Cold hits vs hard facts

The way forward for those wishing to gain prosecutions with DNA evidence now seemed clear.

Large reference databases must be built up as soon as possible to avoid the questions of statistical validity which bedevilled the early days of DNA evidence (such as in Green, Pantoja, Tran and Milat in NSW). Earlier practices, like taking profiles from Red Cross blood banks without the knowledge or consent of donors, would cease to be sustainable as more people learn of how much personal information can be extracted from these samples. Prisoners can easily be exploited to meet this objective however, so they are.

More polymorphic alleles must be added to the range of loci tested so that the product rule can be used to generate match odds of jury-boggling proportions.

The theories of subpopulation variation in allele distribution which had been developed by population geneticists in response to known ethnic variations in the distribution of blood groups must be discreditted to avoid complicating the already difficult explanations of match odds calculations.

Any attempt to quantify the possibility of laboratory error must be resisted to avoid providing statistical evidence which would undermine the astronomical match odds used in court.

Wherever possible the chance of a DNA match arising from a blood relative, or even an identical twin, must be ignored.

In meeting the objective of gathering large databases authorities in most jurisdiction have swamped DNA testing facilities with samples taken from prisoners. The resultant mass of data quickly suggested a new possibility for DNA prosecution - full database trawls in which the DNA of a huge number of individuals would be checked against that taken from crime scenes in an attempt to gain 'cold hits'; matches which might crack cases for which there isn't even a suspect.

However huge databases bring their own problems.

For instance, if a DNA test capable of distinguishing between 'unrelated' people with one million to one confidence was used to create a database of two million personal profiles from a population of 20 million potential suspects you could be pretty certain that most crime stain profiles run against it would produce at least one cold hit.

You could also be more than 90% certain that it would be the wrong cold hit. You could further expect that around 80% of the personal profiles on the database would match at least one other on record from a different person.

Problems like these had led the 1996 National Research Council publication "The Evaluation of Forensic Evidence" (NRC-II) to recommend that "When the suspect is found by a search of a DNA database, the random match probability should be multiplied by N, the number of persons in the database".

Director of the Virginia Division of Forensic Science and Chair of the Laboratory Funding Working Group of the National Commission on the Future of DNA Evidence, Dr Paul Ferrara, was quick to deal with the problems the NRC-II recommedation caused him, with his database of 150,000 prisoner and arrestee samples which he regularly trawls for hits. In his 1997 article "DNA databanking - the next step", he tells us

He concludes his essay optimistically

Dream on Dr Ferrara.

Dr John Buckleton told the Commission that there are at least 10 cases of different individuals sharing the same DNA profile recorded on ESR's database of 10,907 6-loci profiles. In two of these cases the people sharing profiles had no known family relationship.

This problem is not likely to afflict Dr Amanda Sozer of Fairfax Identity Labs. In her evidence to the Commission she says

Her solution is to assume that the identifying data is in error and record duplicates as a single individual.

By February 2001 UK police had amassed a database of over 1 million personal profiles along with almost 100,000 crime scene profiles. This database is constantly expanding and police report hundreds of matches every month, using 6-point testing which claims match odds of about 37 million to one. This is to be expected.

What is also to be expected from a thorough trawl of this mammoth database would be around 100 billion attempts to match crime scenes with potential suspects, resulting in over two thousand false cold hits. More if the labs are capable of making mistakes.

The first false cold hit from a database trawl to be recognised in the UK was in 1999, though it did not become public knowledge until the following year after a UK forensic scientist addressed the USDOJ Commission on the Future of DNA Evidence.

A man with advanced Parkinsons disease who could not drive an automobile or dress himself unaided was linked to a burglary which had occured 200 miles from his home. In spite of protestations of innocence and alibi evidence police arrested him because the DNA profiles matched and 'so it had to be him'. It was several months before10-point DNA tests were done on samples from the suspect and the crime scene. The results exonerated him.

He gained his freedom and a brief note from the prosecutor saying that charges were being dropped because "there was not enough evidence to provide a realistic chance of conviction". He still awaits an official apology. Or even an admission of error.

In 1996 five different commercial premises in Birmingham were robbed. Police noted several similarities between the robberies including the manner of entry to the premises, the sophistication of the safe-cracking methods employed and the lack of fingerprints left behind. Oddly, the police recovered several cigarette butts from the crime scenes which they theorise were dropped by an offender who smoked while the safes were being opened. The use of forensic DNA in such cases had received prominent coverage in the British media for several years prior to the robberies.

In December 1998 Mr Robert Watters, who lived near the scene of the robberies, was arrested in relation to unconnected offences and DNA tested. Although the judge in this case instructed the jury to aquit, the profile obtained from his sample was run against the UK database and matched with DNA which had been recovered from the cigarette butts. Mr Watters was arrested, charged and convicted.

During the trial FSS scientist Valerie Tomlinson told the court that the probability of a false match between Watters and the butts was 1 in 86 million. Under cross examination she conceded that there was one chance in 267 that the DNA on the cigarette butts could have come from one of Watters' two brothers.

The prosecution led no other evidence which tended to exclude the defendant's brothers as contributors of the DNA, in fact one of them had previously been questioned by the police in relation to the robberies. The case against Watters consisted of the DNA match, that Watters had bought cigarettes (i.e. he may be a smoker), his proximity to the scenes of the robberies and the fact that like most safe crackers, Watters is male. The trial judge also seems to have drawn inference from the fact that Watters had declined to provide the police with his brothers' home addresses, although he instructed the jury against engaging in such 'dangerous speculations'.

Watters' conviction was quashed in October 2000 when an appeal panel found that the prosecution evidence did not constitute a prima facie case which could be safely left to a jury. During the appeal, reference was made to another UK case in which a man who had been linked to an offence by database trawl was aquitted when it was discovered that someone with an identical DNA profile lived closer to the crime scene than he.

Trawling the massive UK database in the attempt to link unsuspected (and often unsuspecting) people to unsolved crimes raises many questions about mass DNA testing and databasing.

Perhaps the most pertinent is: What has happened to the subjects of the many hundreds of false cold hits which UK authorities have as yet failed to recognise?