How Do Genes Actually Work?

Ask most people what a gene is and you will hear "a piece of DNA that codes for a trait." That is close, and also wrong in ways that matter. A gene is a stretch of DNA that a cell can read to make a specific RNA or protein, and whether that stretch is read depends on the cell, the moment, and the signals around it. Humans have about 19,900 protein-coding genes, a roundworm has more than 19,000, and a water flea has at least 30,907. Here is how genes work, and why counting them says so little about complexity.

Educational content, not medical advice. This article describes general findings from genetics research. For questions about a personal or family health condition, talk to a doctor or a genetic counselor.

Key takeaways

• A gene is a section of DNA that a cell reads to make a protein or a functional RNA molecule.
• Humans have about 19,900 protein-coding genes, a count similar to a roundworm's and lower than a water flea's.
• Most protein-coding genes are active somewhere in the body, so cell types differ by which genes they use, not by which genes they own.
• Complex traits such as height depend on thousands of small-effect variants, not on one "gene for" the trait.
• Genes are predispositions that interact with environment and chance, not fixed destinies.

What is a gene, exactly?

MedlinePlus, the US National Library of Medicine's consumer health site, calls a gene "the basic physical and functional unit of heredity". Human genes range in size from a few hundred DNA base pairs to more than 2 million. Most genes are identical in everyone, and a small number of genes, less than 1 percent of the total, are slightly different between people.

The picture is messier than a tidy definition suggests. In 2007, Mark Gerstein and colleagues traced how the idea of a gene had changed over a century and summarized what the ENCODE project had found about dispersed regulation and widespread transcription. They proposed an updated definition built around "a union of genomic sequences encoding a coherent set of potentially overlapping functional products". In plain terms, a gene is a region that a cell can turn into one or more useful molecules, and those regions can overlap.

An analogy helps. Think of the genome as a very large cookbook kept in a library that never lends out its books. A gene is a single recipe. The cell photocopies the recipe it needs (RNA), takes the copy to the kitchen, and cooks. The book itself stays put.

What is inside a typical gene?

A typical protein-coding gene has several parts:

• Promoter: a stretch of DNA near the start that tells the cell's machinery where to begin reading and helps control when.
• Exons: the pieces that end up in the final instructions.
• Introns: stretches between exons that are cut out after the gene is copied.
• Regulatory regions: switches, some located far from the gene, that raise or lower how often it is read.
• A termination region: the signal for the machinery to stop and finish the RNA copy.

Back to the cookbook: the promoter is the title of the recipe, the exons are the steps that are actually cooked, the introns are notes in the margin that are crossed out in the copy, and the regulatory regions are sticky notes that say "only make this on weekends."

Not every gene makes a protein

Genes that make proteins get the most attention, but cells also read genes whose product is an RNA that does a job itself. The GENCODE project, which annotates the human genome, reports in its current release (Release 50) the following counts, which change as annotation improves:

Category	Number of genes (GENCODE Release 50)
Protein-coding genes	19,442
Long non-coding RNA genes	35,885
Small non-coding RNA genes	7,608
Pseudogenes (damaged or inactive copies)	14,702
Total	78,733

Source: GENCODE human statistics. A longer list is not the same as a longer list of well-understood functions. Many entries in the non-coding categories have been identified by the way they are transcribed, and what each of them does, if anything, is an active question.

How many genes do humans have, and why is the number so small?

When the Human Genome Project finished its work in April 2003, many researchers expected a far larger number. The consortium announced an "essentially complete" sequence covering about 92 percent of the genome, per the National Human Genome Research Institute, and a truly complete sequence arrived only on March 31, 2022. The complete sequence, reported by Sergey Nurk and colleagues in Science, is 3.055 billion base pairs long and adds nearly 200 million base pairs of new sequence to the earlier assemblies. For the story of the project itself, see the Human Genome Project explained and the human genome revolution.

