Can a Paternity Test Be Wrong? The Science of DNA Accuracy

DNA paternity tests are extraordinarily accurate when performed correctly, but they are not infallible. Modern testing typically produces probabilities of paternity above 99.99%, and the chance of a random unrelated man matching by coincidence is often quoted as less than one in a trillion. Yet wrong results do happen, and the causes range from mundane sample mix-ups in the lab to rare biological phenomena where a person genuinely carries more than one set of DNA. Understanding where errors creep in matters, because a paternity result can reshape families, legal obligations, and personal identities.

How Modern Paternity Tests Reach Such High Accuracy

Standard paternity testing works by comparing short, repetitive stretches of DNA at specific locations across the genome. These locations, called short tandem repeat (STR) loci, vary enough between unrelated people that checking a panel of them can distinguish almost anyone. A typical commercial test examines around 15 to 20 of these loci simultaneously using a technique called capillary electrophoresis. At each location, a child should share one version from their mother and one from their biological father. When the alleged father’s DNA matches at every tested location, the lab calculates a combined paternity index and converts it to a probability of paternity, usually well above 99.9%.

Exclusion works the other way around. If the alleged father’s DNA does not match at multiple locations, he is excluded as the biological father. The general rule in most labs is that mismatches at two or more independent loci are required before declaring an exclusion. A single mismatch alone is not enough, because DNA can change slightly from one generation to the next, and that possibility has to be accounted for before ruling someone out.

When Mutations Create False Mismatches

STR loci are inherently unstable compared to the rest of the genome. The repetitive structure makes them prone to gaining or losing a repeat unit when DNA is copied during the formation of sperm or egg cells. These germline mutations are a normal part of human biology, but they create a headache for paternity testing: a true biological father can show what looks like a mismatch at one locus, or occasionally even two, purely because a mutation happened during reproduction.

One published case illustrates the problem vividly. Out of nineteen STR and related loci examined, a putative father showed apparent exclusions at two of them. Yet the combined evidence from the remaining seventeen loci, plus additional testing, produced odds exceeding twenty-eight trillion to one in favor of paternity. With no alternative candidate father available, the laboratory concluded that two independent mutations had occurred and that the tested man was indeed the biological father. The lab subsequently changed its internal policy to require mismatches at three independent loci before declaring an exclusion.¹

Forensic geneticists have developed statistical models that account for the possibility of mutation when computing the likelihood ratio for paternity. Rather than treating any mismatch as automatic proof of non-paternity, these models weigh the probability that a mismatch arose from a mutation against the probability that the man is simply not the father.² A well-run lab applies these corrections routinely. A poorly run lab, or one using outdated interpretation guidelines, might call a true father an exclusion based on one or two mismatches without considering mutations at all.

Chimerism and the Person with Two DNA Profiles

Most people carry a single set of DNA in every cell. But a small number of individuals are chimeras, meaning their body is made up of cells from two genetically distinct cell lines. This can happen naturally when two fraternal-twin embryos fuse very early in development, or it can be acquired later in life through a bone marrow or stem cell transplant. Chimerism is one of the most dramatic ways a paternity test can go wrong, because the DNA collected from one tissue might genuinely differ from the DNA the person passed on to their child.

A striking case involved a man whose buccal (cheek) swab was compared with his child’s DNA in a gestational surrogacy situation. The initial result showed paternity exclusion at six STR loci, a result that would normally be considered conclusive. Further investigation revealed that the father was a tetragametic chimera. When his semen sample, the one actually used for IVF, was genotyped alongside hair follicles, nail clippings, and earwax, all tissues showed mixed profiles from two different cell lines, and all contained the paternal alleles matching the child. His blood, however, came from just one of those cell lines and did not match.³ Had no one thought to test additional tissue types, the true father would have been falsely excluded.

