But perhaps we shouldn’t take DNA evidence quite so seriously. This sounds like heresy. Since its invention, by Alec Jeffreys in the early 1980s, genetic fingerprinting has been hailed as the most brilliant breakthrough in forensics since actual fingerprinting. But an interesting investigation by New Scientist magazine earlier this month shows that this precision tool is often blunted by bias, mistakes and statistical misinterpretations.

This is important. Across the world, thousands of people have been convicted of murder, rape and lesser crimes thanks to 'DNA evidence' found at the scene – traces of genetic material derived from skin, blood or hair that are matched to genetic fingerprints taken from the suspects.

And, importantly, thousands of people have been acquitted; indeed the very first use of DNA evidence in an investigation was to prove that the main suspect in a gruesome double-murder case could not have done it. The very mention of 'DNA' in a courtroom makes jurors and judges swoon.

Perhaps they should be a little more level-headed. Looking at real evidence collected from actual crime scenes in the US, the study found that experts were often unable to agree on how to interpret the evidence and that the mathematical 'precision' of genetic evidence melts under close scrutiny.

The problem lies not with the technique itself (which does indeed promise extraordinary precision) but with the way this evidence is gathered and used by all-too-fallible policemen and forensic scientists. For a start, compared to the early days, far smaller samples are deemed to be suitable for forensic analysis. Since the technique of ‘DNA amplification’ was invented, its application to genetic fingerprinting means that as few as 150 cells – a microscopic amount - can be enough to provide a sample for identification.

That sounds good, but it introduces room for error.  “Labs are trying to results from lower and lower amounts of DNA,” John Butler, a US scientist who is trying to improve standards of DNA testing told NS. And few samples contain DNA from just one person. Teasing out which DNA belongs to whom is a highly skilled job and worryingly subjective.

One study, carried out in Ohio, found that in samples containing DNA from four people more than 70pc were mistaken for samples containing DNA from two or three people. This sounds like an arcane point, but it could make all the difference in, say, a gang rape case or a gangland killing where several people may have handled the murder weapon, the body or the ammunition but with only one having pulled the trigger.
The New Scientist report looked at the case of an actual gang rape committed in the state of Georgia, after which a man named Kerry Robinson was convicted – partly on the basis of DNA evidence.

When the evidence plus an actual sample of DNA taken from Robinson was presented to 17 experienced analysts, and a whole range of ‘conclusions’ resulted, ranging from four of the analysts saying that the evidence was wholly inconclusive and 12 saying he could actually be excluded.

The analysts looked at the data ‘blind’, unlike those asked to testify in the original trial. Just being told that a particular sample comes from ‘a suspect’ seems to make it more likely that experts will deem it to be incriminating when asked to compare with samples taken at the crime scene. Even when investigators are being wholly objective, mathematical misunderstandings may pollute otherwise irrefutable evidence.

Take what is known as the ‘prosecutor’s fallacy’. If one DNA profile is held by just one person in a million, you might think that if the suspect has this profile then that is the odds against his innocence – i.e. one in a million.

But you’d be wrong. In fact, even assuming you are confining your search to the UK population (60 million people) there are 60 people, on average, who could be expected to provide a perfect match.  Thus, all else being equal, the chances of the suspect being innocent (measured purely on DNA evidence) is not one in a million but 59 in 60. This is the stuff of Bayesian Probability which has had a bumpy ride in the English courts despite being logically irrefutable (it was an elementary Bayesian mistake which led to the wrongful 1998 conviction of Sally Clark for the murder of her two boys, largely on the evidence of paediatrician Roy Meadow).

DNA evidence is pretty good for establishing whether a person was at a particular place, but it does not establish guilt. Say police find a DNA trace of a known local crook in a bank which has just been robbed. This proves no more than he was there (perhaps completely innocently) not that he robbed the bank (similarly, one of the main arguments against biometric ID cards is that they merely prove that you are the person the card says you are, rather than establishing your identity, which is something quite different).

After Madeleine McCann went missing in 2007, lurid claims were made in subsequent weeks that ‘DNA evidence’ garnered from a rental car showed that the missing girl’s body had been transported. This was a classic example of blinding with science, the nonsense of these claims being overshadowed by the bogus ‘scientificality’ of what was being said. Indeed, such is the almost magical aura of the very term ‘DNA’ that there is good evidence that juries, prosecutors and even expert witnesses are likely to leave their critical faculties outside the courtroom whenever this modern talisman is invoked.

None of this means that genetic fingerprinting is not a hugely useful tool. Like literal fingerprinting, it has helped nail thousands of villains, and cleared the names of thousands of innocents. This technology can be so effective that it seems like magic. But it is not magic, and the people who wield it are not magicians.

So be suspicious, especially if you are on a jury. None of this is to say DNA evidence is useless; compared to the polygraph, for instance (which is unbelievably dodgy) in the right hands it is a useful too. But, again, it is not magic, and no substitute for old fashioned detective work.