The gene count itself is surprisingly modest. The 1998 sequence of the roundworm Caenorhabditis elegans, a creature with a 97-million-base-pair genome, revealed more than 19,000 genes. The water flea Daphnia pulex has at least 30,907, which the authors of its genome paper attributed to frequent gene duplication. Over a third of those genes had no detectable match in other sequenced proteomes.

So a human does not have more genes than a worm in any way that explains being human. What differs is how the genes are used: when they are switched on, in which cells, in which combinations, and how their RNA is edited after copying. A better comparison is a library. Two libraries can own the same number of books and be wildly different in what readers can do with them, depending on the catalog, the reading rooms, and the staff.

How does a gene become a protein?

The process is called gene expression, and it has three main stages.

Step 1: Transcription

An enzyme called RNA polymerase attaches near the promoter and builds an RNA copy of the gene, one letter at a time. The copy is called messenger RNA (mRNA) when the gene codes for a protein. This happens in the cell's nucleus. Transcription factors, proteins that bind specific DNA sequences, help decide whether the polymerase gets a foothold.

Step 2: Processing and splicing

The first RNA copy is not yet a finished recipe. The cell cuts out the introns and joins the exons together, a step called splicing, and adds protective features to the ends of the molecule. Then the mature mRNA leaves the nucleus.

Splicing is also a source of variety. A given gene can often be spliced in more than one way, so one stretch of DNA can yield several related proteins. In 2008, Qun Pan, Benjamin Blencowe, and colleagues used RNA sequencing to estimate that about 95 percent of human genes with more than one exon undergo alternative splicing, with roughly 100,000 intermediate- to high-abundance events across major tissues. That helps explain how roughly 20,000 protein-coding genes can supply a far larger number of distinct proteins.

Step 3: Translation

In the cytoplasm, a molecular machine called the ribosome reads the mRNA three letters at a time. Each three-letter group is a codon, and each codon specifies one amino acid, the building block of proteins. The ribosome links amino acids into a chain, and the chain folds into a three-dimensional shape that determines what the protein does: speed up a reaction, form a structure, carry a molecule, or send a signal.

How the codes were cracked is its own story, told in the Nirenberg experiment. How the folding is predicted by software today is told in AlphaFold and protein folding.

Stage	Where it happens	What goes in	What comes out
Transcription	Nucleus	DNA of a gene	Primary RNA copy
Processing and splicing	Nucleus	Primary RNA copy	Mature mRNA (introns removed, exons joined)
Translation	Cytoplasm, on ribosomes	Mature mRNA and amino acids	A chain of amino acids
Folding and modification	Cytoplasm and organelles	Amino acid chain	A working protein

Why do a heart cell and a neuron use the same DNA differently?

Almost every cell in your body carries the same genome, yet a heart muscle cell and a neuron could hardly be less alike. The difference lies in which genes each cell reads. In 2015, Mathias Uhlén and colleagues reported in Science a tissue-based map of the human proteome in which their analysis detected more than 90 percent of the putative protein-coding genes across 32 tissues and organs. Put another way, most genes are used somewhere. Cell identity is defined by pattern, not by ownership.

This is also why the question of how a stem cell decides what to become is largely a question about gene regulation; see how stem cells know what to become.

Transcription factors: the readers of regulatory DNA

The regulators are largely transcription factors, proteins that recognize short DNA sequences and promote or suppress transcription. Mapping which sequences each one recognizes is still unfinished work. A 2026 paper in Nature from the Codebook effort noted that several hundred putative human transcription factors lacked known binding motifs, and the project set out to determine their sequence specificity systematically. Nearly two decades after the first human genome, scientists are still compiling the list of who reads what.

Epigenetics: tags on DNA and its packaging

Chemical marks on DNA and on the proteins that package it can make a gene easier or harder to read. These epigenetic marks respond to development, environment, and cell state. A 2014 review by Edith Heard and Robert Martienssen in Cell summarized the field's position: environmental factors clearly can alter gene expression, epigenetic inheritance across generations occurs particularly in plants, and how much of it happens in humans remains unclear. The headline-friendly claim that experiences are routinely passed down in the genome is, so far, a hypothesis.

