A gene is a stretch a cell actually uses
You asked what a gene is because someone named one as if it were a personality, a destiny, or a product. It is none of those. It is a stretch of DNA a cell can transcribe, a recipe card in a library that is mostly not recipe cards.
The real term is gene: a functional unit on a chromosome, with a name such as HBB or CFTR, that the cell treats as a unit of work. Work often means making RNA. Sometimes that RNA is translated into a protein. Sometimes the RNA is the product.
Two copies of a book sit on the shelf for most genes: one from each parent. The gene is the stretch. The editions of that stretch are alleles. The trait you can see is later, and leakier, than the stretch itself.
About 20,000 protein-coding genes, not 100,000
Twentieth-century guesses put the human gene count near 100,000. That number was a hope about complexity, not a count. The Human Genome Project and the catalogs that followed cut it down.
Humans have about 20,000 protein-coding genes. MedlinePlus currently cites a figure near 19,900. The exact integer moves a little as annotators argue over edge cases. It does not move back to 100,000.
That count is protein-coding loci, not every gene-like stretch. The genome also holds genes whose product is RNA, not protein. If someone still says humans have 100,000 genes, they are quoting a discarded estimate.
- Protein-coding genes in humans: about 20,000, not 100,000
- Two copies of most genes, one from each parent
- A gene has a symbol (HBB, CFTR, APOE) and a location on a chromosome
- Most DNA still sits outside those genes
Many genes never make a protein
A gene is not a synonym for a protein recipe. Plenty of genes are transcribed into RNA that never becomes a protein: ribosomal RNAs, transfer RNAs, microRNAs, long noncoding RNAs. They still do work.
Even among protein-coding genes, much of the sequence is not coding. Promoters, enhancers, and untranslated regions decide when a gene is read and how stable the RNA is. A variant outside the protein letters can still matter.
When a headline says researchers found the gene for a trait, ask whether they found a protein-coding gene, a regulatory stretch, or a statistical association nearby. Those are different claims.
Introns and exons in one sitting
Many human protein-coding genes arrive in pieces. The pieces kept in the mature RNA are exons. The pieces cut out are introns. Splicing is the edit that turns a split gene into a continuous message.
One gene can be spliced more than one way, so one locus can yield more than one protein. That is why gene count and protein count are not the same number, and why a variant in an intron can still wreck a gene by breaking a splice site.
You do not need a course in splicing to use the words. Exons are the kept stretches. Introns are the removed ones. The gene includes both, plus nearby sequence the cell uses to decide when to read it.
Alleles are editions of the same stretch
An allele is a version of a gene, or of a site inside a gene. At a given place you typically carry two, one on each copy of the chromosome. They may be identical. They may differ by a single letter.
People talk about the sickle cell allele or the C282Y allele in HFE because those editions have names in medicine. Most alleles have no common name. They are simply the letters present at a site.
A SNP is one common way two alleles differ: a single base among the 3.2 billion. Not every gene difference is a SNP. Some are insertions, deletions, or extra copies of a stretch.
When people say they have the gene for something, they usually mean they carry a particular allele. Everyone has the HBB gene. Not everyone carries the sickle allele. The gene is the stretch. The allele is the edition.
Homozygous and heterozygous in ordinary language
If both of your copies match at a site, you are homozygous there: two of the same edition. If they differ, you are heterozygous: two different editions. That is the whole idea, dressed in Greek.
Homozygous reference means both copies match the reference genome's letter. Homozygous alternate means both copies match a different letter. Heterozygous means one of each. None of those words is a diagnosis.
On a consumer file, a missing site is not homozygous anything. It was not called. Do not fill in a wild-type gene because a chip skipped it.
- Homozygous: both copies carry the same allele
- Heterozygous: the two copies differ
- Two copies of autosomes; X and Y do not follow that rule in the same way
- A missing chip row is not a genotype
One gene is rarely one trait
Some conditions follow a single gene closely enough that a family can name it: HBB and sickle cell disease, CFTR and cystic fibrosis. Those examples are famous because they are unusually clean, not because they are typical.
Height, blood pressure, most common disease risk, and almost every personality headline are polygenic. Many genes, each with a small effect, plus environment. Inheritance still happens. Mendel's pea ratios do not.
A gene can also be necessary without being sufficient. Two people can carry the same alleles and only one shows the trait. That gap has a name, incomplete penetrance, and it is why a gene is not a sentence.
Looking up a gene symbol is a public record, not a genome
CFTR, BRCA1, APOE, HFE: those are public names. NCBI Gene, MedlinePlus Genetics, and ClinVar keep pages for them. You can read what is known about the locus without sending anyone your sequence.
genome.sh will look up a gene symbol or an rsID as a public identifier. That is not a reading of your genome and not a diagnosis. If you have a file, matching the identifier to the two letters you carry is a separate, private step.
A gene page will list transcripts, locations on GRCh38, and variants people have submitted. It will not tell you which edition you carry. For that you need an assay, a file, or a clinic, and even then the trait may not follow.
Questions
How many genes do humans have?
About 20,000 protein-coding genes, not 100,000. There are additional genes whose product is RNA rather than protein. Noncoding RNA genes push the broader count higher, depending on how aggressively annotators include them. The integer moves a little. It does not return to the old 100,000 guess.
What does a gene do?
A cell transcribes it into RNA. That RNA may be translated into a protein, or the RNA may be the product. Nearby sequence also controls when the gene is read. A gene is work a cell can do with a stretch of DNA.
Do all genes code for proteins?
No. Many genes make RNA that is never translated. Even protein-coding genes contain introns and regulatory sequence that do not appear in the protein.
What is an allele?
A version of a gene, or of a site in a gene. You typically carry two alleles at a locus, one from each parent. They may be identical or they may differ.
What is the difference between homozygous and heterozygous?
Homozygous means both copies match. Heterozygous means they differ. The words describe a genotype at a site. They do not name a disease.
What is the difference between a gene and DNA?
DNA is the molecule. A gene is a used stretch of that molecule. Most DNA is not a gene. Genes live on chromosomes, which are DNA packed with protein.
Can you have extra copies of a gene?
Yes. Copy-number variation can delete or duplicate a stretch. Some genes, such as amylase genes, vary in copy number between people. Extra copies are still not a diagnosis by themselves.
Where can I look up a gene?
NCBI Gene, MedlinePlus Genetics, and NHGRI's glossary. genome.sh can query a gene symbol as a public identifier without uploading a genome. That lookup is informational, not a personal result.