Could DNA testing, widely criticised in the US, help us to find Naomi Smith's killer? Daniel Jeffreys investigates
"This is where we refrigerate samples," says Dr Kardon, the director of MediGene, a forensic toxicology lab in New York. Pointing to a chest freezer, Dr Kardon says that business is booming. "We do DNA tests on behalf of various police departments. The test can condemn you or set you free. It can be enormously powerful."

That's certainly what Detective Superintendent Tony Bayliss, who is leading the Naomi Smith investigation, believes. To find out who killed 15-year- old Naomi, Mr Bayliss wants to take DNA samples from 800 young males aged 15 to 28. The tests will eliminate suspects and may force the killer to reveal himself.

In the US, DNA testing has been in use since 1985. With experience, lawyers and scientists have identified many of the pitfalls associated with the tests. American lawyers say a mass testing such as the one Bayliss wants would be unthinkable in the US, in part because the tests are now treated with growing suspicion.

"You put garbage in and you get garbage out," says Peter Neufeld, one of two New York attorneys who helped free O.J. Simpson. He said the same thing in the courtroom and won a rebuke from Judge Ito - and an acquittal for his client.

Neufeld continues: "There was a mismatch between OJ's samples and the evidence because of sloppiness in the tests. Far too many private toxicology labs are under enormous pressure from thousands of police departments. Some cut corners."

Most of us are now familiar with the supermarket-style "bar codes" that represent a suspect's DNA. A lab will take four or five segments of DNA and compare them to samples from a crime scene. The samples can come from blood, skin or semen. If the samples don't match, the suspect is excluded. If they do, the suspect is tied to the crime scene although not necessarily to the crime.

"Sounds easy," says Dr Jerry Coyne, DNA expert at the University of Chicago. "Well, it ain't so. It's easier to exclude than include somebody in a list of suspects and the two processes involve different methods with varying levels of quality."

DNA can be tested by Restriction Fragment-Length Polymorphisms (RFLP) or with a Polymerase Chain Reaction (PCR). The key difference is that RFLP tests are more rigorous and are usually used to suggest a match between a suspect and a forensic sample. The more general PCR tests are only reliable when used to exclude a suspect. Problems arise when PCR tests are put to use against suspects on trial.

"There are people who want PCR tests excluded from prosecutions if that's all there is," says Dr Coyne. "In the OJ case, it wouldn't have made any difference. They did a PCR which did not exclude him and an RFLP test which seemed to place him at the scene. The problem was the testing looked sloppy. Hence reasonable doubt."

Peter Neufeld says the problem is a mismatch between technological advance and good practice. "Juries are beginning to expect DNA evidence in every kind of trial, from murders to burglary. It's getting difficult to get a conviction without some DNA evidence tying the perpetrator to the scene. Juries will say, "C'mon! The guy must have left behind some DNA." As if they could tell DNA from doughnuts!"

Neufeld believes there must be federal regulation of all forensic toxicology labs. "There is a rising tide of cases that use DNA evidence and US labs are just not good enough to cope. The government must act fast before DNA is discredited."

Peter Neufeld and his client Tony Snyder would echo the sentiment. When Neufeld heard about Snyder he was serving 45 years in a Florida jail for aggravated rape. Neufeld used PCR tests to prove Snyder was not the man. "Certain sophisticated DNA tests were not available when Snyder was convicted. So we did the tests on him and that constituted new evidence," says Neufeld.

When a DNA match is made, scientists calculate a statistical probability of the suspect having been involved in the crime. Labs use phrases like, "There is a 1 in 23,000,000 chance that X was not the person who left the blood/semen/skin." Huge numbers - but they can be reduced to insignificance by laboratory errors.

"At private forensic labs, false tests were found to have a frequency of between 1 in 1000 and a staggering 1 in 100," says Dr Coyne. "Tell a jury that there's only a 1 in 50 million chance OJ was not the man but a 1 in 100 chance the lab screwed up and you have an acquittal."

Cellmark Diagnostics was one of the forensic laboratories involved in the O.J. Simpson case. Neufeld has criticised their practices but Cellmark is the only private lab to win full accreditation from the American Criminal Laboratory Society, an independent association of forensic experts. It has seen a threefold increase in requests for tests but does not see that as a risk. "We would never compromise our tests to speed up results, however fierce the pressure from a prosecutor's office," says Cellmark director Mark Stolorow. "We have complete power in the system over how fast we act."

Barry Scheck is the director of the Innocence Project at New York University. Scheck was the other DNA expert on the OJ team and his unit has growing worries about some types of testing. The project was set up to use DNA testing in cases where old-fashioned blood tests or eyewitness evidence may have imprisoned an innocent person. "We are seeing more cases where sloppy DNA testing or poor statistical analysis is suspected of producing a false conviction," says Jonas Kent, the assistant director of the project. "There have been instances where the tests have been done carelessly," William Thompson, a professor of forensic law at the University of California agrees. "If a distracted lab worker accidentally contaminates the suspect's DNA with DNA from the crime scene, the fingerprints could match, even though the suspect is innocent."

Luis Venegas was accused of rape in California, where the weaker PCR testing is most common. According to the prosecution, his tests showed that only one person out of 30,000 would have the same DNA. The defence used a different statistical method and showed that his DNA could be shared with 1 in 35. The jury sided with the defence. There are at least seven different ways of calculating matches within the US and they can produce wildly different results.

"Part of the problem is the gene pool," says Jonas Kent. "You have to collect genetic material from volunteers and convicted criminals to give basis for comparison. Some of these pools are very small. One set of tests at the FBI was based on a gene pool of just 20 FBI agents... Then there's data processing - we don't really have computers big and fast enough to do a lot of these tests properly."

Race is also an issue. "Certain patterns of DNA sequences are much more common among certain ethnic groups than in the public at large," says William Thompson. "This has to be reflected in the statistical tests or you can get false positives."

For all the problems, police departments across the US are committed to DNA testing. Tom Presley is the FBI's DNA chief, and helped to solve the bombing of New York's World Trade Centre. "The test procedures have got better; the statistics need working on, but the main weakness is evidence collection."

He cites the Simpson case: "It would have been safer to collect blood drops from the concrete by cutting out entire chunks of paving, rather than using a swab. Whenever you introduce a new element into the sample you risk contamination."

The US is now building its own national data bank of DNA evidence and labs like MediGene will get busier. At their testing bays, samples pass through as if on a conveyor belt. The lab has a first-class safety record, but it's not hard to see how quickly less diligent companies could get themselves and innocent people into trouble.

US lawyers and scientists disagree on how best to improve reliability but they all agree on one thing. The Naomi Smith mass test would be impossible in the US. "We have too many concerns about individual privacy here for that," says Peter Neufeld. "Besides, the OJ trial was a train wreck for DNA evidence - it will be a long while before another US District Attorney bases his case on so much DNA testing."