Reusing the same genes in new ways

Evolution often works by redeploying existing genes. In 1995, Georg Halder, Patrick Callaerts, and Walter Gehring showed that expressing the eyeless gene in fruit fly tissues that normally make legs, wings, or antennae induced well-formed ectopic eyes, which led them to propose it as a master control gene for eye development. A Pax-6 relative has since been found in a ribbonworm, where it is expressed in the nervous system and the eye region during head regeneration, suggesting an ancient origin (Loosli and colleagues, 1996). Body plans that look nothing alike share control genes used in different settings.

What is the "dark matter" of the genome, and is it junk?

Protein-coding exons make up only a small part of the genome. For decades, the rest was often called "junk DNA." That label has proved too blunt, and the argument over what to replace it with is not finished.

In 2012, publications from the ENCODE Consortium were read as claiming that more than 80 percent of the human genome is functional. In 2013, Dan Graur and colleagues replied that this conflicts with estimates that under 10 percent of the genome is conserved by purifying selection, and argued that the claim rested on a loose definition of "function." Both camps agree that the non-coding genome includes real regulatory elements and also includes sequence with no known role. The disagreement is over where the line sits and which definition of "function" to use.

A regulatory change you can see in a glass of milk

Few examples show regulatory DNA at work as clearly as lactase persistence, the ability to digest milk sugar as an adult. In most mammals the lactase gene (LCT) switches off after weaning. In 2007, Sarah Tishkoff and colleagues studied 470 people in Tanzania, Kenya, and Sudan and identified three variants linked to lactase persistence that boosted LCT promoter activity and arose on different haplotypes from the European variant. Haplotype patterns suggested a selective sweep around 7,000 years ago. They described it as convergent evolution driven by animal domestication and adult milk consumption.

Note what changed: not the protein, which is the same enzyme, but when the gene is switched on. In Europeans, lactase persistence is mainly tied to a variant in a neighboring gene, MCM6, known as rs4988235-T. (A 2024 paper on Siberian populations restates that link and adds that the non-persistent version becomes more common moving northeast across Central Asia and Siberia.)

How do mutations change genes?

A mutation is any change in DNA sequence. Copying errors during cell division, radiation, chemicals, and viruses can all introduce them. How often do new mutations arise between parent and child? In 2012, Augustine Kong and colleagues sequenced the whole genomes of 78 Icelandic parent-offspring trios. With an average father's age of 29.7, the average de novo mutation rate was 1.20 × 10⁻⁸ per nucleotide per generation. Paternal age was the main driver of variation in the rate, adding about two mutations per year, and the authors fitted a doubling time of 16.5 years for paternal mutations. A rough calculation shows what that rate means: applied to a diploid genome of around six billion bases, it implies dozens of new changes in every child. (That is the author's back-of-envelope arithmetic, not a figure from the paper.)

Types of mutation

Type	What changes	Typical effect
Silent	One letter changes, but the codon still specifies the same amino acid	Usually none, because the genetic code has redundancy
Missense	One amino acid is swapped for another	Anywhere from harmless to serious
Nonsense	A codon becomes a stop signal	Shortened protein, often nonfunctional
Frameshift	Letters inserted or deleted, shifting the reading frame	Often severe, because every codon downstream changes
Regulatory	Switches near a gene change, not the protein	Gene read at the wrong time, place, or level
Repeat expansion	A short sequence is repeated too many times	Depends on gene and repeat count

Three disease examples from the textbooks

• Sickle cell disease. A single substitution in the beta-globin gene (HBB) replaces glutamic acid with valine at position 6 of the protein, per MedlinePlus. One amino acid, one letter in the DNA, a very different red blood cell.
• Cystic fibrosis. The most common mutation, called delta F508, is a deletion of one amino acid at position 508 in the CFTR protein. The abnormal channel breaks down soon after it is made and never reaches the cell membrane to move chloride ions.
• Huntington disease. The HTT gene contains a CAG repeat that is usually present 10 to 35 times. Disease is linked to 36 or more repeats, and at 40 or more the disorder is almost always developed. Counts of 27 to 35 do not cause the disease but can expand into the disease range when passed to children.

