Genes, Heredity, and Disease

TL;DR. A gene is a recipe for a protein, and proteins do everything: they carry oxygen, digest food, build structure, and pass signals. A disease-causing mutation is a typo in one recipe, and its consequences follow from what that protein was for. Two very different things get called "genetic." A single-gene disease (sickle cell, cystic fibrosis, Huntington's) is one broken recipe with a predictable inheritance pattern. A polygenic disease (type 2 diabetes, heart disease, depression, most of what people actually get) is the summed nudge of hundreds or thousands of common variants, each nearly meaningless alone, none of them a verdict. Confusing the two is the single most common error in how people think about family history.

Key takeaways

• DNA to RNA to protein. A mutation changes a protein: makes it absent, misshapen, overactive, or made in the wrong amount. Everything downstream follows from that.
• Dominant means one bad copy is enough. Recessive means you need two, and people with one copy are carriers who are usually healthy and can be unaware for life.
• Penetrance is the fraction of people carrying a variant who actually develop the disease. Huntington's is essentially 100 percent. A BRCA1 mutation is high but not certain. Most risk variants are far below 10 percent.
• Germline mutations are inherited and in every cell. Somatic mutations happen during life in one cell line and are not passed to children. Almost all cancer is somatic.
• Heritability is a population statistic about variation, not a personal percentage. "80 percent heritable" does not mean 80 percent of your case was genes.
• Genes set a range; environment picks the point in it. PKU is the cleanest proof: a fully genetic disease, entirely prevented by diet.

The machinery in one page

In short: A gene is a recipe for a protein, and one wrong letter can change the protein's shape and therefore its job.

Your DNA is a four-letter code (A, C, G, T) roughly 3 billion letters long, packed into 23 pairs of chromosomes. One of each pair came from each parent. About 20,000 stretches of that code are genes, and each gene is a recipe for one protein (with variations that let one gene make several related versions).

Reading the recipe is a two-step process. The gene is copied into messenger RNA (transcription), and the RNA is read three letters at a time by a ribosome, which links the corresponding amino acids into a chain (translation). The chain folds into a specific three-dimensional shape, and the shape is the function. A protein works because its surface fits something else: a molecule to be carried, a reaction to be catalysed, a signal to be received.

That is why a single-letter typo can be catastrophic. Change one letter in the gene for haemoglobin's beta chain, and a glutamic acid becomes a valine. That swaps a water-loving amino acid for a water-fearing one on the protein's surface, so the molecules stick to each other when oxygen is low, and the whole red blood cell deforms into a rigid crescent. One letter, and you have sickle cell disease (Chapter 48).

Not every typo matters. Most of the genome is not gene, much of a gene's code is redundant (several triplets code for the same amino acid), and many amino acid swaps change nothing important. Everyone carries millions of variants and a handful of seriously damaged genes without consequence.

The four inheritance patterns

In short: Dominant, recessive, X-linked, and mitochondrial inheritance each leave a recognisable family pattern, and recessive disease often appears with no family history at all.

Because chromosomes come in pairs, a variant's effect depends on whether one copy or two is required, and on which chromosome it sits.

Autosomal dominant. One faulty copy causes disease. An affected parent passes it to each child with probability 1/2. Every generation is affected; the pattern is vertical in a family tree. Examples: Huntington's disease, familial hypercholesterolaemia, Marfan syndrome, most hereditary breast/ovarian cancer syndromes.

Autosomal recessive. Two faulty copies are required. Carriers with one copy are healthy. Two carrier parents have a 1/4 chance per child of an affected child, 1/2 of a carrier, 1/4 of neither. Disease often appears with no family history at all, which is why it feels like it comes from nowhere. Examples: cystic fibrosis, sickle cell disease, thalassemia, Tay-Sachs, phenylketonuria.

X-linked. The gene is on the X chromosome. Males have one X, so a single faulty copy is fully expressed; females have two and are usually carriers. The classic signature is a disease that skips through unaffected mothers to affected sons. Examples: haemophilia A and B, Duchenne muscular dystrophy, red-green colour blindness, G6PD deficiency.

Mitochondrial. Mitochondria (the cell's power plants) carry their own small genome, and you inherit all of yours from your mother. Affected mothers pass it to all children; affected fathers pass it to none. Examples: Leber's hereditary optic neuropathy, MELAS.

Don't be confused: "runs in the family" is not the same as "genetic." Families share far more than DNA. They share diet, smoking, income, language, neighbourhood, and air quality. Type 2 diabetes clusters in families partly through genes and partly because a family eats together. Conversely, a genetic disease can appear with no family history at all, either because it is recessive and both parents were silent carriers, or because the mutation is brand new in that child (de novo), which accounts for a large share of Duchenne muscular dystrophy and achondroplasia cases.

Penetrance: the difference between a gene and a fate

In short: Penetrance is the proportion of carriers who actually develop the disease, and for most variants it is low.

Penetrance is the probability that someone carrying a variant develops the disease. It is the number people most need and most rarely hear.

