For decades molecular genetics could reason about DNA but hardly read it. Two inventions changed that. In 1977 Frederick Sanger in Cambridge published a method for reading DNA's sequence letter by letter. Sanger sequencing copies DNA using a mix of normal building blocks and modified ones that stop each copy at a known letter, producing fragments of every length that can be sorted to spell out the sequence. It earned Sanger his second Nobel Prize in Chemistry, in 1980; his first, in 1958, was for working out the sequence of the protein insulin.
The second invention came, by his own account, to the chemist Kary Mullis on a night drive in California in 1983. The polymerase chain reaction, or PCR, repeatedly heats and cools a sample so that an enzyme copies a chosen stretch of DNA again and again. Each cycle doubles the amount, so thirty cycles turn one molecule into about a billion copies. PCR, which makes billions of copies of a short stretch of DNA from a tiny sample, made it possible to study DNA from a single hair, a drop of blood or a Neanderthal bone. Mullis shared the Nobel Prize in Chemistry in 1993, and during the COVID-19 pandemic PCR tests became familiar to billions of people.
The Human Genome Project
The Human Genome Project, launched in 1990 by the United States with partners in Britain, France, Germany, Japan and China, set out to read all three billion letters of human DNA. A private company, Celera, led by Craig Venter, raced it with a faster method from 1998. The two announced rough drafts together in 2000 and published them in 2001, and the public project declared the work essentially complete in 2003, at a cost of about 2.7 billion dollars. The public project released its data freely every day, a decision that shaped the openness of genetics ever since.
The biggest surprise was how few genes there were. Some had guessed 100,000 or more. Humans turned out to have roughly 20,000 protein-coding genes, about as many as a tiny roundworm. Only about 1 to 2 percent of the genome codes for protein; the rest includes switches, RNA genes, repeated sequences and remnants of ancient viruses, much of it still poorly understood. The 2003 sequence also had gaps in hard-to-read repetitive regions. In 2022 the Telomere-to-Telomere consortium published the first complete, gapless human genome, filling in the last 8 percent.
From millions to hundreds
New machines read millions of fragments at once. The cost of sequencing a human genome fell from around 100 million dollars in 2001 to about a thousand dollars by the late 2010s, and lower since, a decline far faster than the famous fall in computer chip costs. Cheap sequencing now helps diagnose rare diseases in sick newborns, tracks the spread of viruses, and guides cancer treatment by reading tumor DNA. It also guides drug choice. Before prescribing the HIV drug abacavir, doctors test for a gene variant called HLA-B*57:01, because carriers of the variant risk a dangerous allergic reaction. Matching drugs to genes in this way is called pharmacogenomics.
DNA as evidence
On 10 September 1984 the geneticist Alec Jeffreys, at the University of Leicester, looked at an X-ray film of DNA patterns and realized that some repetitive regions vary so much that they could identify individuals. He called the method DNA fingerprinting. Its first criminal use came soon after, in the murders of two teenage girls in nearby villages. A young man had confessed to one of the killings, but the tests cleared the man who had confessed, showing he could not have committed either murder. He became the first person cleared by DNA evidence. After the police took blood from thousands of local men, Colin Pitchfork was convicted in 1988. Since then DNA testing has freed hundreds of wrongly convicted people, many after long prison terms.
Ancestry tests
Consumer tests compare a customer's DNA with reference groups of people whose families have lived in a region for generations, then estimate what share of the customer's DNA most resembles each group. The results are measures of resemblance to modern reference groups, not records of where ancestors lived. They shift as companies add data, and different companies give different answers. Finding relatives, by contrast, is far more precise than estimating regions.
Worked example: an ancestry report
Maya's report says she is "12 percent Italian." Her family has no Italian stories, and she wonders what it means. It does not mean a great-grandparent was born in Italy, though that is possible. It means that some stretches of her DNA resemble those of the company's Italian reference group more than its other groups. Neighboring populations share much ancestry, so DNA common in Greece, southern France or the Balkans can look Italian, and national borders are recent compared with the mixing that shaped genes. A year later, after the company adds reference samples, her report may say 6 percent Italian and 8 percent Greek, though her DNA has not changed at all. The precise-looking percentage invites more confidence than the method can bear.