These three show the three kinds of damage. One is a letter swap, one is a small deletion, and one is a stutter that runs too long. For the history of the idea that single genes make single enzymes, see the Beadle-Tatum experiment. The way mutations accumulate in cells that then divide without restraint is covered in what causes cancer.

Why is there rarely "a gene for" a trait?

Public conversation tends to ask for "the gene for" height, intelligence, or a disease. A few conditions are close to single-gene stories, and sickle cell disease and Huntington disease are among them. Most human traits are not.

Height as the test case

Height is among the most carefully studied complex traits. In 2022, Loïc Yengo and a large consortium analyzed data from about 5.4 million people of diverse ancestries and identified 12,111 independent SNPs (single-letter variants) associated with height. Those variants account for nearly all the heritability that common variants can capture, and they cluster in 7,209 segments covering about 21 percent of the genome. In people of European ancestry the variants explain about 40 percent of the variation in height; in other ancestries the figure was 10 to 20 percent, which the authors attributed mainly to differences in linkage and allele frequency.

The practical lesson is that a trait can be strongly inherited and still be nothing like a one-gene trait. Thousands of small nudges add up.

What "heritability" does and does not mean

Heritability is a statistic about a population, not about a person. A Finnish twin study of 8,753 adult twin pairs, for example, estimated a broad-sense heritability of 49.7 percent for diurnal type, meaning morning or evening preference. That does not mean any individual's preference is half genetic. It means that about half of the variation among people in that population is associated with genetic differences. A different population or environment could give a different number.

Polygenic scores

Researchers add up many variants into a polygenic score to estimate inherited risk. In 2018, Amit Khera and colleagues reported scores that identified about 8 percent of people as having more than threefold higher risk of coronary artery disease, which they noted was roughly 20 times the share of people who carry a rare single-gene mutation conferring comparable risk. The finding supports continued study of polygenic prediction, and its limits are real: a 2019 analysis by Alicia Martin and colleagues showed that current scores are much less accurate for people of non-European ancestry, because most genetic studies have enrolled people of European descent.

Genes interact, and environments differ

Two further complications deserve names. Pleiotropy means one gene affects several traits. Epistasis means the effect of a variant depends on variants elsewhere. Add environment, and the same genotype can produce different outcomes. This is why "gene for X" stories mislead, and why researchers speak of risk and predisposition more often than of cause.

What can gene editing change, and what can't it?

Editing tools make it possible to alter specific genes in living cells. On December 8, 2023, the US Food and Drug Administration approved two gene therapies for sickle cell disease in patients 12 and older, Casgevy and Lyfgenia. The agency called Casgevy the first FDA-approved therapy using CRISPR/Cas9, a type of genome editing technology; Lyfgenia uses a lentiviral vector for genetic modification instead.

That a disease caused by a single well-understood letter change was the first target is not a coincidence. Conditions driven by many genes and the environment are far harder targets. For the mechanism and an honest accounting of results so far, see what is CRISPR and how does gene editing work and CRISPR in 2026: cures and hype. The same logic of reading and rewriting nucleic acid instructions underlies mRNA vaccines, which deliver a temporary message and do not alter DNA.

Common myths about genes

"We decoded the genome, so we know what it all does." The 2003 announcement covered about 92 percent of the genome; a complete sequence came in 2022. Having the sequence is different from knowing what each region does, when, or with what partners. GENCODE's gene counts are still revised with each release.

"Humans and chimpanzees are 98 percent identical, so we are almost chimps." The Chimpanzee Sequencing and Analysis Consortium reported in 2005 approximately 35 million single-nucleotide changes and 5 million insertion or deletion events between the two genomes. Thirty-five million changes across roughly three billion letters is on the order of 1 percent for single-letter differences, and the figure grows when insertions and deletions are counted. The percentage you see depends on what is counted, and it says little about biology, because a small number of changes in regulatory switches can matter a great deal.

