, and laws against misuse by employers or insurers. In India, prenatal tests are covered by the Pre-Conception and Pre-Natal Diagnostic Techniques Act of 1994, which bans their use to choose the sex of a child.

Data Worked estimate: 100,000 people tested, 99% sensitivity and 99% specificity
CaseCondition affects 1 in 1,000Condition affects 1 in 10 (a high-risk family)
People who have it10010,000
True positives (99%)100 x 0.99 = 99 (calculated)10,000 x 0.99 = 9,900 (calculated)
People who do not99,90090,000
False positives (1%)99,900 x 0.01 = 999 (calculated)90,000 x 0.01 = 900 (calculated)
Chance a positive result is correct99 / (99 + 999) = 9% (calculated)9,900 / (9,900 + 900) = 92% (calculated)

The same test is right 9 times in 100 for a rare condition in the general population but 92 times in 100 where the condition is common. It is the base rate, not the machine, that changes.

Source: Hewitt, Lyons, Suchocki and Yeh, Conceptual Integrated Science, 2nd edition (Pearson, 2013), ch. 16, 'Genetics'
  • Hewitt, Lyons, Suchocki and Yeh, Conceptual Integrated Science, 2nd edition (Pearson, 2013) · ch. 16, 'Genetics'
  • National Human Genome Research Institute (NIH), A Brief Guide to Genomics · web
  • Pre-Conception and Pre-Natal Diagnostic Techniques Act, 1994; Environment (Protection) Act rules of 1989 under which GEAC works · statutory background: the 1994 Act bans prenatal sex determination; the 1989 Rules govern genetically engineered organisms
  • Standard histories of genetics, as summarised in Encyclopaedia Britannica (Gregor Mendel; Rosalind Franklin; Francis Crick; James Watson; DNA; Meselson-Stahl) and the Nobel Prize site (Physiology or Medicine 1962, 1968, 1978 and 1980; Chemistry 1993) · dates and findings for Mendel (read 1865, published 1866, rediscovered 1900), Sutton and Boveri (1902 to 1903), Morgan (1910), Griffith (1928), Avery, MacLeod and McCarty (1944), Hershey and Chase (1952), Chargaff, Franklin and Gosling (Photo 51, 1952), Watson and Crick (Nature, 25 April 1953), Meselson and Stahl (1958), Crick and others (1961), Nirenberg and Matthaei (1961), Jacob and Monod (1961), Luria and Delbruck (1943), Cohen and Boyer (1973), Sanger (1977), Mullis (1983). Checked in November 2026: Mendel read 8 February and 8 March 1865, printed 1866 (Villanova Mendel text); Photo 51 taken May 1952 and published in Nature on 25 April 1953 beside Watson and Crick (King's College London); Meselson and Stahl at Caltech, PNAS 44 (1958) (Arizona State University Embryo Project); Mullis developed PCR in 1983 at Cetus, Nobel 1993 (Britannica)

Check your understanding

A palindromic site of 6 bases occurs, by chance, about once every

There are 4 to the power 6 = 4,096 possible 6-base sequences, so a given one turns up about once per 4,096 bases in random DNA.

After 20 perfect cycles of PCR, one starting molecule has become about

Each cycle doubles the number, and 2^20 = 1,048,576. A billion needs 30 cycles.

A screening test with 99% sensitivity and 99% specificity is used on people with a 1-in-1,000 base rate. Most positive results will be

Per 100,000 people there are 99 true and 999 false positives, so about 9 in 100 positives are right.
← Part 4: When the code changes: mutation, damage and cancerMark read · Part 6: Changing organisms: breeding, GM crops and gene editing →