TO CATCH A KILLER

AT OJ SIMPSON'S first trial, the jury was treated to what must have seemed an interminable disquisition on the analysis of DNA samples. The sight of the meandering hieroglyphics on a marker board in that cluttered courtroom was a sorry image of how public perceptions of DNA forensic testing have changed. It was presented in its early days as a laser beam that could bore straight through the fog of crime scenes to the truth, backing up its findings with odds so astronomical that they were beyond even unreasonable doubt. Now it seems fiddly, obscure and equivocal.

Mention OJ to forensic scientists in this country, however, and they aren't fazed. Their doubts centred on the way the samples were collected and presented, rather than on the techniques themselves. And the techniques are maturing. The Forensic Science Service (FSS) took up DNA methods 10 years ago, and has raised several generations of them since. Its policy is to replace traditional forensic methods with DNA techniques where possible.

This implies two different goals. The first, now largely achieved, has been concerned with improving the efficiency of the initial wave of DNA tests. These are intended to match a sample to a name. They match bloodstains with suspects or bodies with missing persons.

In one case on which FSS personnel worked, bloodstains were linked to a missing person whose body remained unrecovered. But after a Cleveland doctor reported his wife missing, police inquiries revealed that he had a wealthy mistress, and that he had also previously owned a garden shed. The shed was located, and bloodstains recovered from gaps between the floorboards, which had been coated with two layers of creosote. FSS scientists constructed a theoretical DNA profile for the missing woman on the basis of blood samples they took from her relatives. It matched the profile they extracted from the stains, and the doctor was convicted. (The term "profile" is the proper one; the widely used expression "DNA fingerprinting" is misleading. DNA tests are based on examining small sections of the genome, not recording the entire pattern in the manner of a fingerprint.)

The second goal is to develop methods by which DNA profiling may be used to predict physical characteristics of possible suspects, in the same way that psychological profiling is used to predict what kind of person has committed a particular crime. This project is still in its infancy.

At present, forensic science requires two things of DNA tests. First, they must be reliable, giving consistent results which will not lead to endless courtroom wrangles between lawyers and expert witnesses. Ideally, the chance they will identify the wrong individual must be vanishingly small, but if not, then it must be possible to specify the likelihood of misidentification accurately. Second, forensic experts also need tests that are robust enough to make effective use of the evidence available, which may be poor. When DNA tests are conducted to determine whether a man is the father of a child, or whether would-be immigrants are related to people already resident in the country, blood samples of optimal size can be taken under laboratory conditions. Crime scenes may yield only tiny amounts of dried blood. The DNA from such sources is often badly degraded.

The first generation of DNA tests (known as multi-locus profiling, or MLP) required a bloodstain the size of a 10p coin. Nowadays, evidence can be extracted from bloodstains as small as three millimetres across, as well as from degraded samples. DNA can be recovered from semen samples more than 10 years old, or from semen not containing any sperm. The key is called the polymerase chain reaction, or PCR, and it amplifies DNA sequences by duplicating them millions of times.

The FSS introduced a PCR technique in 1992, but it was limited by its low power of discrimination. With MLP tests, the likelihood of two individuals having the same profile was one in millions. The first PCR method took the levels of likelihood back to between one in five and one in 50, probabilities similar to those obtained from conventional identification techniques based on blood groups.

Forensic scientists had one technique which was highly discriminating, but insensitive; and a second which was extremely sensitive, but not very discriminating. To create a test which was both sensitive and discriminating, they turned to a class of DNA sequences known as microsatellites, or short tandem repeats. DNA molecules contain stretches of varying lengths in which the pairs of bases that form the molecule go into repetitive sequences whose function, if any, is unclear. When the repeating unit is between one and six base pairs long, the pattern is classed as a short tandem repeat, or STR. Their shortness makes them more likely to survive intact as the sample degrades, and also makes them suitable for amplification by PCR. The variation in length of the sequences allows them to be used to distinguish individuals from each other.

The first version of the PCR-STR technique the FSS deployed is known as a quadruplex, because it examines DNA sequences at four different locations on the molecule simultaneously. It is a rapid process which can give completed profiles in 24 hours compared to several weeks for other types of tests. It is also a highly discriminating one, distinguishing between stretches of DNA that differ in length by a single base pair. The sequences are identified by the number of repeat units they comprise. Reduced to a string of numbers, a profile can be stored in computer memory. Thus the FSS has been able to establish the world's first national DNA database, which went into operation at the FSS headquarters in Birmingham two years ago. It now contains about 150,000 profiles, and a second unit has been established in London.

Calculating the likelihood that individuals share a profile requires comparison against a reference population. One of the criticisms levelled against DNA profiling is that it can give misleading results if comparisons are made across racial groups. A black suspect whose profile is compared against a white reference population may seem more likely to have committed a crime, because his profile is less likely to occur among whites than among blacks. To answer this criticism, the FBI compiled "Caucasian", "Black" and "Hispanic" reference populations. But Caucasians comprise subpopulations derived from regions stretching from India to Ireland; African-Americans are a varied mix of African and European stocks; and all Hispanics have in common is the Spanish language.

In this country, the meaningfulness of ethnic categories for STR analysis was examined by two FSS scientists, Peter Gill and Ian Evett. Americans think of themselves as divided principally into black, white and Hispanic; Britons think of our major ethnic categories as black, white and Asian. Gill and Evett followed this conventional perception, although it departs from the traditional schemes of racial taxonomy in which India is seen as the east end of the Caucasoid region. But the accentuation of difference worked for them: they reported that the genetic distances between the major groups were at least an order of magnitude larger than those between subgroups, within these larger divisions. Other researchers had found significant differences between closely related groups, such as Gujaratis and Pakistanis.

