Safety of genetic evidence is put to test: Steve Connor reports on doubts over the use of DNA profiles in court

Six years after genetic fingerprinting was first used in a British court, there is grave concern that it is not the powerful identification tool it was initially said to be.

Defence lawyers in two recent court cases in the UK successfully challenged DNA evidence on the grounds that it could not be relied on to produce a safe conviction. One of the judges involved ruled that if scientists themselves could not agree on the value of genetic fingerprinting, it was unfair to expect a jury to do so.

Next week, forensic experts and leading geneticists will attempt to resolve the issue at a conference in London organised by the Metropolitan Police.

Some of the world's most eminent geneticists are astonished at the British court ruling. They believe anxieties - first expressed more than two years ago - should have been dispelled by now.

'It's the lawyers who keep the pot boiling,' Victor McKusick, professor of medical genetics at Johns Hopkins University in Baltimore, Maryland, said. 'It's their business to defend clients and anything they can say to cast doubt on genetic fingerprinting, they will. I think it's unjustified.'

Professor McKusick chaired a committee of American scientists who last year produced a report that should have answered the criticism. He said the report 'lays their concerns to rest'.

Although the committee, which was called together by the prestigious US National Academy of Sciences, confirmed the general reliability of the technique, the scientific consensus failed to prevent DNA evidence being thrown out by the courts.

Professor McKusick was mystified: 'I can't see why they find it unacceptable.'

Alec Jeffreys, professor of genetics at Leicester University and inventor of the technique, said the case against DNA fingerprinting was groundless. The dispute had led to a situation where 'scientific logic and legal logic get lost'.

The problems are perhaps the inevitable result of the technique's complexity. Defence lawyers have bombarded judges and juries with a welter of contradictory claims about scientific intricacies. Genetic fingerprinting works by comparing the DNA present in a tissue sample left at the scene of a crime - blood, semen, hair, skin or even saliva - with that of the defendant. Forensic scientists use a process of separating DNA fragments - dark bands in a gel - in an electric field. When the positions of two bands coincide, then they are said to match.

One of the arguments used by defence lawyers is that it is not always possible to agree on whether a 'match' is genuine. Even using the same scientific procedure on the same person can produce bands that end up in slightly different positions to each other.

However, Professor Jeffreys said the question of matching bands was not an issue because the interpretation of a match was biased in favour of a guilty defendant. But this has not stopped defence lawyers successfully challenging the forensic scientist's definition of a match.

More importantly, defence lawyers have raised the possibility that a 'match' between defendant and sample is the result of an innocent defendant and the guilty person sharing the same characteristics used in the analysis.

This is why DNA evidence is always accompanied by a statistical estimate of the probability of a chance match. There is more than one way of estimating this probability and different estimates can result in what appear to be huge discrepancies to a judge or jury. In one celebrated murder case in New York, the probabilities of a chance match were given as 1 in 500 using one statistical method, and 1 in 739 billion using another. This extreme example illustrates the difficulties of using a statistical procedure called the multiplication rule, whereby extremely small probabilities arise as a result of multiplying individual probabilities together.

The multiplication rule does not take sub-grouping of genetic traits - such as different races - into account because it assumes the genes in question are independent of each other. Defence lawyers have exploited this limitation.

Critics say there are subgroups of people - such as Scandinavians or Africans - who are likely to share the same genetic profile.

The argument's two chief proponents were Richard Lewontin, a geneticist at Harvard University, and Daniel Hartl, professor of biology at Harvard. They said genetic differences between races within a society mean that some genetic traits used in DNA profiling may not occur independently of each other.

Lewontin and Hartl were right in principle, but in practice it makes no difference, said Professor Jeffreys and others. For the genetic traits used in DNA profiling, there are no major differences between racial groups that could make the technique unsafe. In fact, genetic differences within races rather than between races are greater.

Professor McKusick's scientific committee concluded that even assuming there were differences between groups within the population, this could be taken into account by being ultra-conservative with the statistics, biasing the evidence in favour of the defendant.

Providing scientists follow this procedure and take due care in analysing tissue samples 'DNA typing should be able to provide virtually absolute individual identification', the committee said.

Professor Hartl accepted these conclusions and now endorses genetic fingerprinting. 'It's a terrific technique. It's so powerful at identifying people you don't have to fudge the data.' Even though probabilities will now have to be amended, so that some smaller estimates such as 1 in 100 million are changed to 1 in 10 million, they are 'still within the range that you can convict', he said.

Professor Lewontin, however, is sticking to his position, providing comfort for defence lawyers. 'I don't think it does change anything. I have the same problems I've always had,'he said.

Meanwhile, Professor Jeffreys is adamant that genetic fingerprinting is safe. 'It's here to stay.'