"Junk DNA is useless." Some non-coding DNA has clear regulatory functions, and some has no known role. The accurate statement is that the proportion in each category is disputed.

"Genes determine destiny." Genes shift probabilities. Environment, behavior, and chance contribute, and even for single-gene conditions, severity can vary.

"Genes only come from your parents." Within a lifetime, cells accumulate mutations, and in nature, microbes pass genes between unrelated species, a process described in horizontal gene transfer. Inheritance from parents is the main route in people, but not the only way genetic material moves in the living world.

Frequently asked questions

Q: How many genes do humans have? A: About 19,900 protein-coding genes, according to MedlinePlus, and 19,442 in GENCODE Release 50. The total of all genes, including those that make non-coding RNA and pseudogenes, is much higher at 78,733 in the same release. Counts change as annotation improves.

Q: What is the difference between DNA, a gene, a chromosome, and a genome? A: DNA is the molecule. A gene is a stretch of DNA that a cell can read to make an RNA or protein. A chromosome is a long package of DNA and supporting proteins that carries many genes. The genome is the complete set of DNA in an organism, which for the first complete human sequence totaled 3.055 billion base pairs.

Q: What is gene expression? A: It is the process by which a gene's information is used: DNA is copied into RNA, the RNA is processed, and for protein-coding genes it is translated into protein. Cells control which genes are expressed, so a liver cell and a neuron use the same genome differently.

Q: Can my genes change during my life? A: The DNA in your cells can accumulate mutations, which is why cancer risk rises with age and exposure to mutagens. Mutations in sperm or egg cells can be passed on. The rate of new mutations transmitted from parent to child was estimated at 1.20 × 10⁻⁸ per nucleotide per generation in an Icelandic study.

Q: Is intelligence, height, or personality controlled by a single gene? A: No. Height, the best-studied example, involves thousands of variants of small effect. Behavioral and cognitive traits are also influenced by many variants plus environment, and prediction from DNA alone is limited.

Q: Why are humans and chimpanzees so different if their genomes are so similar? A: Differences in regulatory regions, gene copy numbers, and when and where genes are switched on matter more than the raw count of changed letters. A 2005 comparison found about 35 million single-letter differences and 5 million insertions or deletions.

Q: Has any gene therapy been approved that uses CRISPR? A: Yes. The FDA approved Casgevy on December 8, 2023, for patients 12 and older with sickle cell disease and recurrent vaso-occlusive crises, describing it as the first approved therapy using CRISPR/Cas9. Whether and how it suits any individual is a clinical decision.

The takeaway

A gene is best understood as a recipe that a cell can choose to cook, not as a switch that sets a trait. Humans carry about as many protein-coding recipes as a roundworm, yet build far more elaborate bodies because of what surrounds the recipes: regulatory DNA, transcription factors, alternative splicing, chemical marks, and the history of the cell reading them. That also explains why most traits have no single gene behind them, and why the first approved gene-editing therapy targets a condition traced to one well-understood letter change. For the questions that matter most about health and variation, the unit of explanation is rarely one gene. It is a network and its context.

Sources

• MedlinePlus (US National Library of Medicine). What is a gene?
• GENCODE. Human gene statistics, Release 50.
• Gerstein MB et al. (2007). What is a gene, post-ENCODE? Genome Research.
• Nurk S et al. (2022). The complete sequence of a human genome. Science.
• Yengo L et al. (2022). A saturated map of common genetic variants associated with human height. Nature.
• Kong A et al. (2012). Rate of de novo mutations and the importance of father's age to disease risk. Nature.
• Heard E, Martienssen RA (2014). Transgenerational epigenetic inheritance: myths and mechanisms. Cell.
• US Food and Drug Administration (2023). FDA approves first gene therapies to treat patients with sickle cell disease.

Related Reading

• What Is CRISPR and How Does Gene Editing Actually Work?
• How Does DNA Work? The Molecular Blueprint of Life Explained
• How Do Vaccines Actually Work?
• How Does the Human Eye Actually Work?
• How Does the Human Heart Actually Work?
• Mendel's Pea Plant Experiments