As well as background research of this type, FSS teams have been able to test STR techniques in a couple of set-piece demonstrations. One involved analysis of remains said to be those of the Romanovs, the Russian royal family executed by the Bolsheviks. Tests were also performed on mitochondrial DNA, which is originally present in ample quantities and so tends to survive in old remains. A reference sample was provided by Prince Philip, a grand-nephew of Tsarina Alexandra. The investigators considered that the STR analysis indicated a family group had been present in the grave; on the basis of a DNA sex test and the mitochondrial evidence, they concluded that the remains were "almost certainly" those of the Romanovs.

The major showcase for STR analysis arose from a rawer and more apocalyptic tragedy. Scores of members of the Branch Davidian cult died in the great conflagration that ended the Waco siege in Texas in 1993. Among them were a number of British citizens, recruited by the cult leader, David Koresh, from Seventh Day Adven-tist churches in Britain. It was appropriate for the FSS to take part in the identification of the bodies.

Koresh had amassed a large stockpile of munitions - more than a million rounds, according to the FBI, and dozens of gren-ades - which began to detonate as the fire took hold. The result was that the bodies were not just burned, but torn apart as well. In addition, many of the victims had been crowded into a small area of the compound. Before the forensic investigators could identify the remains, they had to sort out which parts belonged together.

Because of the risk of further explosions in the embers, three days elapsed before recovery workers could begin to remove the bodies, many of which eventually lay in the ruins for a week. Decay set in; and continued even after the remains had been transferred to the mortuary, since they were kept cold but not frozen. Under such conditions, the best places to look for usable DNA samples are in the bones and deep muscle tissues, which may act as protective packaging.

In their efforts to identify the corpses, the investigators exhausted the battery of traditional distinguishing features, such as dental records and only managed to identify 18 victims.

Sixty-one samples were supplied for DNA an-alysis, which was carried out in duplicate, at the FSS Wetherby Laboratory in Yorkshire, and at the Armed Forces Institute of Pathology in Washing-ton, DC. The results were in agreement. Full profiles were obtained from 50 samples, and partial ones from six. Positive identifications were made for 13 of the 14 British victims and for 21 others.

The task was made still more complex by the fact that many of the victims were closely related. Among those investigated was an entire family group: a heavily pregnant woman, her two other children, and her husband. Identifying her was relatively straightforward: half the bands in one DNA profile matched those found in a sample from the woman's mother, and the rest were present in her father's profile; and the body itself contained a foetus. The husband was identified in the same way, and there were two bodies of the same sexes and ages as the children. Half the bands in the children's profile matched bands in the woman's profile, but the rest did not correspond to those of her husband. A match was eventually obtained, with that of Koresh. He appeared to be the father of at least a dozen of the child victims, including the unborn one.

Writing up their reports of the Waco investigation, the FSS scientists were able to claim that the quadruplex test had been validated. A new version of the quadruplex test, the second generation multiplex test is now in use, based on seven rather than four loci, one of which is a sex test. Whereas the quadruplex's discrimination was of the order of one in 10,000, that of the second generation multiplex (SGM) is around one in 50 million.

Later this year, the FSS will introduce a third generation test. It is based on seven loci, one of which is already used in the SGM test, and its power is about one in 30 million. It is intended to make DNA profiling a "unique identifier", taking it to a level of confidence achieved in earlier years by fingerprinting. The idea is that the third generation test can be added to the second generation if required. It could be used, for example, to counter the "my brother did it" defence, where a defendant claims that the true culprit is a close relative. Together, the tests will give a discriminating power of one in 30 million multiplied by one in 50 million - a lot more than there are people in the world.

Further performance improvements will come from reducing the amount of manual work in the process. With identification techniques firmly founded upon the PCR-STR combination, and the database it has generated, the new conceptual developments will be in the area of investigative techniques. To explore these possibilities, the FSS maintains a programme of research into the genetics of physical traits. It is certainly an ambitious project. Nobody knows how many genes are involved in the determination of skin colour. In mice, which the FSS uses as models, the total is thought to be around 50. The figure for humans may be similar, though many of the genes may have only a minor influence.

The FSS researchers are interested in the colour of hair and eyes as well as of skin. In particular, red hair, because it is produced by a distinctive pathway in melanin metabolism, which appears to be relatively easy to investigate. More ambitiously, the FSS wants to look at the possibility of identifying genes that underlie facial characteristics. But it has no desire to go beyond appearances into ethically sensitive areas. Dr Dave Werrett, Director of Research and DNA Services, stresses that they absolutely do not want to collect any information which might have implications about an individual's health. The programme is simply intended to give the police an idea of what a suspect or a missing person might look like.

Dr Werrett says that FSS personnel made this decision themselves. But although they have done the right thing without being told to, that still leaves the question of who will make such decisions in the future - particularly if forensic science continues its progress into the marketplace.

Unless the onward march of behaviour genetics is halted, we will be faced with increasingly confident claims about genes conferring predispositions to one psychological characteristic or another. If these gain acceptance, it is possible to imagine a discipline of DNA-based forensic analysis in which physical profiling merges with psychological profiling. The emphasis would have moved from who an individual actually is, through what they might look like, to what they might be like. We may end up wishing DNA really was as dull as it looked in Judge Ito's courtroom.

! Marek Kohn's 'The Race Gallery: The Return of Racial Science' is published in paperback by Vintage pounds 7.99