How forensic DNA evidence actually works in court

A DNA report handed to a jury looks like certainty. It usually contains a sentence such as: the probability of obtaining this profile from someone unrelated to the defendant is one in a billion. Jurors hear that number as a verdict. It is nothing of the kind. It is a statement about how common a genetic profile is in the population, and the distance between that and "this person is guilty" is where wrongful convictions live.

Start with what the laboratory actually measures. Human DNA is largely identical between people, so forensic scientists ignore the shared parts and look at short tandem repeats — regions where a short sequence stutters and repeats a variable number of times. The number of repeats at each region differs between individuals, and by counting repeats at several regions you build a profile that is rare in combination even though each region on its own is not. Since 2014 the standard system in England and Wales has been DNA-17, which reads sixteen such regions plus a marker that indicates biological sex, superseding the older ten-region SGM Plus system. The sample is amplified using the polymerase chain reaction, which copies the target regions billions of times so that even a trace can be measured, then separated and read as a series of peaks.

From that profile the scientist calculates a random match probability: pick an unrelated person at random from the population, and this is the chance their profile would match too. For a full, single-source profile that figure is genuinely tiny, often quoted as one in a billion because the databases used to estimate the frequencies cannot reliably support anything rarer. The National DNA Database, launched in 1995 and the first of its kind in the world, holds several million profiles and is where a crime-scene profile is searched for a "hit". A database match is an investigative lead, not evidence in itself — it tells the police who to look at, and the evidential work starts afterwards.

Where the number gets misread

The most notorious error is not scientific but logical. The prosecutor's fallacy takes the random match probability — the chance of a match given innocence — and silently flips it into the chance of innocence given a match. Those are different quantities. A one-in-a-billion match probability does not mean a one-in-a-billion chance the defendant is innocent, because the match probability has to be weighed against every other piece of evidence and against how many people could plausibly have been the source. The Court of Appeal set out how this evidence should be explained in R v Doheny and Adams in 1996, after R v Deen exposed exactly this confusion, and directed that scientists give the frequency and leave the ultimate question to the jury. Courts now increasingly prefer the likelihood ratio: how much more probable the evidence is if the defendant left the sample than if someone else did. It is a comparison of two hypotheses, not a declaration of fact.

Then there is the physical reality of crime-scene samples, which rarely arrive as tidy single-source deposits. Mixtures — DNA from two, three or more contributors — are the norm on a door handle or a weapon, and untangling whose peaks are whose is genuinely hard. Laboratories now lean on probabilistic genotyping software such as STRmix to estimate a likelihood ratio for a mixed sample, but the output is only as sound as the assumptions fed in, and defence access to the underlying code has been a running argument.

Transfer is the quieter danger. DNA moves. You shed it on a handshake, and that person can carry it onto an object you never touched — secondary and tertiary transfer, repeatedly demonstrated in the literature. A match confirms your DNA is present; it says nothing about how or when it arrived. Low-template techniques that amplify a few cells make this worse, which is why the judge in the 2007 Omagh bombing trial, R v Hoey, was scathing about low copy number analysis.

Contamination is not hypothetical either. In 2011 Adam Scott was charged with a rape in Manchester after his DNA appeared on a swab; he had never left Plymouth, and the Forensic Science Service later admitted a plastic tray had been reused between cases. The statutory Forensic Science Regulator, whose powers were finally put on a legal footing by the Forensic Science Regulator Act 2021, exists partly because such failures keep happening. The genetics are extraordinarily powerful. What juries most need is a clear account of what the powerful number does not say.