Even outside these natural chimeras, people who have received bone marrow or stem cell transplants can show mixed DNA in unexpected places. Blood samples from transplant recipients almost always show the donor’s DNA profile rather than the recipient’s. But the mixing does not stop there. Buccal swabs from recipients after transplant show highly variable ratios of donor to recipient DNA, ranging anywhere from zero to 100 percent donor, with many samples falling in the 10 to 30 percent range.⁴ The degree of contamination also increases with time since transplant. Recipients tested many years after their procedure showed substantially more donor DNA in their cheek cells than those tested within a year.⁵

The practical takeaway is that hair follicle samples consistently preserve the recipient’s original DNA profile even after transplantation, making them the preferred sample type for identity or kinship testing in anyone who has undergone a transplant.⁵ If you or the person being tested has ever had a transplant, that information needs to reach the testing lab before a sample is collected.

When Brothers, Uncles, or Other Relatives Are the Alternate Possibility

Paternity testing is built on the assumption that the alleged father is an unrelated man drawn from the general population. The math behind the paternity index depends on how common each DNA variant is in that population. When the true biological father is a close relative of the man being tested, the statistical power of the test drops sharply, because relatives share much of their DNA by descent.

One study created artificial paternity scenarios using 15-locus STR testing and found that among 125 child-uncle pairs, about 31 percent of cases showed zero, one, or two mismatches between the uncle and the child. In five of those pairs, there were no mismatches at all, and the calculated paternity probabilities for the uncle ranged as high as 99.997%.⁶ In other words, an uncle tested without including the mother’s DNA can look, by standard metrics, like a highly probable father. A separate simulation study confirmed that the false inclusion rate for a parent’s sibling posing as the parent was statistically comparable to a grandparent posing as the parent, and that the problem worsened when the biological parents were themselves related to each other.⁷

Including the mother’s sample helps considerably, because it lets the lab determine exactly which alleles the child inherited from the maternal side, leaving the paternal contribution more clearly visible. When the mother’s sample is unavailable, additional marker types beyond the standard STR panel, such as single nucleotide polymorphism (SNP) markers combined with STRs, can help distinguish a father from his brother.⁸ Anyone ordering a paternity test who suspects that a close relative of the tested man could be the biological father should mention that possibility upfront, so the lab can adjust its analysis accordingly.

The Identical Twin Problem

Identical twins present the ultimate challenge for DNA-based paternity testing. Because they originate from the same fertilized egg, they share essentially the same genome. Standard STR profiling cannot tell them apart. If one identical twin fathered a child, a conventional paternity test would give an equally strong positive result for his twin brother.

For years, this seemed like an unsolvable problem. Researchers eventually demonstrated that the solution lies in extremely rare mutations that arise after the embryo splits into two. By performing ultra-deep next-generation sequencing on sperm from both twins and blood from the child of one twin, investigators identified five unique mutations present in the true father and the child but absent in the uncle twin.⁹ Subsequent work confirmed the theoretical basis: because mutations continue to accumulate after the blastocyst splits, whole genome sequencing can in principle identify even a single post-zygotic mutation in the offspring that matches one twin and not the other.¹⁰

A parallel line of research has explored epigenetic differences, specifically DNA methylation patterns, as a way to distinguish identical twins without sequencing their entire genomes. Although twins start with the same DNA sequence, chemical modifications to that DNA accumulate differently over a lifetime. Studies have identified methylation markers in blood that differ between twins, and early forensic validation work has examined whether these differences hold up with trace-quality samples.¹¹ The technology is still not routine or widely available, but it represents a second front in the effort to resolve twin-related paternity disputes.

Laboratory Error and Sample Mix-Ups

For all the attention given to rare biological scenarios, the most common source of a wrong paternity result is far more mundane: human error. Samples can be mislabeled during collection. Tubes can be switched in the lab. Data can be entered into the wrong case file. Chain-of-custody procedures can be breached, allowing questions about whether the tested sample actually came from the person it is supposed to represent. Accredited laboratories maintain strict protocols to guard against these problems, but no system run by humans is immune to mistakes.¹²

Even when samples are correctly handled, the technical process of reading STR profiles is not perfectly automatic. Genotyping errors at the interpretation stage are real and surprisingly consequential. One analysis of microsatellite genotyping found that the most common errors involved misreading allele banding patterns, with over a third of those errors stemming from confusion between a sample that carries two identical copies of an allele and one that carries two slightly different copies. Simulations showed that even a per-allele error rate as low as one in a hundred could produce a false paternity exclusion rate exceeding 20 percent.¹³ That figure does not describe the error rate of accredited human paternity labs specifically, but it illustrates how sensitive paternity conclusions are to even small errors in allele calling.