Expressivity is the separate question of how severely it shows. Two siblings with identical cystic fibrosis mutations can have markedly different lung disease, because other genes and environment modify the outcome.

Polygenic: how most common disease is actually inherited

In short: Common diseases are hundreds of small genetic nudges rather than one broken gene, which is why risk scores predict populations far better than people.

Sickle cell has one gene. Type 2 diabetes has hundreds of contributing variants, and so do coronary artery disease, obesity, depression, schizophrenia, asthma, and height.

Genome-wide association studies compare millions of common variants between people with and without a condition. They routinely find hundreds of hits, each shifting risk by a few percent. Sum them, weighted, and you get a polygenic risk score, which does have real predictive power at the extremes of its distribution: people in the top few percent of a coronary disease score can carry risk comparable to a single-gene familial cholesterol disorder.

Three honest limits on those scores. They predict populations far better than individuals. They were derived overwhelmingly from people of European ancestry and transfer poorly to others, which is an active fairness problem in genomics. And for most diseases they add modestly to what you already know from age, blood pressure, smoking, cholesterol, and family history.

Heritability, the most misunderstood number in medicine

"Schizophrenia is about 80 percent heritable." Almost everyone reads that as "80 percent of the disease is genetic." It does not mean that.

Heritability is the proportion of the variation between people in a specific population at a specific time that is attributable to genetic variation. It is a property of a population, not of a person or a disease.

The consequences are counterintuitive. If everyone in a population smoked identically, the variation in lung cancer between them would be almost entirely genetic, and lung cancer would look highly heritable, even though smoking causes most of it. Improve the environment for everyone and heritability goes up, because you have removed environmental variation. Height in well-fed countries is about 80 percent heritable and was much less so when childhood nutrition varied enormously.

So heritability tells you nothing about whether an intervention will work, and nothing about any individual's case.

Genes and environment, together

In short: PKU is entirely genetic and entirely preventable by diet, which is the cleanest proof that a genotype is a range rather than a verdict.

The cleanest demonstration in medicine is phenylketonuria (PKU). A recessive mutation disables the enzyme that breaks down the amino acid phenylalanine. It builds up and causes severe, permanent intellectual disability. This is 100 percent genetic and 100 percent preventable: put the infant on a low-phenylalanine diet from birth and development is normal. That is why every newborn in most of the world gets a heel-prick blood test in the first days of life, a programme running since the 1960s.

Other clean examples:

• Lactase persistence. Most adult humans worldwide lose the ability to digest milk sugar. Variants keeping the enzyme switched on arose independently in Northern Europe and in East African and Middle Eastern pastoralist populations. Whether that matters at all depends entirely on whether your diet contains milk.
• ALDH2 deficiency. About a third of people of East Asian descent carry a variant that slows breakdown of acetaldehyde, alcohol's toxic intermediate. It causes flushing, and it substantially raises oesophageal cancer risk if the person drinks. Never drink, and it is invisible.
• G6PD deficiency. Common around the historic malaria belt. Harmless until certain drugs (including some antimalarials) or fava beans trigger red cell breakdown.

The pattern is general: a genotype is a range of possible outcomes, and the environment selects within it.

Germline vs somatic, and why cancer sits in this chapter

Germline mutations are inherited, present in every cell, and passed to children. Somatic mutations arise during life in one cell and are copied only into that cell's descendants.

Cancer is fundamentally a somatic genetic disease: a cell accumulates mutations in the genes that control division and death until it escapes the rules (Chapter 24). Only about 5 to 10 percent of cancers begin with an inherited high-risk variant. When they do (BRCA1/2, Lynch syndrome, familial adenomatous polyposis), the inherited copy is a head start: every cell already carries one of the several hits required, so cancers appear earlier, in multiple sites, and across generations.

What is actually testable today

In short: The gap between what can be found and what can be done about it is the ethical core of genetic testing.

The gap between what can be found and what can be done about it is the ethical core of this field. Testing a healthy 25-year-old for Huntington's returns certain, unalterable information about their fifties. Roughly 10 to 25 percent of at-risk people choose to be tested, and that choice is a legitimate one either way.

Sources and notes

Molecular genetics and inheritance patterns are standard (Alberts, Molecular Biology of the Cell; Nussbaum, Thompson & Thompson Genetics in Medicine). BRCA1 penetrance figures are from Kuchenbaecker et al. (2017, JAMA), a large prospective cohort. APOE4 effect sizes vary by ancestry and are given as commonly cited ranges. The hereditary share of cancer (5 to 10 percent) is the standard estimate from cancer genetics literature. PKU newborn screening began with Robert Guthrie's test in 1963. Lactase persistence convergent evolution: Tishkoff et al. (2007). Polygenic score portability across ancestries: Martin et al. (2019, Nature Genetics).

Open questions. How much of the "missing heritability" for common diseases lies in rare variants, gene interactions, or measurement noise is unresolved. The clinical utility of polygenic risk scores outside research settings is actively debated.

Next: what happens after you swallow the pill. 👉