How much DNA do siblings share?
Full siblings, half-siblings, twins and what DNA tests report
Matic Broz, PhD8 min
Full biological siblings share about 50% of their DNA by descent on average, the figure the National Human Genome Research Institute gives when it says fraternal twins share half their genomes "just like any other siblings". Measured pairs scatter around that average. In a study of 4,401 sibling pairs, the mean was 49.8%, most pairs fell within a few percentage points of it, and the extremes were 37.4% and 61.7%.
Fraternal twins are ordinary full siblings genetically and share the same 50%. Half-siblings average 25%, three-quarter siblings 37.5%, and identical twins virtually 100%. None of these numbers conflicts with the 99.9% sequence identity shared by any two people, because they measure something different: stretches of DNA inherited from the same recent ancestor.
How much DNA do full siblings share?
Full siblings share an average of 50% of their autosomal DNA, the 22 pairs of non-sex chromosomes, by descent. The NHGRI glossary states this for fraternal twins and applies it to all brothers and sisters.[1] The figure comes from inheritance rules, and large studies confirm it.
"By descent" means that two people carry the same stretch of DNA because they both inherited it from a shared parent, not because it happens to look alike. At any position in the genome, full siblings have a 25% chance of inheriting the same copy from both parents, a 50% chance of sharing the copy from one parent, and a 25% chance of sharing neither. Averaged over the genome, that gives 25% × 2 + 50% × 1, or half of the two copies each person carries.[2]
Visscher and colleagues measured this directly in Australian twin families. Across 4,401 full-sibling pairs typed at hundreds to thousands of markers, the average share was 49.8%, with a standard deviation of 3.6 percentage points and a range from 37.4% to 61.7%.[3] That range describes one large sample, not biological limits; a larger sample would find a few pairs further out.
Fraternal twins develop from two eggs fertilized by two sperm, so their inheritance follows the same rules as siblings born years apart.[1]
Why do siblings not share exactly 50%?
Each child receives one set of 23 chromosomes from each parent, and every egg and sperm carries a different mixture of that parent's two copies. Before an egg or sperm forms, each chromosome pair exchanges segments through recombination, creating a new mosaic of the grandparents' DNA.[4] Two siblings draw independently from those mosaics, so they can match on more or less than half.
The spread is predictable. One Morgan of genetic map length corresponds to one expected crossover, and the human autosomes measure about 35 Morgans, so each parent passes on only a few dozen crossovers per child. Siblings therefore inherit long blocks rather than independent letters, and long blocks average out slowly. From that 35-Morgan length, Visscher and colleagues predicted a standard deviation of about 3.9 percentage points, close to the 3.6 they observed.[3]
The same randomness makes inheritance from each grandparent uneven. One sibling can carry noticeably more DNA from a particular grandparent than another sibling does, even though both average 25% from each.
How much DNA do twins, half-siblings and three-quarter siblings share?
Identical twins come from one fertilized egg and share virtually all inherited DNA. They are not perfectly identical: Jonsson and colleagues found that monozygotic twins differ on average by 5.2 mutations that arose early in development, a tiny fraction of a genome of about 3 billion base pairs.[5]
Half-siblings share one biological parent, so they inherit matching DNA through only one route and average 25%. 23andMe gives an expected range of 17% to 34%.[6]
Three-quarter siblings share one parent while their other parents are full siblings, for example when two brothers have children with the same woman. We calculate their expected share as the 25% of half-siblings plus the 12.5% that first cousins inherit through two sibling parents, or 37.5%.
Consumer DNA tests also report shared DNA in centimorgans (cM), a unit of genetic map length rather than physical length. The Shared cM Project, which pools matches submitted by test users of known relationship, gives the averages and 99th-percentile ranges most genealogists use.[7]
| Relationship | Expected share by descent | Observed or reported range | Shared cM, average (range) |
|---|---|---|---|
| Identical twins | About 100% | Differ by about 5 early mutations | Not listed |
| Full siblings, including fraternal twins | 50% | 37.4% to 61.7% (23andMe: 38% to 61%) | 2,613 (1,613 to 3,488) |
| Three-quarter siblings | 37.5% | Not reported | Not listed |
| Half-siblings | 25% | 17% to 34% | 1,759 (1,160 to 2,436) |
| Parent and child | 50% | Fixed at one full copy | 3,485 (2,376 to 3,720) |
Why does my DNA test show 38% or 2,600 cM for my sibling?
The percentage depends on how a company counts the places where siblings match on both copies. Full siblings share half-identical regions, where one of their two copies matches, over about 50% of the genome, and fully identical regions, where both copies match, over about 25%. Counting fully identical regions twice gives the textbook 50%. Counting each shared region once gives about 38%, which is what The Tech's geneticist found behind a MyHeritage result for a brother and sister.[8]
Totals in centimorgans follow the second convention. A region counts once whether one or both copies match, so full siblings show a match over about 75% of the genome's length. The Shared cM Project numbers bear this out: the full-sibling average of 2,613 cM is 75% of the 3,485 cM shared by a parent and child, who match on one copy almost everywhere.[7] Half-siblings rarely share both copies, so their 1,759 cM is about half the parent-child figure and their percentage reads close to 25% under either convention.
A full sibling near 2,600 cM and a half-sibling near 1,750 cM are therefore the normal pattern. The ranges overlap only at their extremes, which is where the next question begins.
Can a DNA test tell whether two people are half-siblings?
A DNA test can usually support a half-sibling relationship, but the total amount of shared DNA rarely identifies it on its own. Half-siblings, grandparents and grandchildren, and aunts or uncles with nieces or nephews all average about 25%. 23andMe notes that a 17% match could fit any of those relationships, or even a first cousin.[6]
Testing companies separate these cases using the number and length of shared segments, ages, and matches shared with other relatives.[6] Segment patterns carry real information. Simulating perfect knowledge of shared segments, Kruijver found that grandparents could be separated from half-siblings with over 99% accuracy, while half-siblings and uncle-nephew pairs were harder to tell apart.[2]
Maternal and paternal half-siblings also share their DNA in different patterns, though both average 25%. Eggs undergo more crossovers than sperm, so a mother's DNA reaches her children in more pieces. Simulated maternal half-siblings shared 51.9 segments on average against 37.1 for paternal half-siblings.[9] Mitochondrial DNA, passed down by mothers, can support a shared maternal line, and the Y chromosome can support a shared father for two brothers.
A total near 50% does not fit ordinary half-siblings. It falls far above 23andMe's 17% to 34% range and points to full siblings or a parent and child instead. Endogamy, pedigree collapse or relation through more than one family line can raise the total, and the Shared cM Project ranges do not account for them.[6][7]
Do siblings share 50% or 99.9% of their DNA?
Both figures are correct because they compare different things. The NHGRI fact sheet states that all human beings are 99.9% identical in their genetic makeup.[10] That figure lines up two genomes letter by letter and counts the positions where they differ, mostly the few million single-nucleotide variants each person carries. The same comparison puts humans and chimpanzees at about 98.8% identity.
The sibling 50% asks a narrower question: which stretches of DNA two people inherited from the same recent ancestor. Two unrelated strangers are almost identical in sequence yet share essentially nothing by descent within a genealogical time frame, while siblings share half by descent and are only marginally more similar letter by letter. One number measures common humanity, the other recent family, and a DNA test that finds relatives reads only the second.