What does this mean in practice? If you receive a surprising result, especially an exclusion you were not expecting, the single best step is to request a second independent test using a new sample. A reputable lab will have no objection to this. If a biological anomaly like chimerism or mutation is at play, retesting with a different tissue type or an expanded marker panel can reveal the true picture. If a sample handling error occurred, a clean retest will catch it immediately.

Prenatal Paternity Testing

Paternity can now be established before birth using a simple blood draw from the pregnant person. During pregnancy, fragments of fetal DNA circulate in the mother’s bloodstream. Non-invasive prenatal paternity testing (NIPAT) isolates this cell-free fetal DNA and compares it with the alleged father’s profile. The approach avoids the small but real risks of older invasive methods like amniocentesis.

Validation studies show high accuracy. In one study, the test correctly confirmed paternity in all twenty cases tested against the biological father. When tested against over 36,000 unrelated males, it correctly excluded paternity 99.95 percent of the time, with the remaining 0.05 percent yielding indeterminate results rather than false positives.¹⁴ A separate NIPAT approach using a panel of hundreds of SNP markers reported strong separation between true fathers and unrelated men across more than 900 validation samples.¹⁵

One limitation of some prenatal approaches involves Y-chromosome-based testing, which works only when the fetus is male. A study using Y-chromosome mini-STR markers and next-generation sequencing successfully recovered paternal haplotypes from all fourteen male fetal DNA samples in the study but detected no signal from the ten female pregnancies, as expected. In one of the fourteen male cases, a single-locus mutation reduced the calculated probability of paternity dramatically.¹⁶ Autosomal SNP-based approaches do not have this sex limitation and are generally preferred for prenatal paternity work.

How Next-Generation Sequencing Is Improving Results

Traditional STR analysis measures the length of DNA fragments but does not read their actual sequence. Two fragments of the same length can have different internal sequences, a phenomenon known as isoalleles. These hidden differences are invisible to conventional testing but can be detected by massively parallel sequencing (MPS), also called next-generation sequencing. MPS reads the actual base-by-base sequence of each STR, which means it can resolve ambiguities that length-based methods cannot.

In a study of 29 family trios, paternity index values generally increased when sequence-based data were used instead of length-based data. The technique also resolved allele inconsistencies that would have looked like mutations under conventional analysis, revealing whether a one-repeat difference was a genuine mutation or simply two different sequences that happened to have the same length.¹⁷ Separate work comparing next-generation sequencing with traditional capillary electrophoresis for Y-chromosome STR markers confirmed all previously observed mutations and identified new sequence-level variants between unrelated males that would not have been visible with older technology.¹⁸

In practical terms, MPS means fewer ambiguous cases, better discrimination between close relatives, and a clearer picture of whether a mismatch represents a true exclusion or an innocent mutation. The technology is still more expensive than conventional STR analysis and not yet standard in every commercial lab, but it is increasingly available through specialized forensic and paternity testing services.

When Degraded or Unusual Samples Are All That Remain

Paternity questions sometimes arise when a potential father is deceased and conventional sample collection is impossible. In forensic and legal contexts, laboratories may need to extract DNA from teeth, bone, preserved tissue, or other postmortem sources. Teeth are particularly valued because the hard outer structure protects the DNA inside the pulp chamber from decomposition and environmental damage.

A recent case report involving canine parentage testing demonstrated two independent DNA extraction methods applied to teeth, both of which successfully produced STR profiles sufficient to assess paternity. While the study involved dogs rather than humans, the same principles and extraction technologies apply to human forensic casework.¹⁹ The broader point is that DNA paternity testing is not limited to fresh cheek swabs and blood draws. When a case demands it, genetic material can be recovered from sources most people would never think to test.

Degraded samples do come with caveats. Partial profiles are more common when DNA is damaged, and a partial profile with fewer usable loci means lower statistical power. A paternity assessment based on eight usable loci is less definitive than one based on twenty. Labs experienced in forensic casework know how to interpret partial profiles honestly and will flag cases where the evidence is strong but not conclusive, rather than overstate what the data can support.